Doom Vulkan with different CPUs

shankly1985 · 27 Jul 2016 at 04:18

Doom112 said:
AMD has a bottleneck with Vulkan.
One user posted some benchmark here.
http://forums.anandtech.com/showthread.php?t=2479898&page=9

3rd party reviews made a pretty big mistake that they are doing benchmark on mid tier card with a high end CPU. Most of the people who will buy 480s will be on 5 to 7 years old i7. There is no person who will be buying a $350 CPU for a $2xx GPU and this is the reason experience>>>>>>>benchmark.

The purpose of DX12 and Vulkan was to reduce CPU bottleneck ,which clearly is not happening on AMP GPUs.

Charts don't tell the full story!

flinco said:
I tried Vulkan on my i5 750 and 970. It stuttered like crazy and was horrible to play.

This my point why I said above we need more testing. Looking at a chart doesn't show issue like what you saying.

Charts are old school we all need to move on.

/end off

More users on that forum reporting same issue, didnt need to look far, i'll take user feedback All day long before a chart!!!

HurleyBird said:
The Vulkan path feels very choppy compared to OpenGL on my 980/4770K rig.

AdamK47 said:
I played the game last night with Vulkan enabled. I noticed some stuttering in the Argent Energy Tower map when looking down the long walkway. Looking from either end feels choppy. Keep in mind, Doom with Vulkan does not use mGPU. When I enable the framerate monitor, it still shows it locked at a solid 60 FPS. That's odd. I went ahead and enabled OpenGL (with SLI enabled). The choppy feeling in these areas was gone. It was entirely smooth.

I'm going to stick with OpenGL.

Scougar · 27 Jul 2016 at 04:33

Has anyone verified this users results?

Kaapstad · 27 Jul 2016 at 06:03

Doom112 said:
AMD has a bottleneck with Vulkan.
One user posted some benchmark here.
http://forums.anandtech.com/showthread.php?t=2479898&page=9

3rd party reviews made a pretty big mistake that they are doing benchmark on mid tier card with a high end CPU. Most of the people who will buy 480s will be on 5 to 7 years old i7. There is no person who will be buying a $350 CPU for a $2xx GPU and this is the reason experience>>>>>>>benchmark.

The purpose of DX12 and Vulkan was to reduce CPU bottleneck ,which clearly is not happening on AMP GPUs.

I have both a RX480 and GTX 1060 which I use with a 6950X.

nashathedog · 27 Jul 2016 at 08:00

RavenXXX2 said:
So basically you need a £300 CPU to max out the 480 whist any old cpu will give great performance with nvidia. Seems to me Nvidia have the edge when it comes to driver optimiazation.

That's nothing new though is it. I think this shows that that the newer api's have not provided a magic fix for the deep rooted problems inherent with AMD's hardware.

nashathedog · 27 Jul 2016 at 08:03

Kaapstad said:
I have both a RX480 and GTX 1060 which I use with a 6950X.

You're an exception to the rule, I'm sure you're not the only one but I imagine 99.8% of 6950x owners are not pairing there cpu with a 480 or 1060.

Griffildur · 27 Jul 2016 at 09:19

nashathedog said:
You're an exception to the rule, I'm sure you're not the only one but I imagine 99.8% of 6950x owners are not pairing there cpu with a 480 or 1060.

Yeah whoever buys such a processor won't buy a low end kinda card

JanxSpirit · 27 Jul 2016 at 09:44

Kaapstad said:
If the faster of the two GPUs is not getting CPU bottlenecked then there is no way the slower of the two will.

And your evidence that the faster of the two isn't CPU bottlenecked is where?

I can tell you from personal experience that an HD6850 and an HD7950 are both heavily CPU bottlenecked in nearly all recent games by anything less than an i5 6600K at 4.5GHz. The link I posted earlier proves that this is also the case for a GTX980 Ti.

It really is all about CPU IOPS, and there's currently no CPU on the mainstream market fast enough to not represent at least a partial bottleneck in DX11 or DX12 gaming.

N19h7m4r3 · 27 Jul 2016 at 10:03

Griffildur said:
Yeah whoever buys such a processor won't buy a low end kinda card

I have the 5820K and I'm getting the RX 480 for its encoding hardware, and because it's decent at gaming for the price as well.

