Sign in to follow this  
Followers 0
lutorm

1.31 FPS tests and CPU vs GPU

103 posts in this topic

Your Processor will also be causing you lower framerates. How much lower I'm not sure' date=' but I wouldn't be surprised if that's most of your loss.[/quote']

A difference of 200 mhz is halving my framerate?

Share this post


Link to post
Share on other sites
I need a signal boost.

Is the issue here that ATi cards are still not performing up to snuff?

Yes and not only that, relatively lowly card like my Radeon 4830 is outperforming chimaeras 4870x2 which is just plain wrong. Needless to say ATI cards are being comprehensively beaten by just about any nVidia card.

Edit: Out of curiosity, why would you write "still not performing up to snuff"? My understanding was that they never were.

Edited by amorosa

Share this post


Link to post
Share on other sites

Wow, can't believe this thread is still alive!

I need a signal boost.

Is the issue here that ATi cards are still not performing up to snuff?

That might be an issue, though for me most low-fps situations are due to the cpu limit. Unlike most modern games that are solidly gpu-limited, this game seems like a prime candidate for multithreading. The payoff (if it was possible to efficiently split up the workload) would be huge.

This is the problem with the way performance has been going up, by more cores. Not at all all tasks are helped by this. Unlike graphics cards which have seen a massive increase in performance for massizely data-parallel problems like pixel rendering.

Share this post


Link to post
Share on other sites
I need a signal boost.

Is the issue here that ATi cards are still not performing up to snuff?

Yes. From tests we have seen on close to identical machines (except GFX) that the higher end ATI cards are performing awfully on some of the benchmarks.

One I have seen very often from posters is the Antwerp Bench.

For example my I7 stock will beat other I7's with same OS that have 5770's, 4850's, 70's, 90's and even 58500 series.

I'm getting 46FPS on stock (with old 192 core GTX 260) they are getting 20 something.

I am sure from the bench tests we did you'll have plenty of better data on this but just check through this thread and see what people on Nvidia are posting and People on higher end ATI's - They are worlds apart with same or better setups on the ATI side.

Share this post


Link to post
Share on other sites
Wow, can't believe this thread is still alive!

That might be an issue, though for me most low-fps situations are due to the cpu limit. Unlike most modern games that are solidly gpu-limited, this game seems like a prime candidate for multithreading. The payoff (if it was possible to efficiently split up the workload) would be huge.

This is the problem with the way performance has been going up, by more cores. Not at all all tasks are helped by this. Unlike graphics cards which have seen a massive increase in performance for massizely data-parallel problems like pixel rendering.

Partially right but what we are seeing and hence the thread is that the CPU is running flat out even I7 processors. The GFX cards are runnign at 40, 50 or 60%.

Even so with the GFX NOT getting stressed the same setup with an ATI is scoring much lower then a similar Nvidia card. Both cards I bet are not getting pushed but something slows down a machine with ATI card.

Share this post


Link to post
Share on other sites

The best way to test if there's something going awry with ATI cards is to test in a situation that is GPU bound. If it's not GPU bound, you're not measuring the relevant quantity. The implication is that you should test when looking through the binos at the speedtrees. This seems to be so gpu bound that it basically won't even matter what CPU you have, the fps should be solidly determined by what your graphics card can do.

Here's my arbitrarily defined test case:

  • Set maximum quality, shadows high, all effects on, vsync off, res 1920x1200.
  • Spawn a british rifle offline, go to the poles by the doors between the spawn buildings and look down the road (like in the perf test). Your view should look like this:
    binotest2.jpg
  • Switch to binos and observe fps. It's sensitive to the exact area you look at, so try to replicate the same view as below:
    binotest1.jpg
  • Note the fps. As you can see in the screenshot, my 4870 using Catalyst 10.2 gives 45 fps here.

If we get some reports of both ATI and Nvidia cards we should be able to determine if the ATI cards underperform relative to their theoretical performance. Note the driver version, too.

Share this post


Link to post
Share on other sites
Even so with the GFX NOT getting stressed the same setup with an ATI is scoring much lower then a similar Nvidia card. Both cards I bet are not getting pushed but something slows down a machine with ATI card.

Agreed, that's weird. You have an nvidia card, right? Want to try the test I just posted?

Share this post


Link to post
Share on other sites
The best way to test if there's something going awry with ATI cards is to test in a situation that is GPU bound. If it's not GPU bound, you're not measuring the relevant quantity. The implication is that you should test when looking through the binos at the speedtrees. This seems to be so gpu bound that it basically won't even matter what CPU you have, the fps should be solidly determined by what your graphics card can do.

