• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Is Eyefinity + Crossfire Possible?

Rock&Roll said:
I just got around to testing that. And I was surprised. Running BC2 at 6028x1200 (Bezel compensated 5760x1200), All settings turned high, HBAO off, 1x AA, 16x AF. vsync off 10.3B drivers.

Most Testing was done on an empty server, so as to maintain some consistency in results. Though, these are not real benchmarks.....

@ 5040x1050 Single card, averaging 55-60fps. No stutter

@ 6028x1200 Single card, averaging 34-40fps. No stutter. But not smooth.....

@ 5040x1050 Dual card, averaging around 75 fps in high-detail areas (Lowest FPS I could get). I could detect no stutter.

@ 6020x1200 Dual card, averaging around 63 fps in high detail areas (Lowest FPS I could get). I could detect no stutter (surprised)


The experience has remained mostly the same when I play on a 24 or 16 man server that doesn't lag. FPS will dip (about 5-8 FPS less than other times) in instances where there's a lot of smoke and debris from explosions.

I'm quite surprised, what with almost everyone having problems with BC2 in terms of dual-card performance as well as with eyefinity. I am having absolutely zero issues with BC2 running xfired-eyefinity.
Click to expand...

What DirectX are you running in these? Did you try re-running with vsync turned on?
 
As noobish as it sounds, I completely forgot to check the DX level.

And, vsync on a MP FPS would be a bad idea anyhow. Say what you want about tearing. But, vsync fucks up aiming.
 
Rock&Roll said:
As noobish as it sounds, I completely forgot to check the DX level.

And, vsync on a MP FPS would be a bad idea anyhow. Say what you want about tearing. But, vsync fucks up aiming.
Click to expand...

Yeah, I can see how vsync would be bad for aiming but vsync makes everything look better/smoother to me.
 
lowres said:
Contary to my earlier post, rFactor is indeed still crap with CF+EF, just like every other game i have attempted to play doesn't matter if the frames are over 60, still have stuttering. I'm at a loss of why it could be so smooth on 1 screen and total garbage on 3, i'm also at a loss as to why a few people are doing just fine. Unless it really does come down to PCIe speed.
Click to expand...

I too am at a loss why 1 screen is perfect and 3 screens with CF suck. In my mind x8 bandwidth limitation is ruled out as the culprit because the same stuttering mess happens with a 5970.
 
Hrm. I'll bust out rFactor and see if it happens to me.

Edit:

Done. I wasn't running any mods for the game. Just default out of the box. All settings to max, with 16x MSAA forced in the drivers. Still was getting 100fps out on the track. No stutter.

I'm not so sure the 8x thing can be ruled out at this point. I guess I could swith one of my cards to an 8x slot on my mobo....but...I'm not sure I'm willing to do all that goofing around just to find out. http://www.techpowerup.com/reviews/AMD/HD_5870_PCI-Express_Scaling/25.html
 
Last edited:
I get without crossfire 74 fps in BC2, with the same settings with crossfire i get around 23. This is in eyefinity.

Without eyefinity I can do crossfire with max settings and get 119 fps.

This needs to be fixed really soon.....
 
Not that I'm suggesting anyone subscribe to this idea...But, i personally have to wonder if people are having eyefinity + crossfire problems due to a saturated PCI-E bus @ 8x. A single 5870 at 8x saturates the bus. Dual would be just as bad or worse.

It would be nice if ATI could give some definitive answer to the 16x/16x vs 16x/8x vs 8x/8x question.
 
My poor EF+CFX performance was on a Skulltrail which is supposed to have full 16x lanes. I didn't try switching different slots, though. Just used the top 2 slots by default.
 
Rock&Roll said:
Not that I'm suggesting anyone subscribe to this idea...But, i personally have to wonder if people are having eyefinity + crossfire problems due to a saturated PCI-E bus @ 8x. A single 5870 at 8x saturates the bus. Dual would be just as bad or worse.

It would be nice if ATI could give some definitive answer to the 16x/16x vs 16x/8x vs 8x/8x question.
Click to expand...

I too thought it was x8/x8 bus saturation that was causing stuttering with my CFX 5870s until I tried 1 5970 on a x16 slot and had the same result.

I'm sure x16/x16 in eyefinity will give higher fps but I don't think x8/x8 is the cause of the CFX + EF stuttering mess.
 
I can't jump to that conclusion based on 5970.

Data drops and retransmissions happen on any digital transmission system. I'm sure PCI-E isn't much different. If your bus is saturated and you need to retransmit data, there's a longer waiting line for that round trip.

I'm almost certain a saturated bus would have more influence on performance than FPS alone.
 
The saturated bus speculation rests on shaky ground. Tests on the P55 and X58 platforms revealed miniscule performance differences between x8/x8 and x16/x16 using SLI GTX 280s; moreover, these differences did not widen with increased resolution.

