AMD Reportedly In Advanced Talks To Buy Xilinx for Roughly $30 Billion

[EverydayMuffin]

https://www.tomshardware.com/news/amd-reportedly-in-advanced-talks-to-buy-xilinx-for-roughly-dollar30-billion

[MadTux]

Much better than if Nvidia or Apple would devour them.

[SMB784]

No surprise there, we'll see major consolidation in the chip maker market in the near future, with the big 3 (Intel+Altera, ARM+Nvidia, AMD+Xilinx) offering all 3 major types of chips (CPUs, GPUs, and FPGAs).

My bet is that this merger will happen, and then ARM+Nvidia will buy up what's left of the FPGA market (Lattice, etc)

[Sal Ammoniac]

 

[EverydayMuffin]

My bet is that this merger will happen, and then ARM+Nvidia will buy up what's left of the FPGA market (Lattice, etc)

Do you really think Arm+Nvidia would buy Lattice? Someone mentioned this on the FPGA subreddit also, I just don't see this happening at all. Why would Nvidia buy a low-end FPGA supplier? Lattice barely compete with Xilinx/Intel, they are going after very different markets. No?

[SMB784]

My bet is that this merger will happen, and then ARM+Nvidia will buy up what's left of the FPGA market (Lattice, etc)

Do you really think Arm+Nvidia would buy Lattice? Someone mentioned this on the FPGA subreddit also, I just don't see this happening at all. Why would Nvidia buy a low-end FPGA supplier? Lattice barely compete with Xilinx/Intel, they are going after very different markets. No?

I'm no industry insider, but my guess would be because it's a lot cheaper to just buy the wheel rather than reinventing it. Currently Nvidia has no FPGA offerings or IP that I know of. If they are interested in making a move into the FPGA end of the market it would be helpful to have a baseline to work from even if it's from a lower tier low power/embedded FPGA chip maker.  That might make sense anyway given that ARM already specializes in embedded low power processor design which might pair quite well with offerings from Lattice.

[ralphrmartin]

I this happens, and Nvidia buy up Arm, we will have the interesting state where AMD are selling chips with Nvidia cores... (Zynq with integrated ARM processors).

[SiliconWizard]

My bet is that this merger will happen, and then ARM+Nvidia will buy up what's left of the FPGA market (Lattice, etc)

Do you really think Arm+Nvidia would buy Lattice? Someone mentioned this on the FPGA subreddit also, I just don't see this happening at all. Why would Nvidia buy a low-end FPGA supplier? Lattice barely compete with Xilinx/Intel, they are going after very different markets. No?

I'm no industry insider, but my guess would be because it's a lot cheaper to just buy the wheel rather than reinventing it. Currently Nvidia has no FPGA offerings or IP that I know of. If they are interested in making a move into the FPGA end of the market it would be helpful to have a baseline to work from even if it's from a lower tier low power/embedded FPGA chip maker.

Not just cheaper. If they are interested as you mentioned, they are just instantly killing the corresponding competition if they buy them instead of starting afresh.

Given that Intel made that exact move with Altera a while ago, this is yet another simple reason why AMD would do just the same. AMD is a direct Intel competitor, and as such, they have little choice but expand their offering in the same way, otherwise they're bound to lose ground one way or another in the long run. This is pretty common in the industry. Major competitors tend to expand their offerings in the same way. Just my 2 cents.

[helius]

I this happens, and Nvidia buy up Arm, we will have the interesting state where AMD are selling chips with Nvidia cores... (Zynq with integrated ARM processors).

Altera/Intel also have ARM cores in the Cyclone/Arria/Stratix SoC series.

[coppice]

Do you really think Arm+Nvidia would buy Lattice? Someone mentioned this on the FPGA subreddit also, I just don't see this happening at all. Why would Nvidia buy a low-end FPGA supplier? Lattice barely compete with Xilinx/Intel, they are going after very different markets. No?

I'm no industry insider, but my guess would be because it's a lot cheaper to just buy the wheel rather than reinventing it. Currently Nvidia has no FPGA offerings or IP that I know of. If they are interested in making a move into the FPGA end of the market it would be helpful to have a baseline to work from even if it's from a lower tier low power/embedded FPGA chip maker.  That might make sense anyway given that ARM already specializes in embedded low power processor design which might pair quite well with offerings from Lattice.

I think you are looking at the name FPGA as though its a homogeneous market, but it isn't. Saying someone will buy Lattice because they can't buy Xilink is like saying they will buy an MCU maker because they can't buy AMD. They are both good businesses to be in, but with very little technological overlap.

[james_s]

I'm really not a fan of all this consolidation. Every time it happens a lot of parts get discontinued, a lot of people lose their jobs, and we are left with fewer choices and less competition. Maybe it's just part of the inevitable race to the bottom that is going to leave us with the Chinese companies owning the entire market.

[SMB784]

Do you really think Arm+Nvidia would buy Lattice? Someone mentioned this on the FPGA subreddit also, I just don't see this happening at all. Why would Nvidia buy a low-end FPGA supplier? Lattice barely compete with Xilinx/Intel, they are going after very different markets. No?

I'm no industry insider, but my guess would be because it's a lot cheaper to just buy the wheel rather than reinventing it. Currently Nvidia has no FPGA offerings or IP that I know of. If they are interested in making a move into the FPGA end of the market it would be helpful to have a baseline to work from even if it's from a lower tier low power/embedded FPGA chip maker.  That might make sense anyway given that ARM already specializes in embedded low power processor design which might pair quite well with offerings from Lattice.

I think you are looking at the name FPGA as though its a homogeneous market, but it isn't. Saying someone will buy Lattice because they can't buy Xilink is like saying they will buy an MCU maker because they can't buy AMD. They are both good businesses to be in, but with very little technological overlap.

From what I have read, Lattice ranks 3rd in the world when it comes to sales of FPGAs (and second when it comes to CPLDs), where their main competitors are Xilinx, Altera (now Intel), and Microsemi.  Furthermore, they specialize in low power FPGA solutions for the mobile market (e.g. their iCE line of FPGAs), which is in line with much of ARM's product portfolio.  If Nvidia wanted to make a move into the FPGA business (as its two main competitors AMD & Intel already have), they could do much worse than buying Lattice.