Here's my arbitrarily defined test case:

  • Set maximum quality, shadows high, all effects on, vsync off, res 1920x1200.

  • Spawn a british rifle offline, go to the poles by the doors between the spawn buildings and look down the road (like in the perf test). Your view should look like this:

    binotest2.jpg

  • Switch to binos and observe fps. It's sensitive to the exact area you look at, so try to replicate the same view as below:

    binotest1.jpg

  • Note the fps. As you can see in the screenshot, my 4870 using Catalyst 10.2 gives 45 fps here.

If we get some reports of both ATI and Nvidia cards we should be able to determine if the ATI cards underperform relative to their theoretical performance. Note the driver version, too.

Looking at speedtrees (the densest pocket I could find) and looking through binocs, I got 45 FPS and 100% GPU usage.

That doesn't change the fact that looking at the antwerp Benchmark I get 20 FPS and 0% GPU usage (until I press ~ to bring up my mouse, then it goes to 20ish%, about the same used for the menu screens)

*edit* with 10.2 drivers on a 4850

Edited by dkamerad

Share this post


Link to post
Share on other sites
Agreed' date=' that's weird. You have an nvidia card, right? Want to try the test I just posted?[/quote']

Well seems things are ok using this test. For example in the first SS I'm getting 99 - 103 FPS with a stock I7 920 2.66ghz

GOing to bino view I'm only getting 48 FPS.

******* Correction - Sorry never set it to max detail (best Quality) with everything maxxed. Only getting 60 and 30 FPS with everything up and on. so that sort of takes it away from any graphical performance. ************

The Antwerp test seems the killer for many similar systems - My rig is beating I7's clocked at 3.5ghz etc and with much better cards so somethings going on.

Pulled some peoples reports here. For reference I'm gettign 46 FPS on antwerp on stock I7 with old 260GTX

-------------

Just for comparison chimaera. My i7 920 on stock (asus p6t deluxe v2, 6gb 1600 Ram), radeon 4830, W7 64, game version 1.31.25, catalyst 10.1, 1280x1024 resolution:

remagen - 70fps

antwerp - 34fps

vehicles - 28fps

-----------------------------------------------------------------

Windows7 Enterprise(64bit)

i7-920 (stock, no O/C)

ATI 4870 X2

6 GB DDR3 RAM

1680x1050x32

.benchremagen

62fps

.benchvehicles

31fps

.benchantwerp

29fps

------------------------------------------------------------------

(CPU: Intel E8500 running at 4.18GHz, GPU: ATI 4870):

4.18GHz: 67 / 32 / 35 FPS for benchremagen/vehicles/antwerp

---------------------------------------------------------

Here are my specs:

Core I7 @ 3.8Ghz

6 Gigs DDR3 Triple Channel Ram @ 1600

2 x 5770 in Xfire

Intel X25 SSD 80Gig drive (Game Installed)

2 Seagate Drives in Raid 0 for 250Gigs total

I run the Offline test as per the forum post. My .benchantwerp gives me 40 FPS.

---------------------------------------------------------------------

I just test again with my new computer

I5 750 + ATI 5770+ 4Go+ Win7 64

my old computer is

E6600 +GF7950Gt+ 2Go +XP

strange result to confirm there is an issue with ATI card (or drivers ?) or win 7 perhaps.

--------------New comp------------old--

antwerp:--------18 ----------------- 26

remagen:---- 26 ------------------- 32

vehicles:----- 16 ------------------ 19

Edited by badger77

Share this post


Link to post
Share on other sites
Well seems things are ok using this test. For example in the first SS I'm getting 99 - 103 FPS with a stock I7 920 2.66ghz

GOing to bino view I'm only getting 48 FPS.

******* Correction - Sorry never set it to max detail (best Quality) with everything maxxed. Only getting 60 and 30 FPS with everything up and on. so that sort of takes it away from any graphical performance. ************

Ok, your GTX260 gets 30 and my 4870 gets 45. That's 50% better for me. I just looked through some old game benchmarks (generally gpu bound) and at 1920x1200, the 4870 generally was 10-25% faster than the GTX260. So here the ATI card is actually doing comparatively better than it should.

I have a hunch, but we need more data. If we don't see much difference between ATI and Nvidia in the binos test, where (hopefully) the gpu is really the limit, that would point to a difference in cpu performance, i.e. drivers.

What I mean is that, if the ATI cards are normally fast in situations that are really GPU bound but lags in framerate in situations that are CPU bound, the only way to explain that is if it takes much more CPU to render a frame on an ATI than an Nvidia card. That means it's the drivers.

