The Results Were Surprisingly Consistent: A Deep Dive into GPU Performance
As we began testing various graphics cards, including the high-end NVIDIA GeForce GTX 1080 Ti, we were surprised to find that the results were relatively consistent across different models. We started with the GTX 1070, which yielded a time of one second difference compared to the GTX 1060. This seemed promising, but things took an unexpected turn when we added the AMD Radeon Vega card, another high-end option, and again got exactly the same result within one second.
We decided to test our hypothesis further by adding even more cards to the mix, including the lower-end Nvidia GeForce GTX 1050, which initially yielded the same result as the other high-end cards. However, we were eager to explore what would happen when paired with a much slower card. The Nvidia GeForce GTX 660, an older model that doesn't seem to be relevant for modern gaming and content creation, proved to be a fascinating test subject.
To our surprise, the time difference between using just the CPU and adding a GPU was still significant, but not as pronounced as we had initially thought. When disabled, the hyper-threading feature on the 20-core processor effectively transformed it into a 10-core CPU, which yielded a baseline run time of seven minutes and 47 seconds. However, when paired with any GPU – including the GTX 660 or the high-end Nvidia GeForce GTX 1080 Ti – the overall render time decreased by approximately three minutes.
This pattern held true for various graphics cards, regardless of their computational performance or CUDA/OpenCL capabilities. We tested several other models, including the AMD Radeon RX Vega and the lower-end Nvidia GeForce GTX 1078, but consistently found that adding any GPU to our high-end CPU resulted in a similar reduction in render time. This lack of significant difference was quite unexpected, especially considering the substantial difference in performance between these cards.
This outcome led us to question what factors actually contribute to the performance benefits offered by GPUs in video rendering tasks. It became apparent that, at least for these specific tests and applications, the most important factor is not the GPU itself but rather the raw processing power of the CPU. The differences in CUDA or OpenCL capabilities between various graphics cards seemed to have little impact on the overall outcome.
The discovery has significant implications for content creators and professionals involved in video rendering tasks. If you're looking to reduce render times without sacrificing too much performance, it may not be necessary to invest in high-end GPUs like Vega or Nvidia GeForce GTX 1080 Ti. Instead, a mid-range option paired with a high-performance CPU could provide sufficient results.
However, there is one notable exception: cryptocurrency mining. The highly competitive and rapidly evolving world of cryptocurrency mining has created a demand for specialized hardware that can take full advantage of the massive computational power required to solve complex mathematical equations. It appears that some miners have found ways to utilize OpenCL or other programming frameworks to unlock significant performance gains in these specific applications.
As we continue to explore the capabilities of different graphics cards and CPUs, we will be testing their performance in various professional workflows, including those mentioned earlier like real-time auto-cad rendering and 3D modeling. It's also worth noting that if you're a content creator looking for a GPU with minimal impact on render time, you can choose any model without significant differences in performance.
In conclusion, our tests revealed some surprising patterns when it comes to the performance of different graphics cards paired with high-end CPUs. While it may not be what we initially expected, this experience has provided valuable insights into the relative importance of CPU processing power and GPU capabilities in video rendering tasks. We will continue to explore these findings and provide updates on our testing results as they become available.
