Poor performance with Crossfired R9 270x's.

PivMan · May 9, 2014

Hi there everyone. Posting here is my last resort because I'm about to pull my hair out. My internet history is literally only Google searches at this point.

A few days ago I got a few months old Gigabyte R9 270x 2GB via Ebay to accompany my current retail bought one in a Crossfire setup. I had the specs to run the setup, PSU size, all that, so I figured I'd get the performance I've been needing without having to wait several months to scrounge the money for a really high end card.

I'm very well aware and have been aware of the shortcomings of a Crossfire solution. I know all of the possible issues and such, but AMD has been devoting themselves to fixing them with decent success for a while now so I figured what the hay. I'll go for it.

I plugged in the card (several times up until this point, even switching the master and slave) and connected the bridge. PC started up, I started up CCC, enabled Crossfire no problem. I started up Radeon Pro (no custom profiles or anything, just the OSD to monitor the cards), it detected both cards fine and said Crossfire was good to go. So I started up a few games I play often and went to see how they run. Results were horrendous. When things weren't stuttering to hell while the game returned a playable (but still too low for what the setup should be able to do) framerate, the FPS would fluctuate ridiculously or I'd get approximately the same performance as if Xfire was off. There were points where no matter what I changed the in game settings to the framerate would stay in a range (usually 30 to 57 FPS, never breaking sixty) fluctuating crazily. Neither card ever reaches more than 40% usage, which is awful, and also fluctuates crazily.

Tested games:
Battlefield 4 (experienced all of the issues above)
TES V: Skyrim (Crossfire made nearly no difference)
Fallout 3 (hated Crossfire, ran with heavy stuttering)
Fallout NV (same effect as Skyrim)
ArmA 3 (seemed to make some kind of small improvement, likely placebo)
ArmA 2 (somewhat lower performance)
Planetside 2 (didn't even start, got a BSOD)

My specs:
Asus M5A97 LE R2.0 (aware of the 4x PCI-E slot, I doubt it would make gaming this bad)
AMD FX-6300 @ 3.9 Ghz (will OC further once new heatsink shows up)
Corsair Vengeance 2x4GB @ 1866 Mhz
EVGA SuperNOVA 750W Gold
Gigabyte R9 270X GB in Crossfire (of course)
WD Blue 1TB 7200 RPM

damric · May 9, 2014

Have you tried disabling ULPS? That always made a drastic improvement for me in CF.

PivMan · May 9, 2014

Yes, it is disabled. I did so through the registry. Saw no real difference.

wss_003 · May 9, 2014

I had similar issues and found out that my issues was my overclock i had used software to oc rather than bios and when i switched my settings in bios the results were amazing

PivMan · May 9, 2014

I'll look into this as I have done both.

damric · May 9, 2014

turn down your memory overclock. then raise it slowly while you monitor FPS. It often won't crash when pushed too far, but will cause errors that lower performance.

Yoplait95 · May 9, 2014

Its his motherboard the PCIe slots on the LE are x4 when two are in use. :3 I know I tried crossfiring two 7850 on a M5A97 LE R2.0, you should be experiencing %20 -%30 performance loss. As for a single 270x running outside of crossfire. Reinstall drivers and run a gpu stress test. Its possible the card was overclocked and mined on for Scypt coins / Bitcoin

damric · May 9, 2014

Yoplait95 :

Got any benchmarks to prove that theory?

PivMan · May 9, 2014

GPU-Z states that the 16x lane scales to 8x when in Crossfire, the other one remains 4x as usual. Not saying you're wrong, but GPU-Z says otherwise. Appreciate the help, all of you.

Even if it were the case I'd welcome a penalty like that over the hell that's going on right now.

The card was in fact mined with prior to me getting it but the previous owner stated it was done stock (?) and only for two or so months, which sounds correct because the things was practically dust free when I got it.

I'm about to reset my CMOS to see if WWS's idea has any truth to it. I literally just wiped everything driver related from my system and stuck 14.4 back in. Same results.

Yoplait95 · May 9, 2014

Which one ? Crossfire scaling on PCIe or the fact that the cards life span has been used up for mining ?

PivMan · May 9, 2014

Does mining degrade just the lifespan or does it impact the card's performance in general?

I honestly know too little about the bitcoin process to know how it affects GPUs. Should've looked into that.

damric · May 9, 2014

x4, x8, x16 scaling on 270s.

Yoplait95 · May 9, 2014

I'll cut it short.. Mining on GPU's is only really profitable if you overclock them (really high).. Then when AMD releases a new card that is 'better' at opencl computation you will sell their card as is damaged to make a profit / space for newer cards. This is all based on Scrypt mining as SHA-256 mining with GPU is just too underwhelming in profit. You'd overvolt the card and overclock it past what normaly could be done outside mining, due to it not needing to output a display it faults less at high overclocks etc.

Yoplait95 · May 9, 2014

damric :

I just realized the new R series doesnt require a bridge so they might differ :|

Personally I think it would be worse as its all based on the lanes and not a bridge..

damric · May 9, 2014

No one cares about mining since like 2 years ago...

anbello262 · May 9, 2014

About mining and the second card being damaged, you can try a quick thing:
Unplug your old card and only use this new card with your games, try to stress test it, anything you can think of. If it works correctly, then the mining past probably isn't a problem. If it has issues, then you know where the problem is.

damric :

You mean 2 /months/?

Yoplait95 · May 9, 2014

? Ok.. I do, for a living.. but not mixing personal affairs in, Xfire in x4 with a gpu that powerful is definitely going to experience hard bottlenck from the board..

PivMan · May 10, 2014

Define bridge. You mean a Crossfire bridge? The lower grade cards might not but the 270x does for sure.

E: Oh my God you guys reply so fast. I'll probably check the new card on its own, but I haven't seen much issue with it in other regards.

I also really doubt a single PCI-E lane running at 4x will drop BF4 to 30 FPS all the time.

damric · May 10, 2014

Yoplait95 :

Yeah show me some hard evidence of that please.

Yoplait95 · May 10, 2014

Here: http://www.techpowerup.com/reviews/AMD/HD_5870_PCI-Express_Scaling/1.html

Yoplait95 · May 10, 2014

Again not sure how crossfire will work when scaling without a bridge..

damric · May 10, 2014

W1zzard's review says:

PCI-Express 2.0 x4 is where the Radeon HD 5870's discomfort is slightly notable, with a 5% drop

Maybe you should have read the review.

Yoplait95 · May 10, 2014

damric :

With a less powerful 5870. 3(4 including R series) generations behind current GPUs :X The R9 280x is sooo much more powerful and requires more bandwidth to operate.

Yoplait95 · May 10, 2014

So OP? Whats going on ? Any change from single gpu configurations ?

PivMan · May 10, 2014

http://www.techpowerup.com/reviews/Intel/Ivy_Bridge_PCI-Express_Scaling/19.html

I don't know what to tell you Yoplait.

I tested the single card I got from Ebay, and the mining damage myth seems to have no effect. It runs just as good as the other card. Might have a shorter lifetime, but I'll have a new rig by then.

I'm about to try switching master and slave again and see what happens after these driver updates. Wish me luck I guess.

Poor performance with Crossfired R9 270x's.

Reputable

Illustrious

Reputable

Honorable

Reputable

Illustrious

Honorable

Illustrious

Reputable

Honorable

Reputable

Illustrious

Honorable

Honorable

Illustrious

Distinguished

Honorable

Reputable

Illustrious

Honorable

Honorable

Illustrious

Honorable

Honorable

Reputable

Share this page