Can some of the guys that see ridiculous slowdowns with ATI cards please try this.

Share this post


Link to post
Share on other sites

What I mean is that, if the ATI cards are normally fast in situations that are really GPU bound but lags in framerate in situations that are CPU bound, the only way to explain that is if it takes much more CPU to render a frame on an ATI than an Nvidia card. That means it's the drivers.

.

which would make sense for the Antwerp benchmark since that seems to be pretty CPU intensive but not graphically much going on.

Share this post


Link to post
Share on other sites

Just done your test lutorm. All ingame settings on max, all catalyst settings on best performance. Game version 1.31.24 offline. Catalyst 10.1. My machine specs are in the post on page 2. Looking at trees without binos 73fps. Looking at trees with binos 73fps. Thats right, I haven't lost any fps whatsoever!

Edited by elegance

Share this post


Link to post
Share on other sites

Here's my arbitrarily defined test case:

  • Set maximum quality, shadows high, all effects on, vsync off, res 1920x1200.

  • Spawn a british rifle offline, go to the poles by the doors between the spawn buildings and look down the road (like in the perf test). Your view should look like this:

    binotest2.jpg

  • Switch to binos and observe fps. It's sensitive to the exact area you look at, so try to replicate the same view as below:

    binotest1.jpg

  • Note the fps. As you can see in the screenshot, my 4870 using Catalyst 10.2 gives 45 fps here.

Doing this test I actually get an increase in FPS when going to binos. From 80 normal to 85 with binos. All settings on ingame, CCC to performance. 5770 and 10.3 drivers.

Share this post


Link to post
Share on other sites

Performing Lutorm's test, I get a gain of 7 FPS when looking through binos (53 to 60 FPS).

Post of full system specs again.

AMD Athlon 2 x 2.8 GHZ (Windsor Core)

XFX Raoden 4850 512 MB with catalyst 10.2

4 Gigs ram

Windows 7 64 bit

And the more effects I turn on, the greater the disparity gets... from 35-50 if I enable Post-rendering, normal maps, etc.

Edited by dkamerad

Share this post


Link to post
Share on other sites
Doing this test I actually get an increase in FPS when going to binos. From 80 normal to 85 with binos. All settings on ingame' date=' CCC to performance. 5770 and 10.3 drivers.[/quote']

Munchkin, what cpu you got?

Share this post


Link to post
Share on other sites
Just done your test lutorm. All ingame settings on max' date=' all catalyst settings on best performance. Game version 1.31.24 offline. Catalyst 10.1. My machine specs are in the post on page 2. Looking at trees without binos 73fps. Looking at trees with binos 73fps. Thats right, I haven't lost any fps whatsoever![/quote']

Ingame settings on max quality, you say? Res 1920x1200? Something's weird here. Your 4830 should not be able to get 73 fps with binos.

Share this post


Link to post
Share on other sites
Performing Lutorm's test, I get a gain of 7 FPS when looking through binos (53 to 60 FPS).

Post of full system specs again.

AMD Athlon 2 x 2.8 GHZ (Windsor Core)

XFX Raoden 4850 512 MB with catalyst 10.2

4 Gigs ram

Windows 7 64 bit

And the more effects I turn on, the greater the disparity gets... from 35-50 if I enable Post-rendering, normal maps, etc.

The settings for the test should be with ingame settings on max quality, shadows high, all effects enabled. Are you saying that's 35 without binos, 50 with? And 1920x1200?

Share this post


Link to post
Share on other sites

Yea we all need to settle on a resolution that everyone can benchmark on.

I'm lucky enough to have a 1920x1200 monitor, but i know everyone isnt

What about something like 1660x900, Probably wont make much of a difference on high end rigs if its at 1660x900 or 1920x1200. But i guess for the sake of trying to see whats faster then what. everyone should settle on the same resolution.

?

Share this post


Link to post
Share on other sites

I'm lucky enough to have a 1920x1200 monitor, but i know everyone isnt

What about something like 1660x900, Probably wont make much of a difference on high end rigs if its at 1660x900 or 1920x1200. But i guess for the sake of trying to see whats faster then what. everyone should settle on the same resolution.

Yeah, this is a problem but it seems the people that have complained most loudly about the broken ATI performance have 5xxx cards and you'd think that then you can run at 1920x1200, too. And we need the test to be GPU-limited, and lowering the resolution will lower the GPU workload. For this reason I'd prefer to stick with 1920x1200 for now. Hopefully we can see a pattern without too many people.

But it's absolutely essential the resolution be the same. 1660x900 only has 64% as many pixels as 1920x1200, so in a GPU-limited situation, the framerate will be 50% higher.