"WEBVTTKind: captionsLanguage: enwhat's up guys Jays two cents here and I'm gonna ring every single ounce of content out of Vega than it possibly can because you know we talked a lot about gaming performance a lot of people focused on gaming but I'm one of the ones who said it's more than just gaming with this card it's all about compute performance and there's a lot of different ways you can leverage of that especially in CAD drawings live CAD renderings stuff like that but what about video encoding I've talked about it being a great card for video encoders and content creators alike so I figured what we do today is we would just go ahead and compare Vega 56 and 64 versus the gtx 1070 and 1080 and see which is actually doing a better job for content creators do you want to be cooler do you want to be more desirable well you're in luck because right now you can own your very own JS to sense swag and immediately be the cool kid on the block max out your sex appeal by following the link down below now this is an interesting matchup because on paper pricing wise these are pretty equally matched right 1070 versus 56 they're roughly the same price 399 and then you've got the gtx 1080 which is matched on paper versus the vega 64 and he I took the waterblock off to make this fair and as you can see they're all reference coolers so that's important but we live in the real world where the problem is Vegas 64 right now is actually being priced competitively against the GTX 1080 TI which we know this absolutely dominates this in games what about in a professional environment so in order to try and appease both the MSRP which we're hoping the graphics cards will actually come back down to so I'm comparing them 56 versus 2 and 70 64 versus 1080 and then 64 versus 1080 TI to basically account for the way things are right now at the time of making this video and in the future let's just hope that's completely irrelevant test but even on paper it shows that the Vegas 64 should be beating the GTX 1080 Ti in a more professional environment when you look at this the raw single precision teraflop numbers so I want to see if that actually plays out so with that said let's talk about the way we're gonna do the test I'm using an adobe premiere test project it's the bald eagle test we're gonna do is we run to a baseline because I've got my i9 7900 X on water with 32 gigabytes of Dominator Platinum 3030 200 megahertz memory ddr4 of course and then the CPU is overclocked to four point six gigs so I'm gonna do a baseline test to see how it does with no OpenCL and no CUDA to see how you know are we're gonna be doing without a graphics card then we'll run the same test compare the logs with each graphics card no over clocks or anything just setting them to their turbo clock and letting GPU boost 3.0 do its thing on NVIDIA side and then we are going to see what the results are so with that said enough talking and let's do it so we did quite a fit okay we did a lot of testing on this and I'll be honest we had some results we weren't expecting we added some extra tests I'll show you in a minute because we were trying to figure out some of these results now the perspective of this video the proof the perspective the perspective of this video suffering succotash I actually spit I saw it fly didn't hit you so we did our baseline test with just the CPU and it was about five minutes and 26 seconds you guys saw the graphs and then we put on the Vega 56 and used open CL and we saw it come down to exactly half of that about two minutes and 13 seconds which was pretty impressive so we put on the biggest 64 expecting to see it to come down even more because we know there's more stream processors more compute performance on the 64 and the price it was pretty much the same and we weren't really expecting that we thought wow that's interesting okay what happens when we put on the gtx 1070 exact same results I mean one second difference put on the gtx 1018 it was slightly slower than the 1070 even but again within a second so then i thought to myself this is very strange 1080 TI same result once again so what we found here was all the graphics cards were pretty much the exact same result within one second so I really started thinking about this and I went okay what what's the big picture here so I decided to end up taking my gtx 1050 which was something i didn't even mention at beginning this video because we added it later and guess what exact same result now I decided to go to probably the slowest card I have here which is kind of sad to mention I should I wish I had some older cards but I don't my gtx 660 and again the same result two minutes and 13 seconds so that guy does thinking here what what's happening we are seeing that obviously with a high end cpu like this 20 threads overclocked lots of ram that we were seeing a pretty much exactly half of the cpu performance when it comes to rendering time when you add a GPU any GPU it didn't matter if it was a high-end vega car to hide an Nvidia card or something that you can get for around 100 bucks it didn't really seem to matter so what we did was we ended up going into the BIOS N and disabling hyper threading turning this chip effectively into a core 10 thread CPU now we ended up trying to do actually four course eight threads turning into a very high-end 7700 K with a lot of cash for some reason the BIOS wasn't letting us do it it would be disabled but the windows still saw it so once we disabled the hyper threading we did another baseline run and we