AMD's dx 11 CPU overhead is essentially gone (at least on Polaris)

[ivan134]

I was messing around with the flawless widescreen program yesterday and I decided to fire up far cry 3. There are places in the game that task my CPU heavily and my fps would drop to 55. I was now getting 80 - 85 fps on those areas. I didn't think much of it until I found this thread while browsing reddit.

 

It references a benchmark done by hardware unboxed yesterday and the results are amazing. This used to be a problem for AMD in DX 11 and would manifest itself the worst when pairing higher end GCN cards with lower end CPUs. This problem has literally poofed away. The most amazing part to me is GTA 5. The 470 is literally on heels of the 1060 even with a low end CPU. If you remember, even in the updated 480 vs 1060 benchmarks, the 480 was still trailing the 1060 in GTA 5 by a very considerable amount. It looks like it's faster than the 1060 now if we extrapolate how much faster the 480 is than the 470, but we will still need concrete information to state that as a fact. The improvements in overwatch are good too.

 

I don't know which driver this happened with, but I'm on the latest one which I updated yesterday so I'm going to guess it's the most recent driver. We also don't know if this is Polaris only or across all GCN cards. It would be nice if those of you with older gen GCN cards would test any CPU intensive games you have. I'm going to tests borderlands 2 when I get home because there are areas where similar things as in far cry 3 would happen. I still have my 390 and can pop it back in my system if need be.

 

@Morgan MLGman @Clanscorpia @deXxterlab97 @xAcid9 @-rascal-

[Bananasplit_00]

this would give me even more reson to get an R9 Fury...

[tlink]

holy fucking shit amd it looks like you actually pulled this off.

[Guest]

I downgraded my RX 480 to the one came with the card (from their site). It seems to cause crashes. Haven't got the 17.xx driver yet.

So. Where is that damn Kaby Lake board for cheap? Tired of recommending people to buy i3 6100 with a H110/B150 board as a G4560 though cheaper requires a more expesnive Kaby Lake board

[Guest]

2 minutes ago, Bananasplit_00 said:

this would give me even more reson to get an R9 Fury...

Or Vega.

[Majestic]

 

You were saying?

 

I will say Polaris is doing really well. But Fiji is still crap.

[ivan134]

1 minute ago, Majestic said:

 

You were saying?

Good job not reading the thread.

[Majestic]

5 minutes ago, ivan134 said:

Good job not reading the thread.

I later edited it. But as you didn't refute the people pondering over buying a Fury, I felt obligated to post it.

 

Oh, and that video shows numerous examples of the 1060 being saturated with lower hardware than the Polaris chip. The guy uses way too many weaselwords and is clearly biased towards the results. 

 

>not significant

when talking about a +10 fps difference in minimum framerates

[Bananasplit_00]

On 2/1/2017 at 4:59 PM, deXxterlab97 said:

Or Vega.

nope, Vega wont be cheap like the Fury is, a new R9 Fury is $350, thats what i paid for my 970! Vega is probably going to be hitting close to what Pascal is doing and therefor priced similarly

On 2/1/2017 at 5:01 PM, Majestic said:

I later edited it. But as you didn't refute the people pondering over buying a Fury, I felt obligated to post it.

this driver update would push up those FPS numbers if it affected the Fury too, makeing it even more of a great choice for < DX12 games, which most titles are currently. the Fury is increadably cheap used, around $200, because people think its as much of a dissapointment as at launch

[Megah3rtz]

7 minutes ago, deXxterlab97 said:

requires a more expesnive Kaby Lake board

Nei? some boards can get their bios updated without CPU, pick one of those. ( Gigabyte boards: "Qflash Plus" Asus boards: "USB BIOS Flashback" ) regardless, wouldn't a g4560 be a downgrade/sidegrade from your 4460? 

[ivan134]

6 minutes ago, Majestic said:

I later edited it. But as you didn't refute the people pondering over buying a Fury, I felt obligated to post it.

There's no reason for me to repeat myself and your refute is still not relevant because that benchmark is old.

[Guest]

1 minute ago, Megah3rtz said:

Nei? some boards can get their bios updated without CPU, pick one of those. ( Gigabyte boards: "Qflash Plus" Asus boards: "USB BIOS Flashback" ) regardless, wouldn't a g4560 be a downgrade/sidegrade from your 4460? 

I did not say I am going to get one

Yes. Qflash plus or USB BIOS flashback which are found in most high end boards

[Majestic]

1 minute ago, ivan134 said:

There's no reason for me to repeat myself and your refute is still not relevant because that benchmark is old.

But the one you posted has it's fair share of issues aswell. Like his weird subjective interpretation of the word "significant" and the amount of weaselwords used when the results are not in favor of the AMD setup and similarly superlatives when in favor of AMD.

 

Also, this is the budget segment. You're negating the high-end completely and also 144hz users. Not everyone is interested in sub60 results.

[tlink]

2 minutes ago, Majestic said:

But the one you posted has it's fair share of issues aswell. Like his weird subjective interpretation of the word "significant" and the amount of weaselwords used when the results are not in favor of the AMD setup and similarly superlatives when in favor of AMD.

