EEVblog Electronics Community Forum

For the thread that is actually responsible for prepairing and feeding the processed frames over to the GPU, how much CPU time is it using. It is not hard to end up with a CPU bottleneck while having task manager report a low overall CPU usage. On a CPU with 16 threads, a single thread just needs to use a little over 6.25% overall utilization to be fully occupying the time of a single core.

This type of behavior is why you will see behavior like the CPU usage in fallout 4 dropping when you spawn a ton of NPCs, the AI is single threaded and the engine waits on that thread before doing more work.

Beyond that, other bottlenecks can arise if in your editor timeline, you have effects or adjustments that are not multi-threaded. Not sure how many of those limits sony vegas has, but adobe premiere, as well as after effects, have many effects and adjustment options that are computationally expensive while also being single threaded.

Those limits are at least the case when using the nvenc plugin for adobe premiere pro (though it has gotten really unstable as adobe premiere has been updated).

For the thread that is actually responsible for prepairing and feeding the processed frames over to the GPU, how much CPU time is it using. It is not hard to end up with a CPU bottleneck while having task manager report a low overall CPU usage. On a CPU with 16 threads, a single thread just needs to use a little over 6.25% overall utilization to be fully occupying the time of a single core.

This type of behavior is why you will see behavior like the CPU usage in fallout 4 dropping when you spawn a ton of NPCs, the AI is single threaded and the engine waits on that thread before doing more work.

Beyond that, other bottlenecks can arise if in your editor timeline, you have effects or adjustments that are not multi-threaded. Not sure how many of those limits sony vegas has, but adobe premiere, as well as after effects, have many effects and adjustment options that are computationally expensive while also being single threaded.

Those limits are at least the case when using the nvenc plugin for adobe premiere pro (though it has gotten really unstable as adobe premiere has been updated).

Yes, with filters/effects there is certainly a correlation with CPU clock, but for the final encode via NvENC there is little to no gain in having a faster CPU clock.

As for your assumption that the effects may not be multi-threaded, video processing (not encoding) is an extremely SMP friendly computing paradigm. You have thousands of concurrent jobs you can do at once, one for every frame. The only part of that process that must be ordered is the final encode as mpeg uses motion prediction. Most effects are performed on the GPU anyway using shaders, again bringing us back to synchronization spin locks.

Without actually profiling the application, it is impossible to determine if you have high CPU usage, or high CPU wait in spin locks, CPU load alone is no indication of overall system load or performance with this computing paradigm.

Quote from: EEVblog on June 26, 2018, 04:40:24 am

Quote from: Stefan Payne on June 26, 2018, 03:53:52 am

Yes and if you had a Plugin for your Forums that Allows people to thank other for Posting something, you don't need to have that Problem.

The forum does have that feature.

See, Problem solved.

Should use that next time you need to know something

I remember you Stefan, you are that person that decided to argue that my ATX motherboard was incompatible with my ATX standards compliant PSU. People like you are exactly why Dave doesn't consider "advice" from people like you reliable. Too much of it is conjecture, opinion or speculation as to how things work, and are not based on empirical evidence or understanding.

Quote from: gnif on June 26, 2018, 05:06:00 am

Quote from: Stefan Payne on June 26, 2018, 04:54:18 am

Quote from: EEVblog on June 26, 2018, 04:40:24 am

Quote from: Stefan Payne on June 26, 2018, 03:53:52 am

Yes and if you had a Plugin for your Forums that Allows people to thank other for Posting something, you don't need to have that Problem.

The forum does have that feature.

See, Problem solved.

Should use that next time you need to know something

I remember you Stefan, you are that person that decided to argue that my ATX motherboard was incompatible with my ATX standards compliant PSU. People like you are exactly why Dave doesn't consider "advice" from people like you reliable. Too much of it is conjecture, opinion or speculation as to how things work, and are not based on empirical evidence or understanding.

Yes because I have such a situation here. And it happens in some very rare occasions.

I have a perfectly fine Corsair HX750i that works with like half a dozen ATX Motherboards.
And I have a Gigabyte X79-UD5 that also works with half a dozen other ATX PSU.
But the Corsair HX750i does not work with the Gigabyte X79-UD5...

And its the same thing that people working in the industry say, that it can happen.
And replacing either of those component solves the Problem.
That's also what people working in the Industry say. Like jonnyGURU, he mentioned that from time to time, that a PSU just does not work with a  motherboard although both are perfectly fine and within the spec...

but you argued that a missing signal is just a "incompatibility". It wasn't just slightly out of spec, it was completely dead.

In your example, did you measure the PSU to verify why it was "incompatible"? Did you obtain another HX750i to verify it wasn't at fault? Did you perhaps think that the Gigabyte board might be slightly out of tolerance and was playing it safe by refusing to start with a possible faulty PSU? My point remains, you have no empirical evidence, just your assumptions.

But insulting me for something that you don't believe, is that really necessary??

I wouldn't have said a thing, but when you started telling Dave how to use his forum, and your condescending attitude I felt it was time to speak up.

Dave really called it.. so many people with their nickers in a twist about a damn computer.

Because there is so much Misinformation, Bullshit, Lies and Fairy tales in this area you can write books about that...
And now a highly regarded electronics person does all the things those People are fighting against...
That's one of the reasons people are posting here right now...

Epecially in the PSU area. Like the lie of "Single Rail is better, Multi Rail is garbage*", bigger fans equal lower noise (no it doesn't!), Efficiency equals quality (well, there is a bit truth to that -> there are no good quality 80plus Bronze or worse PSU right now. The best bronze ones are just somewhat OK right now because Gold is like 2,5€ more in manufacturing), that the manufacturer is the most important thing...

