ATi 4870X2(R700) PCB Debut

Printable View

Show 100 post(s) from this thread on one page

07-13-2008, 02:37 AM
hias

well, i don't think the cpu will be the bottleneck in crysis at 46fps, but it's not impossible.

Sampsa, can you please try crysis with frametimes somewhere between 33 and 40ms?
07-13-2008, 04:16 AM
shiznit93

Quote:

Originally Posted by Chickenfeed

Were there ever any CRTs that could even do 120Hz above 1600x1200? I'll take higher resolution and slower performance than what a good CRT would do for me any day but thats just me. Could never go back from 1920x1200 :cool:

My FW900 does 1600x1000@110hz easily, and 1920x1200@95hz if I really push it. But 1440x900@120hz or 1280x800@120-140hz is the g spot for me.
07-13-2008, 07:13 AM
fornowagain

nvm
07-13-2008, 11:47 AM
Sampsa

Ok, I found a new game where 3870 X2 shows very bad microstuttering and it is Race Driver: GRID.

I was checking my benchmark results for R700 preview I'm working on and noticed avg FPS was very bad with 3870 X2. So I immediately applied my previous testing methods using Fraps to log frametimes and analyzed results with Excel:

3870 X2 shows similar behavior in Race Driver: GRID than I noticed in Crysis but it seems to be even worse. In Crysis 3870 X2 rendered every other frame after ~21,5 ms and every other after ~49,5 ms. In Race Driver: GRID 3870 X2 rendered every other frame after ~24,9..27 ms and every other frame after ~40,2..42,4 ms.

I repeated the test with R700 and it showed again very stable rendering times between every frame: ~15,3 - 17,6 ms.

I'll continue and test Race Driver: GRID with 4870 CF and 9800 GX2. I've also asked about this issue from AMD and hoping to get some of my questions answered in the beginning of next week.
07-13-2008, 11:56 AM
Sampsa

I have also noticed a lot of people are very skeptical about my tests. I don't claim to be an expert in this case and all these results are based on my own tests with simple tools and methods. I'm not even 100% sure if this issue with 3870 X2 in Crysis and Race Driver: GRID is actually microstuttering, lag or some other problem. I asked AMD about this and hopefully get an answer.

I have to return this sample tomorrow but imho I've deserved couple more from AMD ;)
07-13-2008, 12:12 PM
Lestat

Sampsa i didn't see what board you were using but unless you are using a TRUE 16x/16x board the 2nd PCI-E slot is only going to be 8x,, some even 4x,, and you could and probably are getting more issues because of that, than simply the cards themselves choking...

would you agree ?
07-13-2008, 12:20 PM
Sampsa

Quote:

Originally Posted by Lestat

Sampsa i didn't see what board you were using but unless you are using a TRUE 16x/16x board the 2nd PCI-E slot is only going to be 8x,, some even 4x,, and you could and probably are getting more issues because of that, than simply the cards themselves choking...

would you agree ?

I'm using Asus P5E64 WS Evolution which is based on Intel X48 chipset (2 x PCI Express x16 2.0 slots)
07-13-2008, 12:29 PM
AriciU

Quote:

Originally Posted by Sampsa

I'm using Asus P5E64 WS Evolution which is based on Intel X48 chipset (2 x PCI Express x16 2.0 slots)

When is the NDA over with and you are allowed to post some results? Maybe you could answer a question if you had 2 x 4870x2's. Does CF scale well with them or not?
07-13-2008, 12:31 PM
Sampsa

Quote:

Originally Posted by AriciU

When is the NDA over with and you are allowed to post some results? Maybe you could answer a question if you had 2 x 4870x2's. Does CF scale well with them or not?

You'll notice it when you see previews online. I won't post any benchmark numbers, stop asking.
07-13-2008, 12:33 PM
hias

can you confirm that R700 is microstutter-free when running crysis at 25-30fps (33.3-40ms frametime) ?
07-13-2008, 12:35 PM
annihilat0r

The real question is whether 4870x2 can scale in situations 4870CF can't
07-13-2008, 12:44 PM
Sampsa

Quote:

Originally Posted by hias

can you confirm that R700 is microstutter-free when running crysis at 25-30fps (33.3-40ms frametime) ?

In my opinion that is too low framerate to actually play the game smoothly anyway.
07-13-2008, 01:22 PM
cadaveca

Thanks for taking the time out to address this issue Sampsa, It's greatly appreciated.
07-13-2008, 01:37 PM
Arseface

Quote:

Originally Posted by Sampsa

I have also noticed a lot of people are very skeptical about my tests. I don't claim to be an expert in this case and all these results are based on my own tests with simple tools and methods. I'm not even 100% sure if this issue with 3870 X2 in Crysis and Race Driver: GRID is actually microstuttering, lag or some other problem. I asked AMD about this and hopefully get an answer.