A simple way of verifying the problem may be to force supersampling (does ATI software allow this?). Since supersampling effectively multiplies the resolution to achieve anti-aliasing, it is akin to Eyefinity in the sense that you'll dramatically increase resolution.. If it produces the same input lag that you're seeing in Eyefinity, then maybe there's something to the claim.


If nothing else, I suppose we'll have the clearest proof when we find out whether Nvidia users have similar problems when Nvidia releases its equivalent three-monitor solution this month or next.
 
Borrowing this from another thread:

Even with a GTX 480 [x8/x8 performance decrease] is still minimal (1-2%)
http://www.techpowerup.com/reviews/N...caling/24.html
http://www.techpowerup.com/reviews/N...xpress_Scaling
Click to expand...

So my money is on no. As an [H] review some time ago pointed out, nothing has changed since at least the Geforce 8800 GTX days when it comes to PCI-E 2.0 8x and GPUs. So now we have three generations of video cards that have not maxed out this bus, that have all produced the same negligible 1-3% differences, even though resolutions have gone up from 1280x960 to 2500x1600. Therefore it is baseless to say that Eyefinity is an exception.
 
I think that this is the biggest problem with Eyefinity at the moment. F**k the bezels, it's this choppiness and lack of vsync that kills Eyefinity for me. I wish Kyle would either do some article or speak to ATI on this as it's a deal breaker, imo.

Everything just feels so glitchy and at times I'm not sure the reduced video quality is worth the larger aspect ratio. Crossfire and Eyefinity just sucks and you really need it for anything released since last year.
 
Well I wonder if it's an issue for Kyle since he uses a 5970 to drive his eyefinity setup. Surely he would notice the choppiness if he was affected by it. And I understand he started out with a 5870 so going from 1 to 2 GPU eyefinity I'm sure he'd notice it.

Kyle would you care to comment?
 
As far as the PCIe x8 vs x16 arguement goes, i picked up an i7+motherboard combo at frys this weekend. At the same CPU clocks (3.8) as the i5, and the same 2 5870s, the unlimited frame rates are about the same (in some cases slightly lower) across my portfolio of games. The minimum frames improved by about 30-40%. The Heaven 2.0 benchmark scores improved considerably (+300>) on one and 3 monitors (6000x1080 eyefinity at 4xaf 0xaa is a slideshow even though fps is acceptable). So stuttering while crossfired is still present with eyefinity, however maybe not quite as bad. Still bad enough though. The only game that was immediately better was FSX, that was much smoother on the 1366 platform, even running crossfire (in dx10). Make of that what you will...
 
x58 does not in itself guarantee 16x/16x lanes. That varies motherboard to motherbaord, slot to slot. Check your manual/mobo forums to confirm how many lanes your cards are really seeing.
 
Ok, I checked out your mobo. So long as your cards are in the top two PCI-E slots, you are 16x/16x. The 3rd slot is only 4x.


So, this basically eliminates the PCI-E bus as the root of the issue. Which brings up a memory I have from the 4870 I had.

The 4870 I bought had 512MB of RAM. @ 1920x1200, framerates were grand, but games still stuttered. I eventually gave up and got a GTX280.

So, one now has to ask if the 1GB framebuffer isn't the limitation. Or maybe it's a memory controller issue when using above a certain % of the vRAM.

Does your rig stutter even in old games like HL1?
 
Rock&Roll said:
Does your rig stutter even in old games like HL1?
Click to expand...

The problem remains even in old games with low rez/no AA. In fact, Source engine games are particularly bad in Crossfinity. TF2 is practically unplayable until you disable crossfire. The framerates are in the hundreds, of course.
 
HL1 isn't source. Mostly Quake 2

I do agree with you about source games though. Though, I've been able to attribute the probelm more whit DX9 rendering in that engine. DX8 makes things buttery smooth. But then, also negates the benefits of crossfire.
 
Yes, top 2 slots.

I'll have to find HL1, not sure i even know if i still have it anymore.

To expand on a couple of things, when i said rFactor was not so good, i think i was making the wrong point. The situations where stutter began were definitely cpu related, eg lots of cars in view, without them the fps an play were indeed very smooth. I would say that rFactor is of a similar age (in terms of graphics engine) to HL1, but can try it out.

If your interested in the rFactor test i did (for comparison), go into controls > rules > number of opponents set 30, then on same screen start position to 31, finally go into display options set number of visible cars to over 30, everthing else maxed out. Enter a race (hit next session until race appears) hit pause, and check frames, un pause, see what you get in terms of jitters/frames. On the i5+CF+EF i would get about 30 very jumpy and unplayable fps.