[colorado.rob]

From what I have read, Lattice ranks 3rd in the world when it comes to sales of FPGAs (and second when it comes to CPLDs), where their main competitors are Xilinx, Altera (now Intel), and Microsemi.  Furthermore, they specialize in low power FPGA solutions for the mobile market (e.g. their iCE line of FPGAs), which is in line with much of ARM's product portfolio.  If Nvidia wanted to make a move into the FPGA business (as its two main competitors AMD & Intel already have), they could do much worse than buying Lattice.

I really wish Lattice would jump on the SoC bandwagon.  The three competitors you mention all have them.  An ARM Cortex M4 or equivalent RISC-V paired with an ice40 and A9 with ECP5 would fill a real niche.

[coppice]

From what I have read, Lattice ranks 3rd in the world when it comes to sales of FPGAs (and second when it comes to CPLDs), where their main competitors are Xilinx, Altera (now Intel), and Microsemi.  Furthermore, they specialize in low power FPGA solutions for the mobile market (e.g. their iCE line of FPGAs), which is in line with much of ARM's product portfolio.  If Nvidia wanted to make a move into the FPGA business (as its two main competitors AMD & Intel already have), they could do much worse than buying Lattice.

I think if you ask someone at Altera or Xilink who their competitors are, I think they will just say the other name. If you ask someone at Lattice who their competitors are they are more likely to say Microsemi than Altera or Xilink. They are currently playing in different fields.

[Mr. Scram]

I this happens, and Nvidia buy up Arm, we will have the interesting state where AMD are selling chips with Nvidia cores... (Zynq with integrated ARM processors).

They already do in a major way. AMD's TrustZone is an ARM coprocessor. Many, many AMD processors are shipped with ARM cores on the same die. It's very similar to Intel's Management Engine in the sense that it's a discrete subsystem, with similar worries about the consequences for system security.

https://en.m.wikipedia.org/wiki/AMD_Platform_Security_Processor

[SMB784]

From what I have read, Lattice ranks 3rd in the world when it comes to sales of FPGAs (and second when it comes to CPLDs), where their main competitors are Xilinx, Altera (now Intel), and Microsemi.  Furthermore, they specialize in low power FPGA solutions for the mobile market (e.g. their iCE line of FPGAs), which is in line with much of ARM's product portfolio.  If Nvidia wanted to make a move into the FPGA business (as its two main competitors AMD & Intel already have), they could do much worse than buying Lattice.

I think if you ask someone at Altera or Xilink who their competitors are, I think they will just say the other name. If you ask someone at Lattice who their competitors are they are more likely to say Microsemi than Altera or Xilink. They are currently playing in different fields.

And you would probably be right about that.  However, Nvidia is not in a position to buy either Xilinx or Intel, so if they want to make a play in the FPGA market (which they very well might, given that their main competitors are and there are many advantages to doing so), they will need to acquire a company with significant experience in the FPGA market if they don't want to start from scratch.  There are only two left and Lattice is the bigger of the two.

[EverydayMuffin]

And you would probably be right about that.  However, Nvidia is not in a position to buy either Xilinx or Intel, so if they want to make a play in the FPGA market (which they very well might, given that their main competitors are and there are many advantages to doing so), they will need to acquire a company with significant experience in the FPGA market if they don't want to start from scratch.  There are only two left and Lattice is the bigger of the two.

Microchip is actually the bigger FPGA supplier out of Lattice and Microchip. They passed Lattice out at some point in the last couple of years, not sure when.

Source:

[SMB784]

And you would probably be right about that.  However, Nvidia is not in a position to buy either Xilinx or Intel, so if they want to make a play in the FPGA market (which they very well might, given that their main competitors are and there are many advantages to doing so), they will need to acquire a company with significant experience in the FPGA market if they don't want to start from scratch.  There are only two left and Lattice is the bigger of the two.

Microchip is actually the bigger FPGA supplier out of Lattice and Microchip. They passed Lattice out at some point in the last couple of years, not sure when.

It would appear my information is out of date!  So, perhaps they will consider buying Microchip.  Or maybe both, they have the money.

[tszaboo]

I this happens, and Nvidia buy up Arm, we will have the interesting state where AMD are selling chips with Nvidia cores... (Zynq with integrated ARM processors).

Altera/Intel also have ARM cores in the Cyclone/Arria/Stratix SoC series.

Yes. And when was the last time Intel released a new FPGA that would be remotely interesting? It is all "big data" and "reconfigurable server" and all these buzzwords. Or actually when was the last time intel released a part that us, as a electronics designer would consider using. Or from AMD, or from Qualcom, or from Marvell or Broadcom.

When these huge semiconductor companies buy smaller ones, because they have some parts that look remotely interesting... For each part they get a dozen other parts, that we might be using or considering in our design. When they "consolidate" we always get the short end of the stick.
Have you tried calling any of these companies, and ask them for a few thousand parts?

[coppice]

Yes. And when was the last time Intel released a new FPGA that would be remotely interesting? It is all "big data" and "reconfigurable server" and all these buzzwords.

I imagine the key buzzword for both Altera and Xilink right now is "5G", rather than "big data" or "reconfigurable server".

[SiliconWizard]

I'm really not a fan of all this consolidation. Every time it happens a lot of parts get discontinued, a lot of people lose their jobs, and we are left with fewer choices and less competition. Maybe it's just part of the inevitable race to the bottom that is going to leave us with the Chinese companies owning the entire market.

Agreed.

[Muxr]

At least it's AMD. They are growing at lightning speed and I doubt there will be much in a way of eliminating redundancies. Both AMD and Xilinx needed to invest more in ML and I think they can get where they need to be if they combine their efforts sooner.

This is about competing with both Intel and Nvidia.

Both companies need this.

[asmi]

My only worry is they will turn away from hobbyists and small businesses. Right now I can ask questions from Xilinx sales guys and they will give me an answer (or pass the question along to their FAEs) even if I didn't buy 1000000 of their chips. If I do the same to AMD right now - they will tell me to go do hell, or just ignore me.