Griffildur · 27 Jul 2016 at 10:05

N19h7m4r3 said:
I have the 5820K and I'm getting the RX 480 for its encoding hardware, and because it's decent at gaming for the price as well.

that's fair enough, but 6950x is an expensive 10 core processor, I doubt we will see many people pairing it up with a 480 or 1060.

I am not including Kaap in this category ...

CAT-THE-FIFTH · 27 Jul 2016 at 10:09

LOL,I have a Xeon E3 1230 V2 with a GTX960,another mate has a Core i5 6500 and a GTX960 another mate has a Xeon E3 1230 V3 with a R9 280 and yet another mate has a Core i7 6700 and is getting a RX480. Know a few mates running Haswell Core i5 CPUs with a GTX970.

Plenty of people spending £150 to £250 on a CPU now.

The excuse makers here can't explain why under Vulkan I still get no improvement or 4lower of FPS over using OpenGL in Doom.

OpenGL is better for me than Vulkan and this is despite Nvidia making a big song and dance about it:

https://blogs.nvidia.com/blog/2016/02/16/vulkan-graphics-api/
http://www.geforce.co.uk/whats-new/...-gtx-1080-run-doom-at-up-200-fps-using-vulkan

Watch The GeForce GTX 1080 Run Doom at up to 200 FPS Using Vulkan

Edit!!

I like how someone links to that AT thread to make sure they find the most negative results possible,then missed this userbench:

https://docs.google.com/spreadsheet...uoXdKcAnY/pubchart?oid=797892815&format=image
https://docs.google.com/spreadsheet...oXdKcAnY/pubchart?oid=2130497165&format=image

Look at the scaling, unboxed showed a 4% loss going from an overclocked 6700k to a stock x4 955. Look at GameGPU with a 1080, you see a loss of almost 40% going down to an FX6100. The 955 should be right about the performance of the 6100, maybe a tad faster.

I'm also seeing twice the performance uplift from Vulkan on the q9300 compared to unboxed results and more than GameGPU as well. The reason is simple, their tests are basically running around an empty room. These are not indicative of actual game performance. Basically, if you look at the frame time graph I posted and look only at the spot where I'm essentially looking at a wall (It's a scripted event of throwing a switch in the game) you see the fps is at it's highest and OpenGL and Vulkan perform basically the same just like you see with unboxed reviews. If you want to see performance during real gaming though, look outside of that area in the graph and you'll see a much different story.

GameGPU's benchmark is basically the same thing as well, they run around an empty room but at least they shoot a barrel and have a nest that adds some load to the scene. This is not indicative of actual in game performance. It's similar to me to looking at max fps and determining a hierarchy. I have some more data too though, I'll see if I can parse through that today and get it posted.

Griffildur · 27 Jul 2016 at 10:11

CAT-THE-FIFTH said:
LOL,I have a Xeon E3 1230 V2 with a GTX960,another mate has a Core i5 6500 and a GTX960 another mate has a Xeon E3 1230 V3 with a R9 380 and yet another mate has a Core i7 6700 and is getting a RX480.

Plenty of people spending £150 to £250 on a CPU now.

The excuse makers here can't explain why under Vulkan I still get no improvement or a lower of FPS over using OpenGL in Doom.

yes no one is debating that, all I said was that the 6950x is a 10 core processor and costs about £1400. Whoever buys it doesn't really worry about the cost of components

as for no one explaining the results, all we can do is debate it but we don't really know how Vulkan is coded so ...

CAT-THE-FIFTH · 27 Jul 2016 at 10:25

Griffildur said:
yes no one is debating that, all I said was that the 6950x is a 10 core processor and costs about £1400. Whoever buys it doesn't really worry about the cost of components

as for no one explaining the results, all we can do is debate it but we don't really know how Vulkan is coded so ...

Its annoying Nvidia made such a big song and dance about it and months later nothing despite them making more of a song and dance about it earlier than AMD.

Bethesda and iD are generally more Nvidia focussed devs anyway,as they have in the past tended to work a bit closer with Nvidia.

But being Nvidia it will be swept under the carpet until Volta just like with the async patch which some promised here for Maxwell for months.

Kaapstad · 27 Jul 2016 at 10:56

JanxSpirit said:
And your evidence that the faster of the two isn't CPU bottlenecked is where?