In one game i have found that with CF+EF where i would normally get stutter it can be reduced by x16aa+af, capping frames at 60 (not vsync'd though), however the consequence is quite a bit of tearing. Linking that to your point about %mem usage, MSI afterburner reports both cards working at about 80% in that case. Also, you may be able to make sense of this... without 16xaa selected but 60 cap, there is a periodic stutter (as if the 2 cards were out of sync for a moment), this is completely regular and varies according to the frame cap i impose. In this situation afterburner shows much more erratic usage but still not 100%. Not sure what afterburner means by usage, whether its memory % or just activity... i'd be leaning towards saying its got something to do with the size of the frame buffer thats used in CF situations. Especially if your dual 2gb cards are performing so well by comparison.
 
Rock&Roll said:
So, one now has to ask if the 1GB framebuffer isn't the limitation. Or maybe it's a memory controller issue when using above a certain % of the vRAM.

Does your rig stutter even in old games like HL1?
Click to expand...

I dont' think it's frame buffer. I say this because in the recently posted eyefinity6 review 2 5870 E6s exhibit the issue this thread has highlighted. On the Bad Company 2 page Brent admits what we have been saying all along (http://www.hardocp.com/article/2010/04/12/ati_radeon_hd_5870_eyefinity_6_performance_review/4):

"With multi-GPU CrossFireX we needed to aim for higher framerates in order for the game to feel smoother. When the average framerates were down around 30-40FPS, the game felt choppy even though the framerate was at 30FPS. We found that by keeping the framerate higher, the game smoothed out and the choppiness was diminished."

So even when they upped the frames into the 50s it felt smoother but the choppiness was only "diminished". Seems to agree with my experience with CFX & EF. Sub 60fps is noticeable. 60 to 90fps is a lot better but still there.

Crossfire eyefinity reviews really need to highlight this point more because really the reported fps particularly in demanding games is rubbish. You have a situation where 40fps is really 20fps effective.
 
Add me to the list of Crossfire Stutter Sufferer's. Eyefinity causes a quick stutter every 3-6 seconds in most games. It's very irritating and I hope it's fixed soon.
 
I'm not holding my breath on it getting fixed. If it is ever fixed the cards will probably be a lot cheaper by then. Which is why I've ditched crossfire and will pick up another card if it's ever truly fixed. No point having a wasted card sitting in my box if it's causing problems / not being utilised. I don't game in single screen anymore.
 
Thanks for all the comments guys, now i have a clear idea why not to go for another 5870 at the moment or not go for it at all, and wait for new ati cards 6870s possibly next get with eyefinity and more power overall.

I have 1 5870 running eyefinity 5700X1200 with bezel comp, and everything runs smooth
One of my friends bought the same cards and we decided to try a xfire config and see if it was worth to get the second one.

My experience was horrid, more FPS yea but shuttering and also was experience system hang ups at some point with crysis.

my rig core i7 920 D0 overclocked 4ghz and the mobo is capable of 16/16 Pcie lines for Xfire

So at the end this threat has cleared my doubts on getting or not getting another card, im glad i didnt bought 2 at the beginning.

I guess 1 5870s is the safest route to go with eyefinity setups atm.

ill try to overclock it and probably gain a few more FPS :)
 
NickJames said:
Add me to the list of Crossfire Stutter Sufferer's. Eyefinity causes a quick stutter every 3-6 seconds in most games. It's very irritating and I hope it's fixed soon.
Click to expand...

You my friend are on the same mobo as I. Which slots are you using? You need to be using slots #1 and #3 (you'll need a 100mm crossfire bridge). As shown here:
http://i48.photobucket.com/albums/f241/breaker_19/tripple-mon/100_2294.png
http://i48.photobucket.com/albums/f241/breaker_19/tripple-mon/100_2299.png


With this setup, I have no stutter to speak of.

If you were to duplicate my arangement, then the only differences would be 1GB vs 2GB cards, System RAM clocks and timings, the monitors being used. In my case, all three are hooked up over displayPort.

I'm still curious as to why I'm having so few issues while so many other people are.
 
Last edited:
No I don't have my 100mm crossfire bridge yet, still waiting for it in the mail. But I have my doubts its going to make much of a difference. My setup is 2 DVI and 1 DP to DVI active adapter.
 
Rock&Roll said:
In my case, all three are hooked up over displayPort.

I'm still curious as to why I'm having so few issues while so many other people are.
Click to expand...

I wonder if this might be why. 2 dvi and 1 dp for me also. Don't know about others, but i periodcally have a loss of signal on the dp monitor, its a dell so no adapter required. Is there a difference in how the card handles the data for the 2 types of connection?
 
I haven't experienced any issues with my adapter. Never in the last 24 or so hours I have had these monitors running have I had the no signal, or sleep or flashing screen problem. Eyefinity has essentially been flawless until put in Crossfire.
 