[james_s]

I've never dealt with AMD before so I don't know what they're like. I had an Athlon PC around 20 years ago that was fantastic but then it seems like they didn't have much to offer for a while and I've been on Intel ever since. AMD was nearly irrelevant for a while and I thought they might go under, glad to hear they're growing though because Intel needs some competition.

[Mr. Scram]

I'm really not a fan of all this consolidation. Every time it happens a lot of parts get discontinued, a lot of people lose their jobs, and we are left with fewer choices and less competition. Maybe it's just part of the inevitable race to the bottom that is going to leave us with the Chinese companies owning the entire market.

The pace of consolidation in the electronics market has been frightening the past 5 or 10 years. While I think AMD would be a stronger company with its own FPGA branch, I don't think that condensing a diverse field of respected companies into a few behemoths serves anyone. It doesn't tend to be good for either company for various reasons, nor the customer or state of technology. I guess only the shareholders win. I can't help but feel that the free market and big money inevitably lead to a perverted version of a good thing with people losing out all round.

[coppice]

I'm really not a fan of all this consolidation. Every time it happens a lot of parts get discontinued, a lot of people lose their jobs, and we are left with fewer choices and less competition. Maybe it's just part of the inevitable race to the bottom that is going to leave us with the Chinese companies owning the entire market.

The pace of consolidation in the electronics market has been frightening the past 5 or 10 years. While I think AMD would be a stronger company with its own FPGA branch, I don't think that condensing a diverse field of respected companies into a few behemoths serves anyone. It doesn't tend to be good for either company for various reasons, nor the customer or state of technology. I guess only the shareholders win. I can't help but feel that the free market and big money inevitably lead to a perverted version of a good thing with people losing out all round.

You are going to have behemoths somewhere along the supply chain. The real consolidation has been in fabs, not component vendors. Fabs now need to be so big and expensive to be viable, that there are very few players there, and its hard to see how it could be otherwise. We still have numerous choices of supplier for most semiconductor devices, but the device vendors have very few choices when it comes to fabbing their silicon. When a company like TSMC has to close an old process, as no longer economically viable, its quite possible that every device of a certain type, from every device vendor, has to be withdrawn at the same time, as they were all made in the same fab.

[tszaboo]

I've never dealt with AMD before so I don't know what they're like. I had an Athlon PC around 20 years ago that was fantastic but then it seems like they didn't have much to offer for a while and I've been on Intel ever since. AMD was nearly irrelevant for a while and I thought they might go under, glad to hear they're growing though because Intel needs some competition.

I heared a story, about dealing with them. The company wanted a linux SBC, that time there was AMD geode. So they introduced the new selected part, then managed to obsolete the entire series, before they could fully assemble the first prototype batch. The parts were quite competitive and made sense back in ~2000.
Same with intel and their curie platform. They were targeting the Arduino folks, and microcontrollers with the product. It had like 512 MB of RAM or some other silliness only BGA package, and their datasheet was kept secret. Like, I dont even understand, how they could miss their target audience. From the same company that made the 8051, and sold billions of them. They released something that was not even in the ballpark of being usable.

[brucehoult]

When a company like TSMC has to close an old process, as no longer economically viable, its quite possible that every device of a certain type, from every device vendor, has to be withdrawn at the same time, as they were all made in the same fab.

Which is not going to happen if there are customers using the fab and willing to pay about the same for the dies as they always were, or a bit less as the factory is all paid off and depreciated.

If the equipment in it starts to fail and need a lot of expensive repairs that's a different thing, but I imagine it has a longer lifetime than people want to keep using an old process anyway.

[David Hess]

Given that Intel made that exact move with Altera a while ago, this is yet another simple reason why AMD would do just the same. AMD is a direct Intel competitor, and as such, they have little choice but expand their offering in the same way, otherwise they're bound to lose ground one way or another in the long run. This is pretty common in the industry. Major competitors tend to expand their offerings in the same way. Just my 2 cents.

Intel bought Altera as part of their effort to fill their foundries and provide foundry services.  AMD has no such incentive to do the same with Xilinx since they divested their foundries to GlobalFoundries and now use TSMC.

If Intel had a plan for CPU compatible FPGAs, it has not worked out.

[coppice]

When a company like TSMC has to close an old process, as no longer economically viable, its quite possible that every device of a certain type, from every device vendor, has to be withdrawn at the same time, as they were all made in the same fab.

Which is not going to happen if there are customers using the fab and willing to pay about the same for the dies as they always were, or a bit less as the factory is all paid off and depreciated.

If the equipment in it starts to fail and need a lot of expensive repairs that's a different thing, but I imagine it has a longer lifetime than people want to keep using an old process anyway.

Some people still want to be able to buy parts, especially analogue parts, that were launched 40 years ago. When a fab reaches its end of life, its seldom a sudden catastrophe. Most vendors are pretty good about planning for a shut down, notifying customers, and fulfilling lifetime buys of parts before the fab closes. If you don't order in time, that's your problem.

You must think TSMC is a magic company, immune from the pressures the rest of the industry faces. A lot of the mass removals of parts from vendor's catalogues that have occurred over the years have been due to a fab getting so old and creaky, the only production facility for the part had to close. You do know that TSMC is no longer running all the processes it has had since day one, right?

[David Hess]

Some people still want to be able to buy parts, especially analogue parts, that were launched 40 years ago. When a fab reaches its end of life, its seldom a sudden catastrophe. Most vendors are pretty good about planning for a shut down, notifying customers, and fulfilling lifetime buys of parts before the fab closes. If you don't order in time, that's your problem.

Fabrication facilities may also be closed because the raw materials becomes unavailable.

You must think TSMC is a magic company, immune from the pressures the rest of the industry faces. A lot of the mass removals of parts from vendor's catalogues that have occurred over the years have been due to a fab getting so old and creaky, the only production facility for the part had to close. You do know that TSMC is no longer running all the processes it has had since day one, right?

Unlike Intel, TSMC maintains several generations of older processes.