got seven minutes and 47 seconds up in code time with just the CPU 10 threads no hyper threading but once we hooked up a GPU any GPU 660 versus the 1080 Ti we got three minutes and thirty four or three minutes and 35 seconds with the 660 and three minutes and 34 seconds with the 1080i so once again that showed us that although the whole thing slid kind of as a whole bracket longer with less cores available to it to the CPU adding a GPU made a significant difference in render time but it didn't matter what GPU we used so that was something that was a little bit unexpected now I tell you this part right here that's obviously not the whole story when it comes to CUDA versus OpenCL right we're talking OpenCL here this is not a specific AMD technology it's an open-source technology but you know the amount of teraflops and compute performance really doesn't seem to make a difference when it comes to Vega versus CUDA on video rendering where it seems like the most important thing is that you have the most powerful CPU you can get and then any GPU whatsoever hooked up to it okay so yeah that's that's our results and it wasn't what we expected I had expected Vega to win I really did now unless you're gonna be doing extremely high-end Pro workflows with things like real-time auto cad rendering and 3d modeling which I don't have to actually have a way to test that and maybe I'll try to do that in the future I don't think you're gonna see the benefit of Vega whether it be for games or content creators which is really disappointing until the crowd the price comes back down it's not something that really makes a whole lot of sense at its current price point and unless the guys that are that are doing cryptocurrency mining and I'm hearing rumors that they're they're actually cracking this card using OpenCL and doing some amazing stuff when it comes to hash but no one will actually talk about how they're doing it I have no idea I have no idea who's buying up these cards so anyway if you guys think that we should do future testing on GPUs for performance and professional workflows and let me know what sweet two guys think we should do I'm not actually all that sure because I'm not sure the best way to compare it apples to apples but if you're content creator looking to create videos and you want to get the least amount of render time that obviously doesn't matter which GPU you go with it's still all about CPU anyway guys with that said I'm gonna go tell me what you guys thought about today's video these results were highly unexpected but they were fun to test which is why we ended up throwing more graphics cards at it because we were trying to find some rhyme or reason for it but whatever guys time to go thanks for watching and as always I'll see you in the next onewhat's up guys Jays two cents here and I'm gonna ring every single ounce of content out of Vega than it possibly can because you know we talked a lot about gaming performance a lot of people focused on gaming but I'm one of the ones who said it's more than just gaming with this card it's all about compute performance and there's a lot of different ways you can leverage of that especially in CAD drawings live CAD renderings stuff like that but what about video encoding I've talked about it being a great card for video encoders and content creators alike so I figured what we do today is we would just go ahead and compare Vega 56 and 64 versus the gtx 1070 and 1080 and see which is actually doing a better job for content creators do you want to be cooler do you want to be more desirable well you're in luck because right now you can own your very own JS to sense swag and immediately be the cool kid on the block max out your sex appeal by following the link down below now this is an interesting matchup because on paper pricing wise these are pretty equally matched right 1070 versus 56 they're roughly the same price 399 and then you've got the gtx 1080 which is matched on paper versus the vega 64 and he I took the waterblock off to make this fair and as you can see they're all reference coolers so that's important but we live in the real world where the problem is Vegas 64 right now is actually being priced competitively against the GTX 1080 TI which we know this absolutely dominates this in games what about in a professional environment so in order to try and appease both the MSRP which we're hoping the graphics cards will actually come back down to so I'm comparing them 56 versus 2 and 70 64 versus 1080 and then 64 versus 1080 TI to basically account for the way things are right now at the time of making this video and in the future let's just hope that's completely irrelevant test but even on paper it shows that the Vegas 64 should be beating the GTX 1080 Ti in a more professional environment when you look at this the raw single precision teraflop numbers so I want to see if that actually plays out so with that said let's talk about the way we're gonna do the test I'm using an adobe premiere test project it's the bald eagle test we're gonna do is we run to a baseline because I've got my i9 7900 X on water with 32 gigabytes of Dominator Platinum 3030 200 megahertz memory ddr4 of course and then the CPU is overclocked to four point six gigs so I'm gonna do a baseline test to see how it does with no OpenCL and no CUDA to see how you know are we're gonna be doing without a graphics card then we'll run the same test compare the logs with each graphics card no over clocks or anything just setting them to their turbo clock and