AMD Ryzen - New CPU Series that is cheaper but better then Intel Core I Series

[ZeTeX]

Better performance for cheaper, Intel now will have probably no other choice but to reduce the prices significantly or release new series that can compete.
But we still don't know if there is any problems in the Ryzen series, who knows, it might not explode like Samsung note but something else.

[BradC]

Don't get me wrong. I'm an AMD fan from way back and my main workhorse is still an FX-8350 from 2012. But..... let's wait for the reviews before we crown it king.

Yes, I'm as excited as the next bloke, and I've hovered over that pre-order shopping cart more than once in the last week or so, but having 2 phenoms, a bulldozer and a piledriver in the shed (the two later are still 24/7 machines), I want some hard proof before I drop another grand on more AMD kit.

[Augustus]

How's the power consumption of these? I don't care if it's slightly faster than Intels offerings if it runs hot like hell... 

[Ampera]

How's the power consumption of these? I don't care if it's slightly faster than Intels offerings if it runs hot like hell... 

Fairly low and good. They are a great series of chips for AMD. The socket is actually stupid though. PGA sucks, and it's REALLY dense.

My i7-4790k is still a budget performer, and it's gonna last me several years more.

[wraper]

How's the power consumption of these? I don't care if it's slightly faster than Intels offerings if it runs hot like hell... 

They consume less power than intel counterparts.

[bibz]

AMD's hype machine is rolling out but noone has any actual results from it. They've cherry picked the numbers, ramped up the taglines, but I haven't seen a single real world result. It's worrying imo that their banking on preorders pretty hard with a selected paper launch.

I do hope for great bang for buck though. Happy to have competition, Intel really doesn't see much progress from the CPU division unless you spend waayy big bickies. I think these Ryzens will be great for the multicore nerds. Unfortunately I'm still limited by the per thread clock speed of cpu's still.

[wraper]

PGA sucks, and it's REALLY dense.

Cannot agree, considering how often socket pins get bent on intel mobos.

[rollatorwieltje]

How's the power consumption of these? I don't care if it's slightly faster than Intels offerings if it runs hot like hell... 

direct comparison according to this article.

Ryzen 7 1800X: 95W, Core i7 6900K: 140W
Ryzen 7 1700X: 95W, Core i7 6800K: 140W
Ryzen 7 1700: 65W, Core i7 7700K: 91W

So the Ryzen appears to consume less power as well.

Still would like to see some real world results, especially single threaded performance. But it's good that there's finally some competition, CPU performance has stagnated for years now. My I5 2500 is still on the same chart when comparing it to more recent CPUs (in a similar price bracket of course), even though it's 5 years old. Most performance gains seem to have come from faster memory.

[Howardlong]

I agree, it needs to be out in the wild, not just a few leaks and limited invitation-only controlled tests.

Although there are a number of people with press kits, as far as I am aware they're currently locked into an NDA regarding performance.

Having said that, with the limited results available so far, if they turn out to be reliable, are a very good indication that'll have Intel on the backfoot regarding pricing, particularly on Broadwell-E, which has turned out to be quite a damp squib as the previous Haswell-E performs significantly better once overclocked.

Cooling a 95W TDP processor is going to be a lot easier than a 140W Broadwell-E.

[wraper]

Intel is already cutting prices on i7 and i5, so Ryzen should be very good, obviously.
http://www.tweaktown.com/news/56440/intel-rocked-core-over-ryzen-price-drops-begin/index.html

[MarkS]

LinusTechTips is known for having a strong pro-AMD bias. I'm interested in this architecture as well, but let's wait for actual real-world tests and not give biased videos weight.

[brucehoult]

How's the power consumption of these? I don't care if it's slightly faster than Intels offerings if it runs hot like hell... 

Fairly low and good. They are a great series of chips for AMD. The socket is actually stupid though. PGA sucks, and it's REALLY dense.

My i7-4790k is still a budget performer, and it's gonna last me several years more.

Ryzen doesn't have anything to touch the 4790K (or 6700K, or 7700K) if your task uses four threads or fewer.

If you don't have one of those then the top Ryzen will match your i7 for lightly-threaded tasks, and blow it away for things that can use all eight cores. And be cheaper.

The problem is most tasks *are* lightly threaded and four cores is usually plenty. Video compression/transcoding/filters is one of the few reasonably common non-benchmark tasks that can use more than four cores.

With the recent move to cmake and ninja, building software is starting to be able to use more cores effectively. Still, the common case is rebuilding only a couple of object files and linking ... and both running cmake initially and the linker at the end is where most of the time goes, and a single 4.5 GHz core is way more useful than 100 3.0 GHz ones.

[Howardlong]

It depends on your use case and workflow. If you're a gamer running only one application at a time then yes, there's generally little point going beyond four cores.

But not everyone's using their machine for gaming.

If you're a developer or content creator, there are reasons to benefit from increased cores count, partially because the software can benefit from those extra cores, but mostly because your workflow frequently includes performing multiple tasks simultaneously.

As a developer, it's also not at all uncommon to have VMs running, a case where multiple cores comes into its own.

In the past couple of weeks I updated my daily driver desktop from an Ivy Bridge I7 4C/8T to a 6800K after evaluating the 5820K, 6700K and mich more recently the 7700K. While I may yet put the 5820K back in (despite being older it overclocks far better than the 6800K), I found the 6700K/7700K still limited my workflow in the more extreme scenarios, typically when using multiple VMs.