[brucehoult]

You must think TSMC is a magic company, immune from the pressures the rest of the industry faces. A lot of the mass removals of parts from vendor's catalogues that have occurred over the years have been due to a fab getting so old and creaky, the only production facility for the part had to close. You do know that TSMC is no longer running all the processes it has had since day one, right?

Ahahaha. Of course no.

But they're still doing 180, which debuted in 1998 I think.
And even 350 nm from ... what? 1994 or so? ... is still available.

I think 500 nm has been gone for a while though. And both the above are I think only on older 200mm wafers.

TSMC only started in 1987.

[Berni]

Its mostly digital fab processes that are marching on at such a quick pace.

The analog fab processes have much longer lifetimes since they don't really benefit from miniaturization as much since the transistor counts are much lower and dies are small anyway. So it makes sense to keep a lot of the older fabs running. However when a fab does get shut down its much harder to move to a new one since analog chips might use some special fine tuned steps to get the analog performance required. While transitioning a simple digital IC to a new process is a lot easier (Tho there are cases where the new process made the chips faster than the original, so this made some of the products using them not work anymore as they needed the chip to be slow)

Going back to AMD. They are really kicking ass recently tho. The just released Zen 3 pretty much put the final nail in Intels coffin as AMD has caught up on the only thing that Intel still had an advantage in, the performance per cycle metric. Finally Intel has some motivation to actually do something ever since the 4th Gen i5/i7.

[coppice]

But they're still doing 180, which debuted in 1998 I think.
And even 350 nm from ... what? 1994 or so? ... is still available.

I think 500 nm has been gone for a while though. And both the above are I think only on older 200mm wafers.

I suspect 350nm will hang around for a long time. It has an interesting and useful position. Its the last step before the operating voltages get really low. 350nm devices typically run up to 3.6V throughout the chip. 250nm devices are usually limited to 2.5V, requiring split supplies and low density higher voltage transistors around the perimeter of the chip to be able to interface to a higher voltage external world. This means 250nm is not long for this world. Once you have to split the core and peripheral voltages you are usually better off going for a finer geometry. However, wafer costs for 350nm are now rising over time, so there will be an increasing detterence factor.

[SMB784]

But they're still doing 180, which debuted in 1998 I think.
And even 350 nm from ... what? 1994 or so? ... is still available.

I think 500 nm has been gone for a while though. And both the above are I think only on older 200mm wafers.

I suspect 350nm will hang around for a long time. It has an interesting and useful position. Its the last step before the operating voltages get really low. 350nm devices typically run up to 3.6V throughout the chip. 250nm devices are usually limited to 2.5V, requiring split supplies and low density higher voltage transistors around the perimeter of the chip to be able to interface to a higher voltage external world. This means 250nm is not long for this world. Once you have to split the core and peripheral voltages you are usually better off going for a finer geometry. However, wafer costs for 350nm are now rising over time, so there will be an increasing detterence factor.

Interesting observation. I guess my question is why is running things at 3.6V preferable than say 2.5 or 1.5 if you can get away with it?

[coppice]

I guess my question is why is running things at 3.6V preferable than say 2.5 or 1.5 if you can get away with it?

Low voltages outside the silicon can bring a lot of issues. They are normal in some situations, like PC motherboards, now. However, inside electrically noisy equipment, like household electrical appliances, they can be very troublesome. Deep in the heart of an IC they are usually fine, but if the chip needs split rails to allow for higher interface voltages, it becomes more complex. This overhead is no issue for complex ICs, which usually gain more from a fine geometry than they lose. The increased leakage in fine geometry processes can be an issue in ultra low power applications, but that's the main problem area for complex ICs. For simpler ICs the overhead due to split rail operation can be serious.

[ali_asadzadeh]

This would make the game very easy for Chinese FPGA manufacturers, they will dominate very soon.

[KaneTW]

For anyone that's not a giant, AMD will just tell you to sod off. I have/had a project based on x86 CPUs. AMD referred me to SBCs (which are not suitable at all), despite this project having 6 digit funding. I guess they don't care about you until you're in the 7-8 digits.

[coppice]

This would make the game very easy for Chinese FPGA manufacturers, they will dominate very soon.

This won't affect the existing Chinese FPGA manufacturers much. They are playing in a different, low end, market. What will be affecting FPGA development in China is the big 5G makers being cut off from American FPGAs. That must be spurring a lot of high end FPGA development in China right now.

[ralphrmartin]

When a company like TSMC has to close an old process, as no longer economically viable, its quite possible that every device of a certain type, from every device vendor, has to be withdrawn at the same time, as they were all made in the same fab.

Which is not going to happen if there are customers using the fab and willing to pay about the same for the dies as they always were, or a bit less as the factory is all paid off and depreciated.

It may not be as simple as that. They may need the space the line occupies for a new line, for example.

[SilverSolder]

This would make the game very easy for Chinese FPGA manufacturers, they will dominate very soon.

This won't affect the existing Chinese FPGA manufacturers much. They are playing in a different, low end, market. What will be affecting FPGA development in China is the big 5G makers being cut off from American FPGAs. That must be spurring a lot of high end FPGA development in China right now.

...which will eventually lead to competition in that space.  Own goal by the USA in my view, but what do I know...

[asmi]

...which will eventually lead to competition in that space.  Own goal by the USA in my view, but what do I know...

That's exactly how ITAR-free satellite market came to be. One would think people can learn at least on their own mistakes.

[SiliconWizard]

Given that Intel made that exact move with Altera a while ago, this is yet another simple reason why AMD would do just the same. AMD is a direct Intel competitor, and as such, they have little choice but expand their offering in the same way, otherwise they're bound to lose ground one way or another in the long run. This is pretty common in the industry. Major competitors tend to expand their offerings in the same way. Just my 2 cents.

Intel bought Altera as part of their effort to fill their foundries and provide foundry services.  AMD has no such incentive to do the same with Xilinx since they divested their foundries to GlobalFoundries and now use TSMC.

If Intel had a plan for CPU compatible FPGAs, it has not worked out.

I think you are strictly seeing it from a technical POV. And you seem to consider Intel strictly as a CPU vendor (with foundry services) not willing or able to expand their horizons. Intel already does more than just design and sell CPUS. Filling their foundries may have been an incentive, but the end result is much more than just that.