letting GPU boost 3.0 do its thing on NVIDIA side and then we are going to see what the results are so with that said enough talking and let's do it so we did quite a fit okay we did a lot of testing on this and I'll be honest we had some results we weren't expecting we added some extra tests I'll show you in a minute because we were trying to figure out some of these results now the perspective of this video the proof the perspective the perspective of this video suffering succotash I actually spit I saw it fly didn't hit you so we did our baseline test with just the CPU and it was about five minutes and 26 seconds you guys saw the graphs and then we put on the Vega 56 and used open CL and we saw it come down to exactly half of that about two minutes and 13 seconds which was pretty impressive so we put on the biggest 64 expecting to see it to come down even more because we know there's more stream processors more compute performance on the 64 and the price it was pretty much the same and we weren't really expecting that we thought wow that's interesting okay what happens when we put on the gtx 1070 exact same results I mean one second difference put on the gtx 1018 it was slightly slower than the 1070 even but again within a second so then i thought to myself this is very strange 1080 TI same result once again so what we found here was all the graphics cards were pretty much the exact same result within one second so I really started thinking about this and I went okay what what's the big picture here so I decided to end up taking my gtx 1050 which was something i didn't even mention at beginning this video because we added it later and guess what exact same result now I decided to go to probably the slowest card I have here which is kind of sad to mention I should I wish I had some older cards but I don't my gtx 660 and again the same result two minutes and 13 seconds so that guy does thinking here what what's happening we are seeing that obviously with a high end cpu like this 20 threads overclocked lots of ram that we were seeing a pretty much exactly half of the cpu performance when it comes to rendering time when you add a GPU any GPU it didn't matter if it was a high-end vega car to hide an Nvidia card or something that you can get for around 100 bucks it didn't really seem to matter so what we did was we ended up going into the BIOS N and disabling hyper threading turning this chip effectively into a core 10 thread CPU now we ended up trying to do actually four course eight threads turning into a very high-end 7700 K with a lot of cash for some reason the BIOS wasn't letting us do it it would be disabled but the windows still saw it so once we disabled the hyper threading we did another baseline run and we got seven minutes and 47 seconds up in code time with just the CPU 10 threads no hyper threading but once we hooked up a GPU any GPU 660 versus the 1080 Ti we got three minutes and thirty four or three minutes and 35 seconds with the 660 and three minutes and 34 seconds with the 1080i so once again that showed us that although the whole thing slid kind of as a whole bracket longer with less cores available to it to the CPU adding a GPU made a significant difference in render time but it didn't matter what GPU we used so that was something that was a little bit unexpected now I tell you this part right here that's obviously not the whole story when it comes to CUDA versus OpenCL right we're talking OpenCL here this is not a specific AMD technology it's an open-source technology but you know the amount of teraflops and compute performance really doesn't seem to make a difference when it comes to Vega versus CUDA on video rendering where it seems like the most important thing is that you have the most powerful CPU you can get and then any GPU whatsoever hooked up to it okay so yeah that's that's our results and it wasn't what we expected I had expected Vega to win I really did now unless you're gonna be doing extremely high-end Pro workflows with things like real-time auto cad rendering and 3d modeling which I don't have to actually have a way to test that and maybe I'll try to do that in the future I don't think you're gonna see the benefit of Vega whether it be for games or content creators which is really disappointing until the crowd the price comes back down it's not something that really makes a whole lot of sense at its current price point and unless the guys that are that are doing cryptocurrency mining and I'm hearing rumors that they're they're actually cracking this card using OpenCL and doing some amazing stuff when it comes to hash but no one will actually talk about how they're doing it I have no idea I have no idea who's buying up these cards so anyway if you guys think that we should do future testing on GPUs for performance and professional workflows and let me know what sweet two guys think we should do I'm not actually all that sure because I'm not sure the best way to compare it apples to apples but if you're content creator looking to create videos and you want to get the least amount of render time that obviously doesn't matter which GPU you go with it's still all about CPU anyway guys with that said I'm gonna go tell me what you guys thought about today's video these results were highly unexpected but they were fun to test which is why we ended up throwing more graphics cards at it because we were trying to find some rhyme or reason for it but whatever guys time to go thanks for watching and as always I'll see you in the next one\n"
