Video card rendering speed video

[Prysin]

So @LinusTech just upped a video on "which vendor is best at video editing" video on Vessel.

I have TWO issues with this video. However first i want to congratulate you on somehow getting back on track in terms of content quality. It was a fairly well laid out video, bit quick and a bit on the short end perhaps in terms of in depth testing.

 

A: The build itself is not really that representative to a pro rendering build, nor is it representative of a pro-sumer/casual youtuber. Due to this, we have almost no definitive proof of how thermals/chipset bandwidth/compatibility. It would be more interesting to see two separate builds, one 5960X/5830k/5820k variant (perhaps more CPU limited?) vs a proper server chassis configuration with the lid on (for thermal testing).

 

B: The lack of other programs that is commonly used ASIDE from Premiere (which can cost a lot)

 

There is also signs of the 290X was possibly thermal throttling in some tests. Seeing as you were using the reference model with the shitty cooler, which is notorious for thermal throttling. Yet it was not mentioned in the video, why? Did you not notice that a 290X was occasionally trailing the 290 by a massive margin?

 

@LinusTech@Slick@nicklmg

I would also like to suggest for a future video that you try to source a lower end (read; affordable to normal human beings) Firepro/Quadro. Something like a FirePro W7100 or a Quadro M4000 (600-800 USD on newegg). Because while rather expensive, these cards are tailored to professional workloads through highly optimized drivers and ultra binned GPUs (meaning they run at much lower TDP then normal cards, thus radically reducing the risk of thermal throttling)....

It would be interesting to see if a pro-sumer/youtuber would be better off with one of these, then a 980Ti/Titan X/390X/Fury X, or if these cards make no sense for the average mortal being.

 

[Matt100HP]

6 minutes ago, Prysin said:

Did you not notice that a 290X was occasionally trailing the 290 by a massive margin?

I noticed that when I was watching the video. I wondered if the 290X had done its favourite thing and got too hot, there's no other reason I could think of for it to be beaten by a 290 at anything performance related. 

[Vivilacqua1]

A - you could see the 290 performing better than the 290x, due to I would guess thermal throttling. Why limit the CPU when the test is about gpu's? Isolate the single variable and leave the others constant and controlled.

 

also, I noticed that in the last test, they didn't show the fury cards, just saying that "the 300 series is close I performance to the 200". I agree, but why leave out the fury/x when they cost about the same as the Nvidia 980/ti.

[samcool55]

Also the fact the fury nano is suddenly missing and he explains performance with that exact graph doesn't make me happy.

He could at least add it! Or use the fury X.

[Prysin]

Just now, Vivilacqua1 said:

A - you could see the 290 performing better than the 290x, due to I would guess thermal throttling. Why limit the CPU when the test is about gpu's? Isolate the single variable and leave the others constant and controlled.

 

also, I noticed that in the last test, they didn't show the fury cards, just saying that "the 300 series is close I performance to the 200". I agree, but why leave out the fury/x when they cost about the same as the Nvidia 980/ti.

they had the Nano, which is a full fledged Fury X core... due to thermals and clock speed it generally performs around a R9 Fury though.

[iamdarkyoshi]

1 minute ago, Vivilacqua1 said:

A - you could see the 290 performing better than the 290x, due to I would guess thermal throttling. Why limit the CPU when the test is about gpu's? Isolate the single variable and leave the others constant and controlled.

 

also, I noticed that in the last test, they didn't show the fury cards, just saying that "the 300 series is close I performance to the 200". I agree, but why leave out the fury/x when they cost about the same as the Nvidia 980/ti.

This. I noticed it too and I was like wtf...

[Sakkura]

There was only one 290(X) reference cooler in the case as far as I could see, so I just assumed that was the 290X, and it was thermal throttling badly because of how tightly packed the cards were (and just how bad the reference cooler itself is).

[samcool55]

1 minute ago, Prysin said:

they had the Nano, which is a full fledged Fury X core... due to thermals and clock speed it generally performs around a R9 Fury though.

Yes, but not in the last graph and that's the exact graph linus uses to go in detail of the individual performance of the cards.

 

And the fact he says "the beefiest gaming gpu, in this case the GTX 980 ti, in this system, comes out on top" is the sentence that i really have an issue with.

Because the nano was ALSO in that system, becuase it has been in the 2 graphs before that, so it clearly was there, but it suddenly went missing, and could be ahead of the 980 ti.

[Prysin]

9 minutes ago, samcool55 said:

Yes, but not in the last graph and that's the exact graph linus uses to go in detail of the individual performance of the cards.

 

And the fact he says "the beefiest gaming gpu, in this case the GTX 980 ti, in this system, comes out on top" is the sentence that i really have an issue with.

Because the nano was ALSO in that system, becuase it has been in the 2 graphs before that, so it clearly was there, but it suddenly went missing, and could be ahead of the 980 ti.

no it cant. Simply put, Adobe software IS more optimized for CUDA then OpenCL. Just a matter of fact. I was going to mention it in the OP, but considering the results arent that far apart, i didnt want to nitpick that much.

 

If they had used say Sony Vegas, AMD would have destroyed Nvidia in every test, because Sony Vegas is so well optimized for OpenCL that AMD cards simply crushes.

[samcool55]

Just now, Prysin said:

no it cant. Simply put, Adobe software IS more optimized for CUDA then OpenCL. Just a matter of fact. I was going to mention it in the OP, but considering the results arent that far apart, i didnt want to nitpick that much.

 

If they had used say Sony Vegas, AMD would have destroyed Nvidia in every test, because Sony Vegas is so well optimized for OpenCL that AMD cards simply crushes.

But amd was ahead in every situation apart from the last one.

The last graph was the only one in favor of Nvidia... And even if the nano can't beat the 980 ti, i still want to see the results.

I don't care what card wins in that scenario, i do care about every card being in there and showing us an objective as possible result.