From a business POV, buying Altera means Intel is now also an FPGA vendor (with a significant market). The fact they would mix FPGA and CPU technologies could be interesting, but is almost irrelevant at this point. That means, at this point, that Intel has a bigger market share overall as a company. All big companies need to expand in that way in order to survive long term. Diversification is key.

Now regarding mixing technologies, that's probably not for tomorrow, but there could be some interesting opportunities. Intel is not ARM. I don't see - and don't think that would make sense - them, for instance, embedding one of their CPU cores in FPGAs, the way it has already happened with ARM cores. I doubt that would fit Intel's strategy or model. But I could imagine something similar, but IMO more appropriate for Intel's strategy: embedding some tightly-coupled FPGA in their future CPUs, to be used as reconfigurable accelerators for instance. The difference may seem subtle, except that this way they would sell powerful CPUs with some integrated FPGA fabric, and not FPGAs with some integrated and only modest CPU core.

[wraper]

But I could imagine something similar, but IMO more appropriate for Intel's strategy: embedding some tightly-coupled FPGA in their future CPUs, to be used as reconfigurable accelerators for instance. The difference may seem subtle, except that this way they would sell powerful CPUs with some integrated FPGA fabric, and not FPGAs with some integrated and only modest CPU core.

They already do that https://en.wikichip.org/wiki/intel/xeon_gold/6138p#:~:text=Xeon%20Gold%206138P%20is%20a,by%20Intel%20in%20early%2D2018.&text=This%20microprocessor%2C%20which%20operates%20at,channel%20DDR4%2D2666%20ECC%20memory.

[Dmeads]

https://seekingalpha.com/article/4378735-amd-and-xilinx-prize-is-versal-acap-not-fpgas

found this article: really good read.

Basically it says that AMD's real motivation for the 30Billion comes from Xilinx'x ACAP. The article states that the Xilinx core FPGA market has only risen 4.8% since 2013, which isnt worth the 30 billion dolllars.

It also says, despite the fact that FPGAs will replace GPUs for AI acceleration, "that market is still small but with strong growth potential," also not making the deal worth it to AMD (in terms of hardware acceleration).

The ACAP however is targeted towards the fastest growing and "next generation" semiconductor markets such as 5G, defense, autonomous driving assist, etc, which is why the article says AMD wants Xilinx.

The author guesses that Xilinx will reject the offer, due to AMD not having much to offer them.

Any thoughts?

[helius]

It simply looks like a more advanced SoC FPGA, like the Zynq from 10 years ago. Hardened I/O blocks is not new, although the "AI" vector computation units have not been available in FPGAs before (unless I missed something, which is likely).

[SilverSolder]

It simply looks like a more advanced SoC FPGA, like the Zynq from 10 years ago. Hardened I/O blocks is not new, although the "AI" vector computation units have not been available in FPGAs before (unless I missed something, which is likely).

ISTR something about Tesla using this kind of technology, in a custom chip?

[asmi]

Growing popularity of Nvidia's datacenter products tells me you don't need FPGA to accelerate AI/ML. AMD makes their own datacenter-related products too (based on GPU tech), so if anything I'd expect to see FPGA fabric integrated into datacenter "GPUs" rather than into CPUs.

[SiliconWizard]

But I could imagine something similar, but IMO more appropriate for Intel's strategy: embedding some tightly-coupled FPGA in their future CPUs, to be used as reconfigurable accelerators for instance. The difference may seem subtle, except that this way they would sell powerful CPUs with some integrated FPGA fabric, and not FPGAs with some integrated and only modest CPU core.

They already do that https://en.wikichip.org/wiki/intel/xeon_gold/6138p#:~:text=Xeon%20Gold%206138P%20is%20a,by%20Intel%20in%20early%2D2018.&text=This%20microprocessor%2C%20which%20operates%20at,channel%20DDR4%2D2666%20ECC%20memory.

I didn't know about this device yet. That confirms Intel's interest. This just looks like a first step though IMO: it's apparently just an existing FPGA die coupled to a CPU die. I expect something more integrated in the future, like FPGA fabric possibly directly on the same die and possibly more "tightly coupled".

[asmi]

I didn't know about this device yet. That confirms Intel's interest. This just looks like a first step though IMO: it's apparently just an existing FPGA die coupled to a CPU die. I expect something more integrated in the future, like FPGA fabric possibly directly on the same die and possibly more "tightly coupled".

It's more efficient to use separate dies on interposer, because you are not limited to using the same process for all subcomponents, and yields are generally better for smaller dies. And AMD's success in recent years has proven that SiP designs (which is what all their modern CPUs are) are a way to go.

[SiliconWizard]

I didn't know about this device yet. That confirms Intel's interest. This just looks like a first step though IMO: it's apparently just an existing FPGA die coupled to a CPU die. I expect something more integrated in the future, like FPGA fabric possibly directly on the same die and possibly more "tightly coupled".

It's more efficient to use separate dies on interposer, because you are not limited to using the same process for all subcomponents, and yields are generally better for smaller dies. And AMD's success in recent years has proven that SiP designs (which is what all their modern CPUs are) are a way to go.

It's more efficient in certain contexts. This is not a general recipe.

1. If both elements can't be implemented on the same process (for technical or cost reasons), you just don't have a choice.
2. For multi-core CPUs, that makes sense. You can develop dies with 4 or 8 cores each, and then interconnect several dies on interposer to make CPUs with a lot more cores, giving you a lot of flexibility. In somes cases it also helps with thermal management. It obviously also helps with yield. Throwing away a 32-core on a single huge die is way more expensive that throwing away a 4-core die. Typical interconnections for multi-core designs also lend themselves well to this.

OTOH, with separate dies, you're obviously limited in how much of the internals you are able to expose to die pads.
So integrating some FPGA (or we may more generally say, reconfigurable logic) deeply on the same die would certainly allow tighter integration. You could get access to a core's internals in ways impossible just through the external connections. I can think of a few interesting applications for this that would make the cores themselves reconfigurable to some degree, whereas just interconnecting an FPGA die to a CPU die makes it possible to use it as some kind of coprocessor (which already allows lots of nice stuff), but nothing else much.