Edited by lutorm

Share this post


Link to post
Share on other sites
The settings for the test should be with ingame settings on max quality' date=' shadows high, all effects enabled.[/b'] Are you saying that's 35 without binos, 50 with? And 1920x1200?

I can only go up to 1920x1080.

Using max quality, 35 without binos, 50 with.

Share this post


Link to post
Share on other sites

Benchmarking comparisons between a Nvidia Geforce GTX 260(896 MB) and ATI Radeon HD 4870(1024 MB) shows they are very similar in performance.

Thats why I bought my 4870x2 card, it was at the time a less expensive card than the GTX260 while still outperforming it.

Tom's Hardware - Benchmark The Last Remnant

Why do I link to a benchmark between ATI 4870(1024 MB) vs GTX 260(896 MB)?

Well I figured as BGE doesnt utilize both GPU's on my x2 card it should be equal to a ATI 4870(1024 MB).

The benchmark shows that in 'mainstream' applications/games the cards are very similar in performance, some are 'won' by ATI and some are 'won' by nVidia.

When it comes to testing in BGE we get a picture that shows that nVidia has some kind of advantage, what it is I dont know.

If I would be buying a computer today I would not go for a ATI card which is sad as my latest 3 computer builds has been with ATI cards.

And yes I saw those recent 'binos' tests where ATI seems to have some kind of advantage, interesting...

It could very well be that there actually is no problem, maybe ATI and nVidia is using VERY different ways to handle graphics.

But I hope that with the recent input from us and the test results that has been posted during Beta that the Rats might actually be able to find some unknown bug that hasnt been squashed yet. For example, why is the ATI cards idling at 0% load during some tests? As if the CPU is doing all the work...

Edited by chimaera

Share this post


Link to post
Share on other sites
Ingame settings on max quality' date=' you say? Res 1920x1200? Something's weird here. Your 4830 should not be able to get 73 fps with binos.[/quote']

1280x1024. Its in the post on page 2. Cant do anymore, monitor limited.

Maybe we should throw in some AA and AF to really stress the card and make the test relevant at lower resolutions or otherwise I wont be able to help you.

Specs again: i7 920@stock, 6Gb 1600, Asus P6T Deluxe V2, W7 64, Radeon 4830, Catalyst 10.1 (performance settings), 1280x1024, Game version 1.31.24 offline

Edited by amorosa

Share this post


Link to post
Share on other sites
I can only go up to 1920x1080.

Using max quality, 35 without binos, 50 with.

Ok, the bino situation is so gpu bound for me it works to just rescale by the # of pixels. That would mean you'd get 45 on 1920x1200, which is in line with my 4870.

Using the same method, elegance with 73fps @ 1280x1024 would have 42 @ 1920x1200 with his 4830.

It seems all the 48xx cards are doing about the same, with small differences 4830->4850->4870.

Munchkin's 5770 is killing us with 85 in bino view, basically twice the performance of the 48xx cards.

badger77's GTX260 is doing about what's expected or a little worse with 30fps given that card in relation to the 48xx.

What we need now is some high-end Nvidia cards. Anyone?

Share this post


Link to post
Share on other sites

Well I figured as BGE doesnt utilize both GPU's on my x2 card it should be equal to a ATI 4870(1024 MB).

...

And yes I saw those recent 'binos' tests where ATI seems to have some kind of advantage, interesting...

It could very well be that there actually is no problem, maybe ATI and nVidia is using VERY different ways to handle graphics.

But I hope that with the recent input from us and the test results that has been posted during Beta that the Rats might actually be able to find some unknown bug that hasnt been squashed yet. For example, why is the ATI cards idling at 0% load during some tests? As if the CPU is doing all the work...

chimaera, can you do the bino test, with both one and two cards? If you're ever to see a difference using CrossFire, that would be the one.

And 0% simply means that the scene is fully CPU-limited. Yeah, the "0" should probably be "a few", but given my results, if you have a high-end CrossFire setup, it really could be 1% in the most CPU-limited cases.

Share this post


Link to post
Share on other sites
chimaera, can you do the bino test, with both one and two cards? If you're ever to see a difference using CrossFire, that would be the one.

And 0% simply means that the scene is fully CPU-limited. Yeah, the "0" should probably be "a few", but given my results, if you have a high-end CrossFire setup, it really could be 1% in the most CPU-limited cases.

I'll do it as soon as I get home from work.

Share this post


Link to post
Share on other sites
Sign in to follow this  
Followers 0

  • Recently Browsing   0 members

    No registered users viewing this page.