AMD RX Vega 64 Outperforms NVIDIA GTX 1080 Ti By Up To 23% In DX12 Forza 7

[The Benjamins]

4 minutes ago, MadyTehWolfie said:

You do know the game lowers and raised in game assets to keep frame rate up that =/= good optimization. Also I don't have anything set to dynamic and playing at a higher resolution. If you have resolution set to dynamic as well it'll lower that too. Turn off dynamic settings and see what happens. Also AMD performs better in dx12 maybe a factor as well. 

well, I have the res locked, and I just want a playable experience, my hardware is not the greatest and I always have to much open in the background but the fact that they dynamic settings keeps it smooth the whole time is amazing. 

 

What happens if you run dynamic on some stuff for you?

[MadyTehWolfie]

30 minutes ago, The Benjamins said:

well, I have the res locked, and I just want a playable experience, my hardware is not the greatest and I always have to much open in the background but the fact that they dynamic settings keeps it smooth the whole time is amazing. 

 

What happens if you run dynamic on some stuff for you?

Didn't try it and don't want to. I have a 1080ti ftw3 it shouldn't have dips below the 60hz limit on my monitor period in this game. Though at one point I did lower the settings by one slide to the left and still had the same dips in the same areas. 

[Mira Yurizaki]

3 hours ago, cj09beira said:

amd cards usually to the opposite of becoming slower with time, before memory starvation there are other things, like the same amount of rops, etc of a 290x with 2 times the compute 

In DX11, sure. In DX12 or Vulkan, GCN gets to bare its fangs.

[Aprime]

On 10/1/2017 at 1:48 AM, kiska3 said:

Enjoy some graphs!
 

  Hide contents

 

I feel bad for anyone that's bought a Fury.

[Phate.exe]

11 minutes ago, Aprime said:

I feel bad for anyone that's bought a Fury.

You don't need to talk about me like I'm not here, lol.

 

Then again I bought my Fury for $300 at the peak of the crypto-mining GPU price insanity, after selling my RX 470 for $330, so it wasn't a bad purchase.

 

If I had to guess, Forza 7 at 4k with whatever settings they tested at just uses more than the 4 gigs HBM.  Like a lot more.  Furys seem to do better than you'd expect with only 4 gigs, not sure if that's because of the HBM or whatever, but you're just gonna have a bad time trying to cram 5-6 gigs (based on the 1060 doing well) into 4.  Assuming the 580 in there is an 8 gig card, it's actually doing really well at such a high resolution, but a Fury should still slap it around pretty nicely as long as it's not choking on the ram limit.

 

If anybody cares I'm buying the game tonight/tomorrow so I'll have some more info available about just how far you have to dial things back before the Forza 7 will run smoothly at 4K on a Fury.  All I need to do is keep it in the 50's with pretty graphics at 4k and I'll be happy with it.

[suchamoneypit]

On 10/1/2017 at 5:01 AM, Master Disaster said:

Amazing job AMD, your top of the line GPU manages to smash everything in a 10 year old resolution.

 

All hail the new 1080p king.

vast majority of users are on 1080p. 57% of steam users are on 1080p.

[raultherabbit]

On 10/2/2017 at 11:43 AM, cj09beira said:

could you check individual cpu core utilization plz, 

Attached.  The update today seems to have fixed the stuttering issue which almost disappoints me.  I'm happy to report everything is quite smooth and runs well now, no crashes.  My PC specs are 16gb ram @ 3000mhz, 6700k @ stock clock and a FE edition 1080ti.  You can see that the game uses one cpu core heavily, and the rest low/moderately. 

[Jito463]

On 10/1/2017 at 7:21 AM, Trixanity said:

In this case it appears to be capacity as well. 1080 Ti has 11 GB of memory versus 8 GB for Vega. Their memory bandwidth should be similar. Nvidia has better compression though so the effective memory bandwidth might be significantly higher. 

 

Having to swap data from memory instead of just reading it definitely reigns in the performance of Vega in this case.

Kind of makes one long for the days of video cards with expandable RAM (anyone else remember those days?).  Yeah, the RAM would almost certainly be slower than the onboard, but it would be nice to buy a video card and then add more VRAM as needed.

On 10/2/2017 at 12:02 AM, Phate.exe said:

In the CS:S days I played on a flatscreen CRT at 60hz.

60Hz on a CRT is different from 60Hz on a LCD.  They display the data in very different fashions.  Just because 60Hz is perfectly acceptable on a CRT doesn't mean it's perfectly acceptable on a LCD.

 

Obligatory qualifier: That's not to say that high refresh rate monitors will make a difference for some (most?) people in game - and I'm not saying anyone absolutely needs the higher refresh rate - just pointing out that there are significant differences between the two.

[vanished]

3 minutes ago, Jito463 said:

60Hz on a CRT is different from 60Hz on a LCD.  They display the data in very different fashions.  Just because 60Hz is perfectly acceptable on a CRT doesn't mean it's perfectly acceptable on a LCD.

If anything high refresh was more important on CRTs to avoid the flicker

[Jito463]

1 minute ago, Ryan_Vickers said:

If anything high refresh was more important on CRTs to avoid the flicker

I don't recall that at all.  I do recall dealing pretty heavily with ghosting in the early LCD days, making CRTs the superior (at the time) option.

[vanished]

Just now, Jito463 said:

I don't recall that at all.  I do recall dealing pretty heavily with ghosting in the early LCD days, making CRTs the superior (at the time) option.

In some ways they're still superior.

[Jito463]

Just now, Ryan_Vickers said:

In some ways they're still superior.

