• 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.

GTX970 SLI with overclocked hex-core... GPU-benchmarks lower than expected

It's not reading boost clock properly as String mentioned... at least MSI Afterburner is reporting it as 1420-1450... it has been fluctuating a lot in these last runs.

Fan on 100% has somehow been scoring lower than fan on auto? I don't know how to explain that... but it's what I'm seeing.

And yeah... the score is too low for my liking... I'm unsure where the bottleneck is.

I guess it might be best to test each card individually.

Either it is one GPU or one PCI-E slot error/capacity as the CPU/mem benchmarks are performing as expected... so the rest of the motherboard/cpu/ram must be ok.

:(
 
It's not reading boost clock properly as String mentioned... at least MSI Afterburner is reporting it as 1420-1450... it has been fluctuating a lot in these last runs.

Fan on 100% has somehow been scoring lower than fan on auto? I don't know how to explain that... but it's what I'm seeing.

And yeah... the score is too low for my liking... I'm unsure where the bottleneck is.

I guess it might be best to test each card individually.

Either it is one GPU or one PCI-E slot error/capacity as the CPU/mem benchmarks are performing as expected... so the rest of the motherboard/cpu/ram must be ok.

:(

Are you seeing the core clock drop in 13hz increments during the benchmark? This will tell us if the card is being throttled by the driver, or not.
 
Your GPU, Physics and combined score aren't from running FireStrike Extreme but regular FireStrike :)
Good eye, my bad i must have copy and pasted the wrong one when i had both pages open, these are the real results -
SCORE
9666 with NVIDIA GeForce GTX 970(2x) and Intel Core i5-4670K
Graphics Score 11768
Physics Score 9180
Combined Score 4277

Nice one.
 
Its not reading his boost clock correctly. Mine states 1478 but was really at 1556x

It is reading his OC'ed base clock correctly. What his card boost to is a "lottery".
If he OC'ed it to 1300MHz he base clock he'd get a better score. When I OC'ed my 2x MSI GTX 970 further from 1250MHz to 1300MHz base clock for both cards my FPS increased by 7FPS more in Unigine Heaven 4.0 average score.

It's not reading boost clock properly as String mentioned... at least MSI Afterburner is reporting it as 1420-1450... it has been fluctuating a lot in these last runs.

Fan on 100% has somehow been scoring lower than fan on auto? I don't know how to explain that... but it's what I'm seeing.

And yeah... the score is too low for my liking... I'm unsure where the bottleneck is.

I guess it might be best to test each card individually.

Either it is one GPU or one PCI-E slot error/capacity as the CPU/mem benchmarks are performing as expected... so the rest of the motherboard/cpu/ram must be ok.

:(

3Dmark is reading your OC'ed base clock - just like GPU-Z would. Your max boost clock varies and isn't a fixed speed. MSI Afterburner showing your clocks at 1420-1450Mhz is perfectly normal ! My 2x MSI GTX 970 Gaming cards I run at 1300MHz base clock boosts to a stable 1502Mhz. Increase your base clock by 50Mhz and you see better results.

Also do remember how GPU Boost 2.0 works - it's all based around voltage, temperature. When you run your fans at 100% speed you are also maxing out the voltage fan speed this means that the card sacrifices some GPU speed because the fans run at 100% and then there isn't enough voltage for the GPU to boost higher.
So set your fanspeed to auto. Set your powertarget to the max 106/110% (or more depending on what card you have).
Increase your base clock on the GPU to 1300MHz. Then run some benchmarks again and have an eye on your max boost clock speed. Check stability and then report results.
 
Last edited:
What software are you using to OC base clock and boost clock separately?

I use MSI Afterburner and that one overclocks your base clock - the boost clock automatically follows. You can check with GPU-Z. When you OC the base clock and press apply then you can see the change in GPU-Z and it will also show you the guaranteed (minimum) boost clock.
 
I'm using afterburner too...

I assume you don't quite get the point here... similarly clocked cards are scoring 10-15% higher than mine... that is the problem...
 
I'm using afterburner too...

I assume you don't quite get the point here... similarly clocked cards are scoring 10-15% higher than mine... that is the problem...

Hmm what similar clocked cards ? I haven't seen results from those.:confused:

Again what about background programs have you closed them ?. Remember a "used" system that has been running for a long time and with games and programs installed won't score just as high as a fresh windows install unless you really trim windows often.

When I ran my cards at 1250MHz GPU core I pretty much got the same results as you. Around 107 FPS average in Unigine Heaven 4.0

My FireStrike Extreme score
SCORE 9261 with NVIDIA GeForce GTX 970(2x) and Intel Core i7-4930K
Graphics Score 10718
Physics Score 14687
Combined Score 3599
http://www.3dmark.com/fs/3513767

Check your GPU score with mine - they are very similar. My GPU speed here was 1250Mhz/1800Mhz ram.

What brand of GTX 970's are you running ?
 
Last edited:
Scores do seem a bit low. I find its good to reboot before testing but let the PC sit for a few minutes running the test to make sure everything has finished loading. Also you could bring up task manager with the PC idle and make sure it's not doing anything.

My scores for comparison on Xeon 5650 @ 4.7 with 970 SLI @ 1500/8000

Fire Strike 19195
Graphics 25383
Physics 16277
Combined 7500
 
It was already a relatively clean install of Win 7... decided to try a clean install of Win 8.1 to see if I would get much of a change.

Oddly... now I can't OC as high as I was before.

The highest I can bench stable now is +100 Core and +500 Mem. +105 on the core and I get grey screen + it resets to base boost clock... then continues.

GPU2 sits around 100-105% power usage (slider increased to 120%)
GPU1 sits around 75% power usage (slider increased to 112%)

both power sliders to max... I assume GPU1 has a higher default TDP than GPU2

GPU1 is a Gigabyte Gaming G1
GPU2 is an Asus Strix

It appears to be GPU2 holding back GPU1 as GPU1 on it's own was able to boost to 1550-1600 without trouble.

I would have gone for both of them Gaming G1s... only for the second slot the G1s are a touch too long for the Be Quiet Silent Base 800 case if wanting to leave the bottom hard drive cage in place (which I do).

For comparisons I was looking at the Firestrike benchmarks threads on this forum and trying to compare.

I see what you mean about your score.
 
It was already a relatively clean install of Win 7... decided to try a clean install of Win 8.1 to see if I would get much of a change.

Oddly... now I can't OC as high as I was before.

The highest I can bench stable now is +100 Core and +500 Mem. +105 on the core and I get grey screen + it resets to base boost clock... then continues.

GPU2 sits around 100-105% power usage (slider increased to 120%)
GPU1 sits around 75% power usage (slider increased to 112%)

both power sliders to max... I assume GPU1 has a higher default TDP than GPU2

GPU1 is a Gigabyte Gaming G1
GPU2 is an Asus Strix

It appears to be GPU2 holding back GPU1 as GPU1 on it's own was able to boost to 1550-1600 without trouble.

I would have gone for both of them Gaming G1s... only for the second slot the G1s are a touch too long for the Be Quiet Silent Base 800 case if wanting to leave the bottom hard drive cage in place (which I do).

For comparisons I was looking at the Firestrike benchmarks threads on this forum and trying to compare.

I see what you mean about your score.

Non matched Cards are probably your issue.
 
OK... darn :(

Still have a couple of days left to return it and get a second G1.

But that would require moving my 3TB and 750GB spinning platter drives into server in basement... with limited space.

Or I give up on this relatively small %age gain (which I suppose I am not guaranteed with a second G1 and keep setup as originally intended...).
 
Back
Top Bottom