 

Also, this is the budget segment. You're negating the high-end completely and also 144hz users. Not everyone is interested in sub60 results.

yea but it still illustrates a larger overall point which is what this post is about. refuting bad data with more bad data is not a good way to argue.

[ivan134]

6 minutes ago, Majestic said:

But the one you posted has it's fair share of issues aswell. Like his weird subjective interpretation of the word "significant" and the amount of weaselwords used when the results are not in favor of the AMD setup and similarly superlatives when in favor of AMD.

 

Also, this is the budget segment. You're negating the high-end completely and also 144hz users. Not everyone is interested in sub60 results.

Seriously, go back and read the OP and watch the entire video because I'm going to start ignoring anything else you write here if you don't.

 

• Overwatch is in the video.
• 480 is the highest end GCN 4 GPU available.
• This test had not been done on older GCN cards and I asked people to post their own results.

Stop arguing for the sake of arguing.

[Droidbot]

this is definitely a good thing - 480 and 6100 is now a pretty good pairing on paper and in reality. 

AMD does it again with a lower R&D budget

[Majestic]

34 minutes ago, ivan134 said:

Seriously, go back and read the OP and watch the entire video because I'm going to start ignoring anything else you you're here if you don't.

 

• Overwatch is in the video.
• Polaris is the highest end 480 available.
• This test had not been done on older GCN cards and I asked people to post their own results.

Stop arguing for the sake of arguing.

I did watch the video, and I've been looking at his channel since he had like 1K subs. And without the audience he has now (and thus being toned down) I got a good taste of his personality. He was extremely volatile once you questioned the results or methods of his video's. And in this case there is a lot that can be said. Like the GTA 5 results showing the same performance for the 6700K and 7600K, eventhough i7 is clocked much higher. Or how he keeps saying "the margins are insignificant" when there is >10 fps difference in minimum framerates (at 70-ish fps). Not saying because of the outlet i'm not trusting the results, but i'm inclined to not trust my judgement solely on them either. I'd be more inclined to trust someone like AdoredTV for that. Also self-admitted AMD skewed, but his methodologies are mostly impeccable.

 

EDIT: i rest my fucking case, lol.

 

 

Because I was however, prior to the video, not unfamiliar with the strides AMD made with polaris in terms of overhead. And you haven't seen me complain much on polaris either the last few months. But as rosy as the guy in the video is making (pretty much gone or better than Nvidia) is too far fetched. 

 

These are from 16 december, not ancient but fairly recent.

https://nl.hardware.info/reviews/7063/3/civilization-vi-review-benchmarks-met-20-gpus-testresultaten-full-hd-1920x1080-+-frametimes

 

• Which is a counterargument to what point? And do note it was a bot match.
• People still buy it with 144hz freesync panels
• no comment

[Nena Trinity]

About time! (now I can buy Vega for my FX-8370!)

[Majestic]

9 minutes ago, Nena360 said:

About time! (now I can buy Vega for my FX-8370!)

You do realise reduction of overhead does not increase the absolute performance? Or were you just jesting?

[tlink]

9 minutes ago, Majestic said:

These are from 16 december, not ancient but fairly recent.

https://nl.hardware.info/reviews/7063/3/civilization-vi-review-benchmarks-met-20-gpus-testresultaten-full-hd-1920x1080-+-frametimes

but aren't they still a big update behind? v16 vs v17 is what we should be looking at, not chronological distance right?

[Majestic]

1 minute ago, tlink said:

but aren't they still a big update behind? v16 vs v17 is what we should be looking at, not chronological distance right?

If that's the case, tests should be done between those specific driversets.

[Nena Trinity]

2 minutes ago, Majestic said:

You do realize reduction of overhead does not increase the absolute performance? Or were you just jesting?

I expect that with DX12 and Vulkan replacing DX11 along with the better optimized DX11 games recently I have high hopes for my FX-8370 to perform decently with my future GPU even if its Nvidia or Radeon!

[8uhbbhu8]

Hmmm. Ima try getting the latest drivers for my 7770 and see if it does anything  probably not but....

[Droidbot]

4 minutes ago, Nena360 said:

I expect that with DX12 and Vulkan replacing DX11 along with the better optimized DX11 games recently I have high hopes for my FX-8370 to perform decently with my future GPU even if its Nvidia or Radeon!

 

You realise that the FX processor line is 6 years old, right? The 8350 bottlenecks the RX 480 and 1060 by a good margin even compared to a 26/2700k...

[Majestic]

8 minutes ago, Nena360 said:

I expect that with DX12 and Vulkan replacing DX11 along with the better optimized DX11 games recently I have high hopes for my FX-8370 to perform decently with my future GPU even if its Nvidia or Radeon!

Even in low overhead scenario's the FX series loses to the new G4560. So i'm not sure what longetivty you're expecting out of a budget-level chip at this stage.

 

3 minutes ago, 8uhbbhu8 said:

Hmmm. Ima try getting the latest drivers for my 7770 and see if it does anything  probably not but....

It's mostly Polaris though, so probably not.