[asmi]

So integrating some FPGA (or we may more generally say, reconfigurable logic) deeply on the same die would certainly allow tighter integration. You could get access to a core's internals in ways impossible just through the external connections. I can think of a few interesting applications for this that would make the cores themselves reconfigurable to some degree, whereas just interconnecting an FPGA die to a CPU die makes it possible to use it as some kind of coprocessor (which already allows lots of nice stuff), but nothing else much.

You will destroy all CPU performance going this way. There is a reason you can only get into 1GHz+ logic frequency with hard silicon. Unless by "reconfigurable" you mean few basic settings, but for that you don't need any PFGA, just a register with some config bits.
Having thousands of interconnect lines between dies on interposer is not a problem - look at HBM for example. And then there is a 3D stacking ("die-on-die"), Intel is pursuing this approach with their Foveros packaging, which allows even lower delays due to shorter lines. Xilinx has been using this approach in their SSI (stacked silicon interconnect) devices for a while now.
The time of giant monolithic dies is coming to an end. This generation of GPUs is probably the last one (or one-before-last) that uses a monolithic big-ass die.

[tszaboo]

https://seekingalpha.com/article/4378735-amd-and-xilinx-prize-is-versal-acap-not-fpgas

found this article: really good read.

Basically it says that AMD's real motivation for the 30Billion comes from Xilinx'x ACAP. The article states that the Xilinx core FPGA market has only risen 4.8% since 2013, which isnt worth the 30 billion dolllars.

It also says, despite the fact that FPGAs will replace GPUs for AI acceleration, "that market is still small but with strong growth potential," also not making the deal worth it to AMD (in terms of hardware acceleration).

The ACAP however is targeted towards the fastest growing and "next generation" semiconductor markets such as 5G, defense, autonomous driving assist, etc, which is why the article says AMD wants Xilinx.

The author guesses that Xilinx will reject the offer, due to AMD not having much to offer them.

Any thoughts?

They are making a huge mistake (30B to be exact). They bought ATI, hoping they will integrate the graphic cards into the CPU, and it is going to be stronger than everything and kumbaya. And they failed. They failed so hard, that it led to a decade of Intel dominance on the CPU market. Now they have a strong 2 years, and they are going to buy another company, to ruin their financials and to chase pipe dreams. What is wrong with them?
Google developed the TPU. It is an AI accelerator, very fast. Faster than an  FPGA, because it is custom built to do only that. Amazon has  their own AI accelerator chips.
FPGAs are not really good at doing anything. They are just good enough of doing many things. If there is money to be made somewhere, they will make an ASIC for it.
And if you just want to place a little bit of FPGA in your CPUs, then license the damn thing. You can do that, you dont need the entire company for that.

[coppice]

Google developed the TPU. It is an AI accelerator, very fast. Faster than an  FPGA, because it is custom built to do only that. Amazon has  their own AI accelerator chips.

Yep, those chips, and a few other's like Tesla's, seem to be where the action is right now with AI and ML acceleration. They aren't making silicon for the merchant market, though. Someone needs to.

FPGAs are not really good at doing anything. They are just good enough of doing many things. If there is money to be made somewhere, they will make an ASIC for it.

Actually FPGAs are really good at getting advanced DSP solutions into the market place. That's why cellular has been such an important market for them. Eventually, as specs stabilise, ASICs are typically made for each generation of cellular system. However, anyone committing to an ASIC too early ends up with useless junk as the specs drift. Its complex FPGAs that get each new generation off the ground.

[Berni]

So integrating some FPGA (or we may more generally say, reconfigurable logic) deeply on the same die would certainly allow tighter integration. You could get access to a core's internals in ways impossible just through the external connections. I can think of a few interesting applications for this that would make the cores themselves reconfigurable to some degree, whereas just interconnecting an FPGA die to a CPU die makes it possible to use it as some kind of coprocessor (which already allows lots of nice stuff), but nothing else much.

You will destroy all CPU performance going this way. There is a reason you can only get into 1GHz+ logic frequency with hard silicon. Unless by "reconfigurable" you mean few basic settings, but for that you don't need any PFGA, just a register with some config bits.
Having thousands of interconnect lines between dies on interposer is not a problem - look at HBM for example. And then there is a 3D stacking ("die-on-die"), Intel is pursuing this approach with their Foveros packaging, which allows even lower delays due to shorter lines. Xilinx has been using this approach in their SSI (stacked silicon interconnect) devices for a while now.
The time of giant monolithic dies is coming to an end. This generation of GPUs is probably the last one (or one-before-last) that uses a monolithic big-ass die.

They are actually already reconfigurable.

Modern Intel CPUs come with about 1 to 16 KB of microcode that is sort of the "firmware" for it. It does the job of configuring some of the larger CPU components and tells it how exactly to decode the complex x86 instructions into multiple smaller instructions that actually get executed. During manufacturing they can also load in a special selftest microcode that helps run tests.

They are making a huge mistake (30B to be exact). They bought ATI, hoping they will integrate the graphic cards into the CPU, and it is going to be stronger than everything and kumbaya. And they failed. They failed so hard, that it led to a decade of Intel dominance on the CPU market. Now they have a strong 2 years, and they are going to buy another company, to ruin their financials and to chase pipe dreams. What is wrong with them?
Google developed the TPU. It is an AI accelerator, very fast. Faster than an  FPGA, because it is custom built to do only that. Amazon has  their own AI accelerator chips.
FPGAs are not really good at doing anything. They are just good enough of doing many things. If there is money to be made somewhere, they will make an ASIC for it.
And if you just want to place a little bit of FPGA in your CPUs, then license the damn thing. You can do that, you dont need the entire company for that.

I would not call it a mistake of buying ATI.

It was a perfectly good product back when AMD chips still held on against Intel the APU offering of AMD Core+Radeon graphics on the same die. It was capable of offering decent gaming performance at a fraction of the price. The GPU core could simply reuse all the support circuitry already there for the CPU, so things like the Vcore supply and RAM. The extra cost of the GPU silicon area was not that high and since they owned ATI there was no royalties or supplier monopolies to contend with. This did result in beefyer Vcore required, more cooling and faster RAM to feed it all, but all of this was still significantly less added cost than a dedicated graphics card. So for a span of a few years this was the best bang for buck if you wanted a cost effective gaming PC.