I have to return this sample tomorrow but imho I've deserved couple more from AMD ;)

Hi

I wrote a little tool which quantifies the degree of microstutter from a given FRAPS benchmark. I don't claim it's foolproof, but it should give a roughly consisent approximation of the degree of MS, which scales does not scale with the absolute value of framerate (a non-dimensional index)

You can grab it at: http://www.ubern00b.net/microstutter.rar

The readme explains how to use it and exactly how the index is calculated. But in short, the index is the average percentage varation away from the stabilised local framerate.

An index of less than 8 ish seems typical for single GPU setups. With multi GPU setups we have seen some values as high as 40, although sometimes they can be of similar magnitude to the single GPU results. I think it may be dependent on the rendering mode which is used.

I'm very interested in microstuttering wrt the 4870x2 also, especially given its closer communication between GPUs via the xfir sideport :)
07-13-2008, 01:40 PM
Arseface

Quote:

Originally Posted by Sampsa

In my opinion that is too low framerate to actually play the game smoothly anyway.

Normally I would say agree, but Crysis is a strange one...

I think the effectiveness of its motion blur smooths things along, as it seems very playable at any framerate over about 30fps. Normally I wouldn't be able to bring myself to play any shooter which averaged less than 60 and/or dipped below about 40fps regularly.
07-13-2008, 01:41 PM
flopper

Quote:

Originally Posted by Sampsa

I have also noticed a lot of people are very skeptical about my tests. I don't claim to be an expert in this case and all these results are based on my own tests with simple tools and methods.

Never read people´s opinions when you get hardware for testing.
To many are just full of :banana::banana::banana::banana:.

Look forward your preview.
07-13-2008, 01:43 PM
annihilat0r

In Crysis when I enable Vsync it's sooo smooth you wouldn't believe it, alas when FPS drops to 40's the input lag makes things unplayable.
When I disable Vsync micro stuttering is the most noticeable thing ever. It's not really stuttering but the smoothness is totally gone.
07-13-2008, 01:45 PM
Arseface

Quote:

Originally Posted by annihilat0r

In Crysis when I enable Vsync it's sooo smooth you wouldn't believe it, alas when FPS drops to 40's the input lag makes things unplayable.
When I disable Vsync micro stuttering is the most noticeable thing ever. It's not really stuttering but the smoothness is totally gone.

Give triple buffering a go :)
07-13-2008, 01:54 PM
annihilat0r

Triple buffering + SLI = No go
They're mutually exclusive :P
07-13-2008, 02:01 PM
Arseface

Quote:

Originally Posted by annihilat0r

Triple buffering + SLI = No go
They're mutually exclusive :P

Ah, okay. Furry muff.

That makes sense though, I suppose. In AFR mode at least.
07-13-2008, 03:35 PM
bill_d

Quote:

Originally Posted by annihilat0r

Triple buffering + SLI = No go
They're mutually exclusive :P

hmm i think you can in cf never tried it, at 2560x1600 i thought it would slow it down too much
but the check box is still available in cf,but i'm not putting my 3870's back in to see
07-13-2008, 03:41 PM
shiznit93

So you can't use triple buffer in SLI? Well that's pretty much a deal breaker for me if gaming on LCD. Can someone confirm that triple buffer works in Crossfire? I use D3DOverrider to force triple buffer in games that do not support it and there is no way I will go multi-gpu if it can't be used.
07-13-2008, 03:53 PM
NapalmV5

sampsa, can you test 2 x 4780x2 ?

if ms fixed @ dual gpu/card.. does it help @ dual x2 card ?

thanks
07-13-2008, 04:23 PM
Arseface

Quote:

Originally Posted by bill_d

hmm i think you can in cf never tried it, at 2560x1600 i thought it would slow it down too much
but the check box is still available in cf,but i'm not putting my 3870's back in to see

The great thing about triple buffering is that there is (virtually) no performance hit. Well, you can't go above the monitor refresgh rate of course, but apart from that...

Use D3Doverrider (an application that comes with rivatuner) to force triple buffering, rather than going through the control panel.

Not sure about x-fire or SLI. Would be nice if someone could test it.
07-13-2008, 04:54 PM
Cooper

Quote:

Originally Posted by annihilat0r

In Crysis when I enable Vsync it's sooo smooth you wouldn't believe it, alas when FPS drops to 40's the input lag makes things unplayable.
When I disable Vsync micro stuttering is the most noticeable thing ever. It's not really stuttering but the smoothness is totally gone.

Same happens with Counter-Strike 1.6 and single GPU on LCD screen :D
Vsync on = smooth

Show 100 post(s) from this thread on one page