People have been clamouring for 720p tests to remove GPU bottlenecks since Ryzen hit.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
*** AMD "Zen" thread (inc AM4/APU discussion) ***
- Thread starter Boomstick777
- Start date
People have been clamouring for 720p tests to remove GPU bottlenecks since Ryzen hit.
People have been clamouring for 720p tests to remove GPU bottlenecks since Ryzen hit.
Right.
He's another one who isn't paying attention or doesn't understand it.
Again... its not about the power of the GPU, its how efficiently the GPU can use the CPU.
At this point I think bottleneck has lost its meaning. Ryzen has more necks but they are narrower than intels. If you pour one at a time the intels faster, if you can pour more than one at once the AMD is faster. And it seems like the AMD gpu is a better bartender than NVidia.
Honestly though I think our tried and tested benchmarking methods aren't really giving a complete picture at the moment. Its been so long since CPU has really mattered in gaming that people don't really know a sure fire way to test them. Different tests, different results, different outcomes. I'm starting to think Jay was right to not bother with comparative stuff. Just show the ryzen performance and let people deal with it.
Honestly though I think our tried and tested benchmarking methods aren't really giving a complete picture at the moment. Its been so long since CPU has really mattered in gaming that people don't really know a sure fire way to test them. Different tests, different results, different outcomes. I'm starting to think Jay was right to not bother with comparative stuff. Just show the ryzen performance and let people deal with it.
As AdoredTV said "Tech reviewers just have to have a winner" so you do end up with all these out of the ordinary ways of testing that hardware, to force a difference between them, because in the real world there isn't.
So for the sake of the thread and my insanity sanity or anybody new to this thread catching up...
What's been concluded thus far re these finding ? All i have is images from its a knockout with the latest analogy ...
What's been concluded thus far re these finding ? All i have is images from its a knockout with the latest analogy ...
Last edited:
So for the sake of the thread and my insanity or anybody new to this thread catching up...
What's been concluded thus far re these finding ? All i have is images from its a knockout with the latest analogy ...
The GPU's reviewers use are not good enough to showcase Ryzen's actual performance
Role on big Vega.
So for the sake of the thread and myinsanitysanity or anybody new to this thread catching up...
What's been concluded thus far re these finding ? All i have is images from its a knockout with the latest analogy ...
Nvidia graphics card + Vulcan, DX12 and probably any other API's, maybe even some versions of DX11 = gimped system performance.
I would love to see how the Pro Duo cards perform with Ryzen and DX12.
As usual, AdoredTV is on to something. After his video we had the division, the 470/1060 comparison above, and now this:
https://www.reddit.com/r/Amd/comments/62s11m/german_tech_magazine_ct_did_a_test_with_a_fury_x/
A German magazine dared to compare the Titan X to a Fury X and got surprising results:
I look at these numbers and I can't believe them! April fools?
720p so all probably bottlenecked by the CPU which is why they are so close (apart from ROtTR)
Or am i missing something obvious here
( quite possible)As someone on reddit put -
"All this link shows is that at a CPU bottleneck amd drivers seem to show higher performance for an additional dx12 game, aots. Maybe shadow of mordor. That speculation is just that, and would need to be confirmed.
The rest is expected. At a CPU bottleneck you would expect a furyx and Titan xp to perform the same, despite the real world performance difference."
Last edited:
720p so all probably bottlenecked by the CPU which is why they are so close (apart from ROtTR)
Or am i missing something obvious here
As someone on reddit put -
"All this link shows is that at a CPU bottleneck amd drivers seem to show higher performance for an additional dx12 game, aots. Maybe shadow of mordor. That speculation is just that, and would need to be confirmed.
The rest is expected. At a CPU bottleneck you would expect a furyx and Titan xp to perform the same, despite the real world performance difference."
Again, as mentioned several times, we all agree there's a CPU bottleneck. We also agree that's the only reason the Fury X is anywhere near the Titan X and that the Titan is much faster than the Fury indeed. What you are missing is that we are NOT comparing the Fury to the Titan.
What we're saying is: for RoTR and a bunch of other games, it is evident that the software does a better job at splitting the CPU workload across more cores in the case of the Fury. This is the reason the single-core bottleneck hits later and magically the Fury comes out ahead (they should be flat).
Since there's no GPU bound (if anything the Titan would set a higher GPU bound) and the CPUs are the same, the only remaining variable is the software.
On the software side you have same OS, same game: only variable is drivers/DX12-implementation.
We are comparing AMD driver stack vs NVidia driver stack and saying the AMD one is better at utilising more cores (for the workload of those games) BECAUSE like you said the only bound is the CPU. There are 2 things that limit your performance on the given CPU. The primary one is when your code hits the ceiling of a single-core's performance and all progress is limited by that one core. Imagine now that the CPU is running that limiting core at 100% and the others at 80%. Now, if only you could offload some work from that core to the other under-utilised ones...
EDIT: The point is that if Vega was now available and Ryzen was benchmarked using a Vega, it would likely compete better against the 7700K and other quad-cores. This would also be expected of a 6900K using Vega. Again, it's NOT about the graphics cards. It's about how the software makes use of CPUs. Due to the absence of Vega, we had everyone review Ryzen with NVidia software and GPUs. The GPUs weren't the problem (high-end cards with low resolution, no GPU bottleneck). But apparently the software was. In some games. That's all it is.
Last edited:
Easy. Watch the video with 2 RX480s @1300. The guy is comparing it even to an overclocked TXP at DX12
The 480s trash the TXP by a huge margin on both Intel & AMD Cpus, when they shouldn't be nowhere near. While Ryzen fps almost doubles in DX12, bringing the performance in line with the Intel CPU. As it should, because DX12 is there to relieve the processes from the CPU not slow down.
And that in a Gameworks game
Soldato
- Joined
- 18 May 2010
- Posts
- 23,034
- Location
- London
Why is this not an issue that has already been observed with the Intel 6+ core/thread CPUs?
I.e X99 and up?
---
As an Nvidia GPU owner I've been feeling/saying there is an issue here with DX12 (970 Owner however)
I.e X99 and up?
---
As an Nvidia GPU owner I've been feeling/saying there is an issue here with DX12 (970 Owner however)
Last edited:
Easy. Watch the video with 2 RX480s @1300. The guy is comparing it even to an overclocked TXP at DX12
The 480s trash the TXP by a huge margin on both Intel & AMD Cpus, when they shouldn't be nowhere near. While Ryzen fps almost doubles in DX12, bringing the performance in line with the Intel CPU. As it should, because DX12 is there to relieve the processes from the CPU not slow down.
And that in a Gameworks game
Because... it is?
The 7700K beats both the 6-core 6800K and the 8-core 6900K when using high-end NVidia cards, because of this same issue (i.e. it's not just limiting Ryzen).
Like Panos said: two RX480s in crossfire will trash a TXP also on a 6900K. Because the AMD drivers will exploit those extra cores better.
Your mileage will vary per case (different games, different level of parallelism possible). Also there will be games where the less-parallel NVidia driver will perform better (especially old ones).
Soldato
- Joined
- 18 May 2010
- Posts
- 23,034
- Location
- London
If it is... Why is this news now?
probably because no one bothers to check the site directly, and even if they do you need to subscribe to see the full review.lol... how is it that no one spotted that?
but they did 1800X 2nd march and 1700/1700X 6th march