The same APU chips also ended up being used in previous generation gaming consoles as the Xbox One (50 milion sold) and PS4 (110 milion sold) making them a large success. Fast forward to today the new console generation of Xbox and PS5 are also using AMDs offering of a Ryzen CPU + Radeon GPU on a single chip.

[BrianHG]

https://seekingalpha.com/article/4378735-amd-and-xilinx-prize-is-versal-acap-not-fpgas

found this article: really good read.

Basically it says that AMD's real motivation for the 30Billion comes from Xilinx'x ACAP. The article states that the Xilinx core FPGA market has only risen 4.8% since 2013, which isnt worth the 30 billion dolllars.

It also says, despite the fact that FPGAs will replace GPUs for AI acceleration, "that market is still small but with strong growth potential," also not making the deal worth it to AMD (in terms of hardware acceleration).

The ACAP however is targeted towards the fastest growing and "next generation" semiconductor markets such as 5G, defense, autonomous driving assist, etc, which is why the article says AMD wants Xilinx.

The author guesses that Xilinx will reject the offer, due to AMD not having much to offer them.

Any thoughts?

They are making a huge mistake (30B to be exact). They bought ATI, hoping they will integrate the graphic cards into the CPU, and it is going to be stronger than everything and kumbaya. And they failed. They failed so hard, that it led to a decade of Intel dominance on the CPU market. Now they have a strong 2 years, and they are going to buy another company, to ruin their financials and to chase pipe dreams. What is wrong with them?
Google developed the TPU. It is an AI accelerator, very fast. Faster than an  FPGA, because it is custom built to do only that. Amazon has  their own AI accelerator chips.
FPGAs are not really good at doing anything. They are just good enough of doing many things. If there is money to be made somewhere, they will make an ASIC for it.
And if you just want to place a little bit of FPGA in your CPUs, then license the damn thing. You can do that, you dont need the entire company for that.

It gets worse, you do not need to even license any FPGA unless you want a complete fabric.  The degree of which AMD already has a chunk of firmware gate programable re-configurability already built into their GPUs and CPUs.  They do not need more.  It is standard in such large designs to allow for bug work-arounds once discovered in the field so entire dies will not need re-designing.

On the other hand, 1 reason alone exists at this level.  AMD may be already licensing some core parts of their CPUs and GPUs from third party vendors.  (IE PCIX/USB/anything other)  If Xilinx has a suitable replacement core within their IP, nothing to do with the main FPGA fabric, say just the die transistor layout on the serdes and PLL on the IO pins, a purchase may work out in the books as a profit to switch over.  In this case, Xilinx has a chance of no longer advancing as AMD just got what they want to make their CPUs and GPU more profitable.

[filssavi]

The reason they MUST do something about FPGAs is that they are just about starting to be looked at seriously for the mainstream(ish) data enter market, where the new trend is to add as much smarts as possible to the NIC in order to offload as much as possible from high speed networking (100G and up), as these speeds put a significant load on the CPU, and I am not only talking about network stack level stuff but also potentially layer 7 processing. Here having a real FPGA inside the cpu would be a godsend.

You must remember that while the desktop segment (and gamers in particular) are very vocal, they represent a vanishingly small part of the company revenue, that is dominated by server chips (this is also true for intel and nvidia, not just AMD). So anything that gives you market share there is good

Also more and more compute heavy applications are moved to GPUs, making CPUs less and less relevant by the day, and in that segment nVidia has a chokehold on the market with CUDA, there is many times the CUDA software with respect to openCL, so to  gain market share there AMD would need a GPU with drastically higher performance with respect to nvidia, also keeping in mind that the openCL software ecosystem (third party tooling, libraries, etc) is nearly non existent.
Oppose this to FPGAs where Xilinx is actually the market leader both in term of market share, and especially in term of technology, on the high level synthesis, software on FPGA, field.

[Berni]

Well FPGAs are not quite the magic bullet.

Sure you can massively accelerate certain computationally light but data heavy algorithms, but writing software for FPGAs is still a pain that nobody outside of the digital electronics field wants to deal with. And the efficency of FPGA solutions is not that great, both in power and clockspeed. Yes i know there are fancy compilers that turn C code into HDL but you still have to keep the FPGAs internal workings in mind to be able to write fast and efficient code.

I personally think that the next step is a more I/O capable GPU-like architecture. Today GPUs with compute are taking over all the boring repetitive math by acting as a sort of miniature supercomputer of 1000s of cores that munch away at the data in parallel. They execute a small program much like a CPU and are programed in a similar way with the usual parallelism issues. Only issue they have is that data needs to be fed to them into memory before they can do anything with it.

You could just as well do something similar on say a network card. Give it 100s of tiny processing cores that are suited for the networking tasks, but have the architecture be designed around fast IO too. Then software can be loaded into the cores to do some rough first pass processing on the incoming network data. The functionality of each core can be quickly swapped out on the fly if suddenly the network traffic has more of one type of data packet, so more cores are needed towards that. The computing resources can easily be shared and split between multiple applications running on the host OS.

[brucehoult]

I personally think that the next step is a more I/O capable GPU-like architecture. Today GPUs with compute are taking over all the boring repetitive math by acting as a sort of miniature supercomputer of 1000s of cores that munch away at the data in parallel. They execute a small program much like a CPU and are programed in a similar way with the usual parallelism issues. Only issue they have is that data needs to be fed to them into memory before they can do anything with it.

And that's exactly the problem with GPUs.

Another big problem with GPUs is algorithms that alternate parallel and serial parts. Their serial execution speed sucks, and communication turn around latency and bandwidth between CPU and GPU sucks. This causes Amdahl's Law problems with the maximum speedup possible from parallelization.

Modern vector ISAs avoid both these problems. If you design them carefully (e.g. RISC-V Vector extension) then it's easy to compile GPU languages such as CUDA and OpenCL to them and run with high efficiency. When it makes sense you can put thousands of execution elements in your Vector unit, but exactly the same code will run perfectly on smaller Vector units.