This forum needs to bring back the 'Thumbs Up'.  I can't bring myself to "heart" a post.  

[Phate.exe]

44 minutes ago, Jito463 said:

60Hz on a CRT is different from 60Hz on a LCD.  They display the data in very different fashions.  Just because 60Hz is perfectly acceptable on a CRT doesn't mean it's perfectly acceptable on a LCD.

 

Obligatory qualifier: That's not to say that high refresh rate monitors will make a difference for some (most?) people in game - and I'm not saying anyone absolutely needs the higher refresh rate - just pointing out that there are significant differences between the two.

Oh I know, I held out on CRT's until 2013 on my home setup (at school I just used my TV as a monitor, because lack of room in my dorm).  Early LCD's were pretty trash, but pretty much anything made in the last 6 years or so has been more than fine.

[leadeater]

2 hours ago, Jito463 said:

60Hz on a CRT is different from 60Hz on a LCD.  They display the data in very different fashions.  Just because 60Hz is perfectly acceptable on a CRT doesn't mean it's perfectly acceptable on a LCD.

I used to run an old ex medical grade Philips CRT that could do 120Hz @ 1280x1024 and 100Hz @ 1600x1200, it also used 5 BNC inputs (3 colour, 1 contrast, 1 brightness). Not sure of the exact model but damn that thing was awesome, free stuff ftw.

 

[cj09beira]

4 hours ago, leadeater said:

I used to run an old ex medical grade Philips CRT that could do 120Hz @ 1280x1024 and 100Hz @ 1600x1200, it also used 5 BNC inputs (3 colour, 1 contrast, 1 brightness). Not sure of the exact model but damn that thing was awesome, free stuff ftw.

 

thats a beast of a monitor, and because crts flicker it should feel seems even smoother than a 120hz lcd

 

[Morgan MLGman]

15 hours ago, Aprime said:

I feel bad for anyone that's bought a Fury.

You really shouldn't, an R9 Fury/Fury X are still damn good cards and those benchmarks are pretty much useless - yes, they tested the cards with the same graphics settings but it doesn't show you how those cards actually perform.

They should've dropped the texture setting for cards with >4GB of memory because scores like that only indicate one thing: the cards hit the memory limit hard.

[TheOriginalHero]

23 hours ago, AlwaysFSX said:

No they won't because there's two fundamental problems with Vega and AMD cards in general, shown below:

It's starved for bandwidth.

 

AMD's drivers are terrible at keeping the core filled with work because they're not developing an architecture for today, they're developing an architecture for tomorrow.. (ie: we can't make drivers so let's just have the game engine sort that out for us!) Which'll be underperforming and dead by the time tomorrow rolls around.

Doesnt seem to be that much of an issue here. As with any amd product drivers tend to get siginificantly better overtime. If optimizations can make the vega architecture preform this good then im excited.

[cj09beira]

@AlwaysFSX its totaly not the drivers fault, as amd does scheduling on hardware in the gpu, the problem is that amd gpus all have 4 pipelines and after some point (290x) the only thing getting faster is compute and shaders, and many times other bottlenecks come long before shaders are a problem

[AlTech]

On 10/1/2017 at 11:59 AM, huilun02 said:

I wouldn't be surprised if it wasnt a Gameworks title. 

It isn't. Forza 6 Apex wasn't. Funnily enough that has higher requirements than Forza 7.

[Humbug]

On 10/1/2017 at 7:35 PM, Misanthrope said:

I'm thinking Nvidia will have a game ready driver for this game narrowing down or erasing this win. Maybe not, who knows? It might inspire some miners to take a break to actually enjoy a game in kind of a gaming miracle they'll realize they were frustrated gamers all of this time!!!

WHQL 385.69 released on September 21st was Nvidia's optimized game ready driver for Forza 7.

https://www.geforce.com/drivers/results/123936

 

Further improvements are not  impossible but the only Nvidia cards that are underpeforming badly in this game are the Maxwell GTX 960 and 970 compared to their competitors R9 380 and R9 390. But some would say that is expected as Nvidia's long term driver support is not as good as AMD.

 

When it comes to cards currently on the market Nvidia seems to be already doing well too in this game though. At 1440p maxed out everything from GTX 1060 upwards easily clears 60fps. And at 1440p the GTX 1060 achieves the same average fps as the RX580. GTX 1070 is way ahead of the Fury X also.

 

Seems to be just a case of the Vega architecture in particular over-performing in this game compared to our expectations. Maybe something about this particular DX12 implementation...

[Humbug]

LOL for some extreme shit....

 

Vega56 doing 8K ultra 60fps with crazy weather effects

 

 

[leadeater]

12 minutes ago, Humbug said:

LOL for some extreme shit....

 

Vega56 doing 8K ultra 60fps with crazy weather effects

That is some terrible driving though, Forza 7 MissApex 

[Vode]

It doesn't look like Forza 7 is the only game where Vega beats the Pascal cards.

 

Dirt 4 and PUBG on High preset (PC Games Hardware) show Vega in a similar lead.

 

I wonder if we will start to see this more and more with console releases. Damn I hope Pascal is not going to die the Kepler death.

[ravenshrike]

11 hours ago, leadeater said:

That is some terrible driving though, Forza 7 MissApex 

Not to mention he's clearly got full assists on.

[Humbug]

1 hour ago, Vode said:

It doesn't look like Forza 7 is the only game where Vega beats the Pascal cards.

 

Dirt 4 and PUBG on High preset (PC Games Hardware) show Vega in a similar lead.

Vega beating the gtx 1080 happens in a number of games...

 

it's beating the 1080ti that is unexpected.