GTX 970 G1 Gaming SLI Performance issues.

[z1tu]

For me, going from FX-6350 and PCIE 2.0 to i5 4690K and PCIE 3.0 meant about 1K difference in overall score in 3dmark. With the same video card of course.

Yeah overall score will be improved ofcourse. The interesting thing would be seeing if the GPU score itself increased.

It did by about the same amount, I think it was around 2k, I'll have to double check once I get home, but the difference wasn't as staggering as people here are expecting from the new cards.
My friend, who actually owns a 970, with some decent oc, can get 2k more points than me (gtx 780) in graphics score.

 

[DLGenesis]

gotta say. my 780sli getting better score then some of these 970 scores

 

[z1tu]

gotta say. my 780sli getting better score then some of these 970 scores

It still doesn't mean anything in games though, just means in synthetic tests they're not as good.

 

[bloewens89]

I am having a similar problem with 3D Futuremark except mine seem to be worse. I'm running 2x Gigabyte 970s in SLI, an i5-4690k, Asus Maximus Formula VII, and 16gb Corsair Vengeance 2133mhz. I'm only getting a Firestrike score of 13.5k! However, Crysis 3 seems to run nicely maxed out at 1440P...........

 

[eidairaman1]

I am having a similar problem with 3D Futuremark except mine seem to be worse. I'm running 2x Gigabyte 970s in SLI, an i5-4690k, Asus Maximus Formula VII, and 16gb Corsair Vengeance 2133mhz. I'm only getting a Firestrike score of 13.5k! However, Crysis 3 seems to run nicely maxed out at 1440P...........

Forget benchmarks, game performance counts

 

[RealNeil]

I think his scores are normal.

 

[bloewens89]

Actually, I just went into the Nvidia Control Panel and changed a setting to Highest Refresh Rate Possible. Apparently Futuremark caps rates to 60 in some tests... it pushed up my graphics score to over 19,500 which is much better but still considerably lower than the 24k scores people are getting. 6k change! WOW

http://www.3dmark.com/3dm/4292020

 

[raton]

i have the same problem here with a 970 sli.

Some times i have a normal score of 15500pts, go watch a movie par ex, run firestrike again 3/4 times, a got 13000 scores.

Activating/deactivating SLI does fix the problem for a limited time...

edit : https://forums.geforce.com/default/...y-lower-voltage-than-the-other-driver-bug-/1/

The problem is that one of the gpus is anormaly undervolted in SLI.

0.75 to 0,180v difference between gpu1 and 2 in sli, this cause performance and stability issues.

Both cards are fine when tested individualy.

 

[RealNeil]

The problem is that one of the gpus is anormaly undervolted in SLI.
https://forums.geforce.com/default/...y-lower-voltage-than-the-other-driver-bug-/1/

What PSU do you use?

 

[raton]

the problem is not the psu, it's the drivers.

750w gold.

 

[vega22]

driver issues on a brand new card???


well i never...

fyi i lose about a grand going from oc 4770k to oc 2600k with 1 gpu in those 3dmark tests.

1 gpu which almost matches 2 970s :rofl:

issue is drivers imo btw. immature drivers and maybe a bloated os install. the fact they do so well when installed alone should indicate it is not a hardware issue.

 

[RealNeil]

the problem is not the psu, it's the drivers.

750w gold.

Post your system specs in your control panel please.

I went to your link and read through it, and it seems like others have the same issues, but how is everyone sure that it's not crappy PSUs or voltage handling properties on some motherboards?
I agree that drivers may be the culprit, but that may not be the case either. I wish that I had a few of the 970s or 980s to test with.

 

[raton]

was running 780s just fine before.... higher tdp

it's clearly drivers related to the 9xx series.

 

[RealNeil]

was running 780s just fine before.... higher tdp

it's clearly drivers related to the 9xx series.

780s in SLI with a 750W PSU?
I thought that I was pushing it running two R9-280X OCs in crossfire with a 750W unit.

Until recently, I had two GTX-680s in SLI, but I was using a 1000W PSU with them. I just sold them to go toward a pair of 970s.

 

[raton]

people tends to overestimate power consumption

 

[RealNeil]

When there is a solution to your issue, will you post here on it? I'm interested.

 

[Aquinus]

people tends to overestimate power consumption

I take it with a smug comment like that, you're going to show us how much that tower of your consumes, right?
In all seriousness when both of my 6870s and my CPU are fully overclocked under heavy load 550-600 watts isn't an unrealistic draw off the wall on my machine which is strictly the tower by itself. It idles at 180-200 watts and full CPU load alone will bring it up to 300-350-watts depending on the workload.

All in all, I don't think it's unrealistic to think that the 970s can consume just as much power as stock 6870s, considering the 970 has a 145-watt TDP and the 6870s have 155-watt TDP. So without doing any testing I would suspect that putting two 970s would tax my PSU similarly to what I have now, which is still a good amount of power. The only thing going for the 6870s is the two PCI-E 6-pin connectors, where the second was only required because the GPU broke 150-watts for its TDP.

My point is that you might be confusing overestimates with taking into account efficiency overhead. I never fully load my PSUs and I try to give myself anywhere between 25-50% overhead under a realistic full load and it so happens that most PSUs provide the most efficient power output between 30% and 70% of rated maximum with 50% typically being a sweet spot. Just saying.

 

[SmileyyFML]

This thread is kinda old and dead but I think I figured out op's problem with the i7 2600k.

http://ark.intel.com/products/52214

This list the 2600k as a max number of 16 pci-e lanes, as well as only supporting up to pcie 2.0. Now I could be totally wrong about this, but doesn't that mean with both gpus plugged in they at absolute most be running at Pcie 2.0 x8/x8?

If Im wrong screw me because that's my best guess.

 

[mull54]

This thread is kinda old and dead but I think I figured out op's problem with the i7 2600k.

http://ark.intel.com/products/52214

This list the 2600k as a max number of 16 pci-e lanes, as well as only supporting up to pcie 2.0. Now I could be totally wrong about this, but doesn't that mean with both gpus plugged in they at absolute most be running at Pcie 2.0 x8/x8?

If Im wrong screw me because that's my best guess.

Yes, i have now upgraded to a i7 5820k and everything is much better now. I don't know if it was the PCI-E lanes alone or if the CPU was bottlenecking also.