• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Crossfire & SLI - PCIEX x8 vs x16

Permabanned
Joined
15 Dec 2008
Posts
3,417
Hey guys, I run some tests on another forum between running in x8 and x16 mode to see what effect it had on modern games, It's nothing special but thought some of you might be interested in the results.

Do note that I only have 2Gb of RAM because one of my sticks died.

Because of this frame rates are lower then normal because of a lack of system RAM.

Anyway, results are mixed

Crysis, x16, No texture Packs :

Crysisx16notexturepack.jpg


Crysis, x8, No texture Packs :

Crysisx8notexturepack.jpg


Those looking close will see that the x8 con fig has a higher minimum frame rate then the x16 con fig, That's because I restarted windows so it had a bit more System RAM free.

Crysis, x16, With Rygels Texture pack :

Crysisx16withrygelstexturepack.jpg


Crysis, x8, With Rygels Texture pack :

Crysisx8withrygelstexturepack.jpg


The minimum frame rate took a hit when running in x8 mode, this is due to the texture pack requiring more then 1Gb of VRAM and having to send quite abit of data over the limit x8 connection. Still not as big of a hit as I was expecting.

STALKER : Clear Sky, x16 :

ClearSkyx16.jpg


STALKER : Clear Sky, x8 :

ClearSkyx8.jpg


I was expecting more of a difference then that. Especially with these 2 titles as they are the most demanding games available at the moment.

Mind you, I do run Clear Sky with a texture pack that I couldn't get to load in the Clear Sky benchmark, Maybe that would have put extra stress on the VRAM and made a difference to the results?

I also wonder if my low resolution has anything to do with it?

So basically if you have enough VRAM to handle the games you play then you should be fine with x8x8 BUT if you play VRAM heavy games then you might see a performance drop.

People with 512mb cards beware.

The GPU is a 4890 and my CPU is a 4.2Ghz Pentium Dual Core E6300

It's not the most accurate or controlled but it gives you an idea.
 
Higher resolutions will certainly show the difference. I remember seeing a test, some time last year. Where the difference was noticeable when you started to hit 1920+ (the kind of resolution people going down the multi GPU path will be using).
 
is your results right for STALKER : Clear Sky?? to me thats low for a 4890 or not?? did u use AA?

with my 2 4850 i get this



i thought a 4890 is just alittle behind 2 4850 ?
 
Last edited:
As was said before, the resolution needs to be cranked up to notice the difference between x8 and x16 (PCI-E 2.0 speeds).

I remember a Tom's Hardware article examining this a while back. In most games at the time, the difference wasn't so big. But there were some exceptions, and FSX took a massive hit from the limited bandwidth.
 
is your results right for STALKER : Clear Sky?? to me thats low for a 4890 or not?? did u use AA?

with my 2 4850 i get this



i thought a 4890 is just alittle behind 2 4850 ?

It's running with full AA and as I said in the OP, I only have 2Gb of System RAM which is limiting everything as the game is using quite a lot of memory and my system is swapping in and out of page file like a bitch...
 
I'm very interested in this, (what with the P55 imminent). I'd like a setup that'd run SLI and not be castrated when the next gen gpu's appear. There's a great little write up on Tweaktown which seems to think it'll be ok, though I think it'll be close to borderline with the high end stuff.
 
It's running with full AA and as I said in the OP, I only have 2Gb of System RAM which is limiting everything as the game is using quite a lot of memory and my system is swapping in and out of page file like a bitch...
tbh his clear sky result is about right

i ran the benchmark again with full AA and this is what i got.

1920x1200


1280x1024
 
Crysis Warhead Benchmark, x16 / x16

....

Crysis Warhead Benchmark, x16 / x8

......

FurMark x16 / x16

......

FurMark x16 / x8

......
 
Last edited:
Thanks for that. So not really much of a difference then...strange. Thats convinced me I have absolutely no need for a 790Fx board and 790GX will do me jus fine :D

Thanks again
 
Thanks for that. So not really much of a difference then...strange. Thats convinced me I have absolutely no need for a 790Fx board and 790GX will do me jus fine :D

Thanks again
your welcome

i guess some other games may have more of a hit but i wouldn't worry about it so much.


if i have time later i will post some more results.
 
Off topic but what cpu cooler are you using? 3.7 is a nice overclock...or is water cooled?
Thermalright Ultra-120A Extreme + fan.

summer weather it idle around 40c , under load 52c.

winter weather it idle around 33c , under load 45c.
 
Last edited:
Back
Top Bottom