I still would hold fire on Ryzen until it's had time to settle in. For example, it wasn't clear until some weeks after Broadwell-E was released that it was a crap overclocker compared the the previous Haswell-E, and that you're generally better off going for the Haswell-E series if overclcoking is your thing.

[thm_w]

Fairly low and good. They are a great series of chips for AMD. The socket is actually stupid though. PGA sucks, and it's REALLY dense.

Does anyone know specifics?
AMD used LGA for opteron because it was such a high pin count (1,900) and PGA for everything else.
But it seems to me the cost of producing PGA be lower overall, and the pin count is high for AM4 (1,334).

[Ampera]

PGA sucks, and it's REALLY dense.

Cannot agree, considering how often socket pins get bent on intel mobos.

Which would you rather replace, 100-300 USD motherboard, or 300-1500USD CPU?

[David Hess]

What about ECC support on socket AM4 processors?

[wraper]

What about ECC support on socket AM4 processors?

It's not certain. Gygabyte, for example, says that motherboards support unbuffered ECC RAM but it will work without ECC.

Support for ECC Un-buffered DIMM 1Rx8/2Rx8 memory modules (operate in non-ECC mode)

Asrock on the other hand says ECC RAM is supported (without mentioning if ECC is functional). AMD refuses to answer until release. So we need a few days to know for sure.
http://www.asrock.com/mb/AMD/X370%20Killer%20SLIac/#Specification

- Supports DDR4 2667/2400/2133 ECC & non-ECC un-buffered memory

[Ampera]

Does it support my 72 pin SIMMs? It's FPM, so it should be compatible.

[David Hess]

What about ECC support on socket AM4 processors?

It's not certain. Gygabyte, for example, says that motherboards support unbuffered ECC RAM but it will work without ECC.

Support for ECC Un-buffered DIMM 1Rx8/2Rx8 memory modules (operate in non-ECC mode)

Asrock on the other hand says ECC RAM is supported (without mentioning if ECC is functional). AMD refuses to answer until release. So we need a few days to know for sure.
http://www.asrock.com/mb/AMD/X370%20Killer%20SLIac/#Specification

- Supports DDR4 2667/2400/2133 ECC & non-ECC un-buffered memory

I read the AM4 ASROCK manuals yesterday and thay all say the same thing.  ECC and non-ECC DIMMs are supported but there is nothing about ECC other than that so I suspect ECC DIMMs will used without ECC functionality.  Over on RWT, a response argued that the manuals are incomplete since the DRAM settings are not mentioned either so we will have to see.

AMD has been making product segmentation statements which imply ECC will not be supported except on server sockets so I suspect my next system will be an Intel LGA1151 Intel i3 or LGA1150 Xeon.

[wraper]

I read the AM4 ASROCK manuals yesterday and thay all say the same thing.  ECC and non-ECC DIMMs are supported but there is nothing about ECC other than that so I suspect ECC DIMMs will used without ECC functionality.
...
AMD has been making product segmentation statements which imply ECC will not be supported except on server sockets

Previous AMD CPUs supported ECC (but not nearly all mobos), So I don't see why Ryzen shouldn't.

[BradC]

Which would you rather replace, 100-300 USD motherboard, or 300-1500USD CPU?

That's like saying "which car is cheaper to replace after you crash it, a Toyota or a Ferrari?". Unlike driving (where you can't account for the other idiots on the road), putting a CPU into a motherboard is one of those tasks where you can actually pay attention and apply due care to make sure you don't break your new expensive toy.

We all make mistakes. I dropped a brand new 1GB SCSI drive while getting out of a car (back when that was worth more than 2 weeks salary). I learned to be more careful! I've even bent a pin on a CPU (A Cyrix). It's a mistake you only make once. Hell, I de-lidded a 3770K with a vice and a hammer (I did a 3570K first but didn't video that one). Can't go through life saying "Oh, I'll buy *that* one as it's cheaper to replace when I break it". Plan not to break it instead!

[wraper]

Which would you rather replace, 100-300 USD motherboard, or 300-1500USD CPU?

1. Excuse me, but many intel CPUs cost the same or even less than mobo, down to 10x less than $300 you claim.
2. Pins on AMD CPUS are  rather stiff and not easy to bend, and if few are bent, very easy to straighten. On intel motherboard, you look on it funny, and pins are already bent. Very hard to straighten them if you don't have a microscope, sharp tweezers and straight hands.
3. If CPU pins are bent, you cannot insert it into the socket, nothing bad happens. If pins are bent on LGA socket, very likely they'll short to nearby pins, and pray god something doesn't burn on motherboard and/or destroy CPU.

[g.lewarne]

Whats to stop intel changing to quad core/8 thread low end and standardising 8 core/16 thread mid and high end?  their individual cores are still better than Ryzen and as I understand it, would obliterate them in one swoop.  Intel also have the manufacturing grunt to take a hit on profits and make them cheaper - now they actually have some competition

[wraper]

their individual cores are still better than Ryzen and as I understand it, would obliterate them in one swoop.

They are not. 8 core Ryzen beats 8 core Intel. Actually about the same IPC for single tread, for multithread AMD IPC is higher. Seems only higher clock speed on 4x and lower core count CPUs can save intel.

[g.lewarne]

their individual cores are still better than Ryzen and as I understand it, would obliterate them in one swoop.

They are not. 8 core Ryzen beats 8 core Intel. Actually about the same IPC for single tread, for multithread AMD IPC is higher. Seems only higher clock speed on 4x and lower core count CPUs can save intel.

ahhh, gotcha, I was mistaken.  That's really cool then, gives Intel the kick up the butt they need