I can tell you from personal experience that an HD6850 and an HD7950 are both heavily CPU bottlenecked in nearly all recent games by anything less than an i5 6600K at 4.5GHz. The link I posted earlier proves that this is also the case for a GTX980 Ti.

It really is all about CPU IOPS, and there's currently no CPU on the mainstream market fast enough to not represent at least a partial bottleneck in DX11 or DX12 gaming.

Look at the graph in the OP.

As to recent games and CPUs, you have got it round the wrong way. They are easy for CPUs to drive as the fps are lower than older games.

Doom112 · 27 Jul 2016 at 11:32

CAT-THE-FIFTH said:
Its annoying Nvidia made such a big song and dance about it and months later nothing despite them making more of a song and dance about it earlier than AMD.

Bethesda and iD are generally more Nvidia focussed devs anyway,as they have in the past tended to work a bit closer with Nvidia.

But being Nvidia it will be swept under the carpet until Volta just like with the async patch which some promised here for Maxwell for months.

Yes it is nvidia fault that even on with async AMD still cannot manage to beat nvidia. It is nvidia fault that they are running better on low end.

Kaapstad · 27 Jul 2016 at 11:34

This is how I think these tech sites should test out CPU performance.

ROTTR maxed
1080p
GTX 1080 FE stock
DX11

6950X @4.0
2 cores, no H/T, Total 2 threads

6950X @4.0
10 cores, With H/T, Total 20 threads

Doing the above the only thing that changes in the system is the active cores/threads.

The first example above is lower speced CPU wise than almost any CPU available today.

The second example is at the other extreme.

There is very little difference between the overall scores.

The minimums on the first example have taken a bit of a dive though.

The way tech sites test also means they are testing motherboards, memory, SSDs and windows installations in addition to the CPU, all of which can skew the results.

Doom112 · 27 Jul 2016 at 11:41

Kaapstad said:
This is how I think these tech sites should test out CPU performance.

ROTTR maxed
1080p
GTX 1080 FE stock
DX11

6950X @4.0
2 cores, no H/T, Total 2 threads

6950X @4.0
10 cores, With H/T, Total 20 threads

Doing the above the only thing that changes in the system is the active cores/threads.

The first example above is lower speced CPU wise than almost any CPU available today.

The second example is at the other extreme.

There is very little difference between the overall scores.

The minimums on the first example have taken a bit of a dive though.

The way tech sites test also means they are testing motherboards, memory, SSDs and windows installations in addition to the CPU, all of which can skew the results.

I think you are doing a bit wrong. You need disable AA so that the GPU bottleneck is out of the way and CPU bottleneck comes in.

tommybhoy · 27 Jul 2016 at 11:53

Doom112 said:
Yes it is nvidia fault that even on with async AMD still cannot manage to beat nvidia. It is nvidia fault that they are running better on low end.

Am I missing something as looking at comparable gpu's in those results, AMD is faster, in fact the 280X(7970 rebrand) is smashing the 780Ti's back doors in?

Doom112 · 27 Jul 2016 at 11:56

tommybhoy said:
Am I missing something as looking at comparable gpu's in those results, AMD is faster, in fact the 280X(7970 rebrand) is smashing the 780Ti's back doors in?

Yes Kepler are really murdered in Vulkan.

Rroff · 27 Jul 2016 at 11:58

tommybhoy said:
Am I missing something as looking at comparable gpu's in those results, AMD is faster, in fact the 280X(7970 rebrand) is smashing the 780Ti's back doors in?

Kepler isn't running right at all - most people seem to have forced V-Sync with 780 series and Vulkan in Doom regardless of whether you turn it off or not - for some reason a few people don't though :S

If you look at my screenshots in one of the other threads my 780GHZ on out the box clocks is running 1.5% behind a RX480 in Open GL but drops back around 25-30% when Vulkan is enabled from its Open GL results never mind the 480.

EDIT:

The 369.81 drives bumped the Vulkan performance up a bit but it is still way down on Open GL never mind gains.

Kaapstad · 27 Jul 2016 at 14:00

Doom112 said:
I think you are doing a bit wrong. You need disable AA so that the GPU bottleneck is out of the way and CPU bottleneck comes in.

Err why

I am averaging 60fps on a 60htz monitor.

ROTTR looks bad enough even when maxed out so why would I want to turn the settings down.