Or stuff like you need a 750W PSU for an office PC (well, in ATX Formfactor, there are no really good PSU under 400W), should be loaded only up to 50% (when the good PSU are rated at continuous power at higher ambient temperatures, usually 40°C, sometimes 50°C. The Cheaper ones are only rated for like 30-35°C Ambient).

When we come to Graphics Cards and CPUs, it doesn't get any better.

The worst thing I've seen is that people believe that nVidia Cards are for Intel and don't work well in AMD Systems or vice versa with AMD Cards -> people believe that they only work well in AMD Systems...

*That is essentially Multi Rail in a PSU:
(Pic of a 14.95€ PSU )
A Shunt Resistor or a coil per Rail, a wire per Rail to the supervisior (or something like that like an LM393). And just some way to limit the current for a certain set of cables. And switches it off, if it is over the set limit.

PS: For the Electronic Nuts who want to learn about Power Supply Stuff and are pretty bored, google Texas Instruments UCD3138A.
And that thing is already used in some higher end Power Supplys today. And has has a 31,25MHz ARM7 Core...

but you argued that a missing signal is just a "incompatibility". It wasn't just slightly out of spec, it was completely dead.

I might have been wrong in that case.
And might have my mind at the Corsair <-> Gigabyte Situation because as far as I remember that was a fairly recently occurance for me at the time.

In your example, did you measure the PSU to verify why it was "incompatible"? Did you obtain another HX750i to verify it wasn't at fault? Did you perhaps think that the Gigabyte board might be slightly out of tolerance and was playing it safe by refusing to start with a possible faulty PSU? My point remains, you have no empirical evidence, just your assumptions.

No, because I don't have no warranty on any PSU I have here right now.
What I remember is that a somewhat similarlish PSU, though a bit different Plattform worked -> Corsair RM650i

And I have talked to people who were at the time responsible in PSU Development at a PSU Manufacturer. And they told me that that stuff might happen. They said that it had to do with the with the Tolerance of the PSU and the Board, if both are a bit borderline "in the right way"; it doesn't work.

And I think I've seen a similar configuration by someone else in a Forum in their Signature. So its not a general incompatibility just something that happened with the Board and the PSU I have.
I have two other PSU that use a similar Infineon LLC-Resonant mode Controller as the Corsair, both don't have that Problem. Though the Supervisor is a different one...

Those things are less common these days and probably something like winning the lottery (in a bad way).
The usual Failure rate of a decent PSU should be at 1% or less.
A 2% Failure Rate is considered bad and a major problem.

And you need "the right" motherboard as well...

Although I know jack all about Silverstone, they at least look like a legit company, Taiwanese instead of Chinese, legit presences in the US and EU, been around a long time etc. The supply I got has been independently tested to meet the Bronze 80 standard, which without knowing much about the world of PC PSU's sounded at least OKish. It's got to better than some OneHungLow no-namer from ebay? I just needed an OKish PSU, yet people seem to think this is some sort of crime against humanity 

Silverstone has some decent PSU, the Strider Gold ( ST55F-G) seems OK, below that, they could do better - much better.
The Strider Platinum Series is pretty nice, though the Casing is on the cheaper side.
Here some Pics of a Strider Platinum:
https://img.tweakpc.de/album/lT
Made by Sirfa, all Japanese Caps (made in China of course because of the Import Tax China has, it has to be manufactured there).

Best thing is the pretty beautiful soldering (except for the hand made part of course)...
No polymers though...

80plus is totally useless though because many companys (if not all) do send in golden samples...
Some even totally different units than they sell...

Interlaced 50i and 60i are a disaster for fast moving objects, much worse imo than 30p, 25p or even 24.

All of that is very relative. Raw frame rate tells you nothing, it depends on the persistence of the display.

eg. I used to use 50Hz monitors all day long, no problem, because the phosphors were designed for no flicker at 50Hz.

Then cheap-ass, low-persistence PC monitors came along and anything less than 75Hz was unbearable.

Cinemas project at less than 30fps but nobody complains of headaches.

@EEVblog: Get the MASTERWATT MAKER 1200 MIJ or nothing!

< goes on a rant about Dave's bad component choices because I'm an AMD fanboy and you should have gone with Ryzen >

Solution: Get it opened up on video (alongside the one you took out....)

Wanna bet the Thermaltake Toughpower has bulged caps?
If it is the one I think it is, it should come with Entry level SamXon Caps...
If I'm wrong, it is one of the newer ones with

€dit:
It is the Tough Power XT that comes with DC-DC but only has Teapo Entry Level Caps (SC Series, rated for 2k, not 3k because DOWNSIZE Caps or rather custom size wich makes a recapping almost impossible because you need 10mm 3300µF/16V Low Impedance Caps), Review from good old Gabriel Torres -> https://www.hardwaresecrets.com/thermaltake-toughpower-xt-775-w-power-supply-review/2/
Similar Plattform to the old Corsair HX750 and 850, the ones that were officially 80plus Silver.

But its around 10Years Old or so.

The ironical thing is that its better than the new one. As far as I remember they were always independently regulated -> Dual Mag Amp at the time 
Only the newest ones had DC-DC.
Problem is the age, it can fail any second now and its not that great by todays standards either.

Quote from: pkplex on June 26, 2018, 06:17:49 am

Dave really called it.. so many people with their nickers in a twist about a damn computer.

Because there is so much Misinformation, Bullshit, Lies and Fairy tales in this area you can write books about that...

And you felt it necessary to add to the heap?

Solution: Get it opened up on video (alongside the one you took out....)

I did, a teardown is in the video, it's exactly what I'd expect for the price, and a surprise of a Japanese main filter cap.