Sorry meant to say that, with crossfire i have this issue with the dp monitor, without crossfire i agree everything worked very well...at lower gfx settings though.
 
I don't have any issues with the active DP monitor with crossfire, just the stutter. I can stare at the sky in BFBC2 getting 140 FPS, then it pops to 125 and goes back up almost instantly. These pops happen almost every 4-6 seconds during gameplay. Same happens in Crysis. Not so much in Metro 2033. Doesn't happen at all in NFS Shift (not that I can notice anyway). Happens in Just Cause 2.
 
lowres said:
Sorry meant to say that, with crossfire i have this issue with the dp monitor, without crossfire i agree everything worked very well...at lower gfx settings though.
Click to expand...

Haha - I found I had to have the settings lower than a single GPU to get decent smoothness with CFX EF in some cases (BC2).

Ati - where spending more $ is the dumbest thing you can do.
 
nat559 said:
Any luck with the beta driver? http://hardforum.com/showthread.php?t=1511137

I will test this tonight and report the results.
Click to expand...

I got my new system up and running on Tuesday night. I was running crossfire 5850s (16x/16x) on an X58 with a Core i7 and a single monitor. Everything was silky smooth and I had to end the night.

Wednesday night I got home and installed this new pre-release (BFBC2 fix) driver and got my Eyefinity setup. It started crashing and I initially mistook it for an eyefinity problem, but the crashing continued with a single screen. I rolled back the driver and everything was fine...

Now I am clearly observing the problems everyone else is having with Crossfire + Eyefinity. The game just feels super choppy even when FRAPS reports a FPS of 60 and it makes me not want to play the game. When I disable crossfire, the FPS drops down but the game feels silky smooth. Now I'm mad (1) that I didn't research this enough before buying my 2nd 5850 and (2) that I have to figure out what to do with my 2nd 5850.
 
This is pretty ridiculous, then. Is there no official acknowledgment of the issue? Even HardOCP's description of it in the review is, quite frankly, oblique and almost a whitewash of the situation.
 
I'm hearing vague reports that the 10.4 driver fixes this. Can anyone confirm or deny?
 
nat559 said:
Any luck with the beta driver? http://hardforum.com/showthread.php?t=1511137

I will test this tonight and report the results.
Click to expand...

I tried this for about 2 hrs before i saw that the 'offical 10.4" previews are out (don't know how i missed that they've been out since the 6th....) So switched to those. Immediately notice that the 10.5 betas or whatever they are elimiated my screen flicker on the desktop when the cards are o/c (think this is becuase the 2d clocks are set high already). However the 10.4 pre's the flicker has returned on the dp monitor only (looks like it happens when the gpu switches from/to idle clocks).

Back on 10.4 pre:
Tried the heaven benchmark, which for me is a stuttery mess with 5870 CF+EF. I would say that there is some glimmer of hope that things will improve. At 6004X1080 dx11 0xaa 4xaf the frames aren't great but there is a lot less pause/stutter for me, not as smooth as on one card yet but a step in the right direction. Same settings at dx9, the same as always was, not so much stutter but 2 second good, then a pause, then good again repeated... (the one legged walk is a good analogy). However at dx10 the benchmark ran very smoothly with the same settings, i would say as smooth as a single card but higher fps, shown by the fact that the minimum reported frame rate is 29 compared to dx11 or dx9 - both 15ish,

I checked this result by running FSX in dx9 and 10. Dx9 was crap, same jumpy one legged walk stuff, dx10 was absolutely perfectly stable even in the worst loading situations i could condure up, where CPU is being really screwballed so fps is low but stable, same situations in dx9 were completely unplayable.

I imagine the reason why dx9 has the one legged walk syndrome is that the frames jump from 50+ to 15, whereas the dx11 only has to do 25+ to 15. Why its doing that is a mystery.

Anyway, so if it works for dx10 there must be hope right? I haven't installed much else on this new machine yet (read nother dx9/10/11 games) but i have some reserved hope for these drivers, and even more reserved hope that the hardware is not the issue...

Here is the link to 10.4pre:
http://support.amd.com/us/kbarticles/Pages/ATI-Catalyst-10-4-Hotfix.aspx
 
Last edited:
Matrices said:
This is pretty ridiculous, then. Is there no official acknowledgment of the issue? Even HardOCP's description of it in the review is, quite frankly, oblique and almost a whitewash of the situation.
Click to expand...

I'll take some videos tonight at 60 FPS that should show the problem very clearly. I'll also try again with the new drivers for comparison (which I assume will cause me to crash again).

Oddly, the leaked 4/6/10 driver caused me to crash even when crossfire was disabled and disconnected. It would only work for me if I physically removed the card.
 
You must log in or register to reply here.
Back
Top