ARM SVE is much the same. I haven't verified whether it has the correct things to handle CUDA-style programs (with things such as divergence and reconvergence). It's also a bit limited in terms of the number of execution elements possible, as it is limited to vector registers between 128 bits and 4096 bits in length -- that's only 128 32 bit ints or floats maximum. RVV's limit on vector register length is 2^32 (or maybe 2^31) and some people are seriously planning implementations with 4k or 16k or 64k processing elements.

You can also interleave parallel and serial processing at the level of individual instructions, and the vector unit can operate from the same L1 or L2 cache as the scalar CPU.

[Berni]

Nvidia tho is pushing this Magnum IO and GPUDirect thing lately. Where the GPU is supposed to directly talk to fast SSD storage and networking to make getting data in and out faster and less CPU intensive.

No idea how that actually works under the hood. They are pushing this tech into the general consumer market as competition to the next gen game consoles that will use similar functionality to load and decompress 3D models and textures from SSD directly into video RAM at ridiculously fast speeds.

Does that tech actually deliver on the promise? Since Nvidia has loved to rush out not yet ready tech out into the market just to be there first.

[asmi]

No idea how that actually works under the hood.

The mechanics of it is very simple - since both GPU and NVMe SSD are PCI Express devices, as long as they know each other's bus address, they can talk to each other directly. This feature is called "Bus mastering" and has been a possibility even back in "classic" PCI times. The only caveat with PCI Express is since they are only logically collected to the same bus, but not physically (as PCIE is a point-to-point link, not multidrop like "classic" PCI was), they will need to use a root port to route packets between endpoints. But that routing is happening inside the PCIE root port without any CPU involvement (well, not exactly since PCIE root complex is a part of CPU nowadays, but you get the point).

[tszaboo]

FPGAs are not really good at doing anything. They are just good enough of doing many things. If there is money to be made somewhere, they will make an ASIC for it.

Actually FPGAs are really good at getting advanced DSP solutions into the market place.

You know what I meant.

It gets worse, you do not need to even license any FPGA unless you want a complete fabric.  The degree of which AMD already has a chunk of firmware gate programable re-configurability already built into their GPUs and CPUs.  They do not need more.  It is standard in such large designs to allow for bug work-arounds once discovered in the field so entire dies will not need re-designing.

On the other hand, 1 reason alone exists at this level.  AMD may be already licensing some core parts of their CPUs and GPUs from third party vendors.  (IE PCIX/USB/anything other)  If Xilinx has a suitable replacement core within their IP, nothing to do with the main FPGA fabric, say just the die transistor layout on the serdes and PLL on the IO pins, a purchase may work out in the books as a profit to switch over.  In this case, Xilinx has a chance of no longer advancing as AMD just got what they want to make their CPUs and GPU more profitable.

They probably want to package the FPGA on the same interposer, like their CPUs. Chiplet design. The Chiplet design makes a lot of sense, and it was a big win for them, and they want to be able to place more stuff in one package. I get that.

I would not call it a mistake of buying ATI.

It was a perfectly good product back when AMD chips still held on against Intel the APU offering of AMD Core+Radeon graphics on the same die. It was capable of offering decent gaming performance at a fraction of the price.

I'm quite sure there were almost dozens of people who were happy that AMD made APUs. In the meantime their market share tanked, they had to sell their fabs, and beg for money from investors.

The reason they MUST do something about FPGAs is that they are just about starting to be looked at seriously for the mainstream(ish) data enter market, where the new trend is to add as much smarts as possible to the NIC in order to offload as much as possible from high speed networking (100G and up), as these speeds put a significant load on the CPU, and I am not only talking about network stack level stuff but also potentially layer 7 processing. Here having a real FPGA inside the cpu would be a godsend.

So they spent 30B to have an in house NIC? Like this?
https://www.broadcom.com/products/ethernet-connectivity/network-adapters/100gb-nic-ocp/p2100g

" combining a high-bandwidth Ethernet controller with a unique set of highly optimized hardware acceleration engines to enhance network performance and improve server efficiency."

[coppice]

FPGAs are not really good at doing anything. They are just good enough of doing many things. If there is money to be made somewhere, they will make an ASIC for it.

Actually FPGAs are really good at getting advanced DSP solutions into the market place.

You know what I meant.

I have no clear idea what you mean. I don't know you. I can't read prior knowledge I have of your character into anything you write. All I can work from is the wording of your messages. "FPGAs are not really good at doing anything" is a pretty clear statement. I can't tell whether you were actually thinking clearly when you worded that.

[filssavi]

So they spent 30B to have an in house NIC? Like this?
https://www.broadcom.com/products/ethernet-connectivity/network-adapters/100gb-nic-ocp/p2100g

" combining a high-bandwidth Ethernet controller with a unique set of highly optimized hardware acceleration engines to enhance network performance and improve server efficiency."

I did not really expand to much on the subject, however they will spend 30B to have this:

https://www.servethehome.com/what-is-a-dpu-a-data-processing-unit-quick-primer/

The core of these devices is still a high speed NIC, however it can do a little bit more, such as communicate with storage clusters on the network and emulate an nvme interface, so that the guest OSes on the main processor can treat it as local, they can offer high speed offloading of network processing (things like packet filtering, sorting, etc, on multiple 100/200 Gbit interfaces without occupying half of the main cpu time, etc.

as you see  in the article nVidia is already mooving in this space, thus AMD needs to move fast if they want a piece of this undoubtedly huge pie

[FenTiger]

This is going ahead: https://www.nytimes.com/2020/10/27/technology/amd-xilinx-35-billion-stock-deal.html

[Karel]

Indeed:

"AMD and Xilinx today announced they have entered into a definitive agreement for AMD to acquire Xilinx in an all-stock transaction valued at $35 billion."

https://ir.amd.com/news-events/press-releases/detail/977/amd-to-acquire-xilinx-creating-the-industrys-high

[SilverSolder]

They are all gearing up to compete with something very big...

[ali_asadzadeh]

Sad But true