GTX 970 G1 Gaming SLI Performance issues.

[mull54]

Those graphs are single card even if multigpu, there is the PLX chip on board of those. In multi card it might matter more with all the bandwidth shared. 2600K is pretty old and certainly not comparable with Haswell even if it is only 20% raw difference. With PCIE 3.0 being out for like 2 years 2.0 might not work perfectly, there were cards that wouldn't run perfectly in 1.1, those were 2.0 cards. But mostly the CPU is the problem, those people might be using Ivy/Haswell/E.

It would be logical to see a small performance loss not a loss of 5-6k.

 

[mull54]

BINGOOOOOOOOOOOO
Is that a PCI-E Sound Blaster?

If you have an additional PCI-E card, it will take x4 lanes from your x8/x8 GPU's, making one of them run at x4. Try running the benchmark without the PCI card installed. I could be totally wrong, but these things can happen.

Yes it is. It is a pcie x1. But i will try ut

 

[RCoon]

Yes it is. It is a pcie x1. But i will try ut

I've ALWAYS come up with that as a theoretical problem, but nobody has ever been in the situation where it becomes an issue. If it is the case, and it works, I can finally justify telling people to watch out for extra PCI devices with their multi GPU setups without sounding like a nagging wench.

If not, well, I'm a wench I suppose.

 

[springs113]

BINGOOOOOOOOOOOO
Is that a PCI-E Sound Blaster?

If you have an additional PCI-E card, it will take x4 lanes from your x8/x8 GPU's, making one of them run at x4. Try running the benchmark without the PCI card installed. I could be totally wrong, but these things can happen.

So true, had a similar issue running a 5820k w/ dual GPUs and a ZXR...my scores weren't right in some of these synthetics. Then upgraded to the 5930k and scores skyrocketed, doubled in some cases.

Edit:
I think I saved some of the scores to prove it as well, would have to go digging for that ssd though.

 

[MxPhenom 216]

The weird thing is that i get 12k GPU score with each card without SLI. So that should still me a combined GPU score of 22k+-2000. I even overclocked the core on both cards by 125MHz(stock is 1178). And the score was still 179XX.
Parhaps it is my old SLI bridge that i have had laying unprotected in a drawer for 3 years?

Your not understanding what @cadaveca is trying to say.

 

[mull54]

Your not understanding what @cadaveca is trying to say.

Ofcoruse i do but that big of a difference is not becuase of that. People with the exact same specs get much higher GPU score.

I've ALWAYS come up with that as a theoretical problem, but nobody has ever been in the situation where it becomes an issue. If it is the case, and it works, I can finally justify telling people to watch out for extra PCI devices with their multi GPU setups without sounding like a nagging wench.

If not, well, I'm a wench I suppose.

It didn't work But should i uninstall the drivers and turn on realtek audio in bios before i draw any conclusions. Or shouldn't that matter at all?

So true, had a similar issue running a 5820k w/ dual GPUs and a ZXR...my scores weren't right in some of these synthetics. Then upgraded to the 5930k and scores skyrocketed, doubled in some cases.

Edit:
I think I saved some of the scores to prove it as well, would have to go digging for that ssd though.

I was hoping that was the issue.

What about the molex sli connector?

 

[zsolt_93]

http://www.3dmark.com/search#/?mode...GTX 970&cpuName=Intel Core i7-2600K Processor
Here are the scores, the only higher than 17k with your config is the one with 3x SLI, the same with 3770k, 4770k gets 18k with 2 card SLI. People with 5960K get around 20k. You are just being unrealistic or reading the scores wrong.

 

[mull54]

http://www.3dmark.com/search#/?mode...GTX 970&cpuName=Intel Core i7-2600K Processor
Here are the scores, the only higher than 17k with your config is the one with 3x SLI, the same with 3770k, 4770k gets 18k with 2 card SLI. People with 5960K get around 20k. You are just being unrealistic or reading the scores wrong.

Well, somethings is up with the gigabyte cards in that case, maybe bios or something like that. Because "stock" aftermarket cards(970) get 22k plus.

 

[zsolt_93]

Show me those scores. So far i can only say that you are trolling and not believing any evidence we bring to you.

 

[mull54]

Show me those scores. So far i can only say that you are trolling and not believing any evidence we bring to you.

Are you got damn kidding me. Just look at the gtx 970 owners thread at overclockers.net. Even my old r290 Tri x get 19k in gpu score on firestrike. Look att my profile at 3dmark and there you Will find my scores of tri x for example.

 

[yogurt_21]

It would be helpful if you posted up the actual result or better yet did it in the style of the submissions in the 3dmark threads on this forum. I for one am inclined to believe you might be reading the overall score as the gpu score.

http://www.3dmark.com/fs/1504997
R9 290 single, 19k gpu, 14k overall

http://www.3dmark.com/fs/2830341
970 sli 25K gpu, just under 17k overall

 

[zsolt_93]

Are you that William guy on 3dmark?

 

[OneMoar]

stock cooler with stock clocks your score is right where it should be

 

[mull54]

It would be helpful if you posted up the actual result or better yet did it in the style of the submissions in the 3dmark threads on this forum. I for one am inclined to believe you might be reading the overall score as the gpu score.

http://www.3dmark.com/fs/1504997
R9 290 single, 19k gpu, 14k overall

http://www.3dmark.com/fs/2830341
970 sli 25K gpu, just under 17k overall

Here you go mate. This is my score. If you check my other test you Will find the test with my recent 290 tri x cf.

13k overall score and 17k gpu score. Im not mixing it up.

Are you that William guy on 3dmark?

That correct.

stock cooler with stock clocks your score is right where it should be

Not stock cooler anymore. Zalman cnps and 4.2 ghz Clock.

 

[yogurt_21]

so is this you? http://www.3dmark.com/fs/2827077

an interesting graph there is that out of 37 similar systems, yours is right in the middle of the pack along with the bulk of 24 others who range from 12600-14699pts.

In fact there are only 4 scores above 15K and if you look at the lowest over 15k you'll see rather low clocks on the cards 1051/1753. If you look at the highest of the sub 15k scores, you'll see them clocked at 1259/1928. It looks as if 3 of the top 4 scores are invalid as they are playing around with textures. Ie they are not running at the same detail level you are. The guy at 15825 looks like he just insanely clocked his cards.

 

[mull54]

so is this you? http://www.3dmark.com/fs/2827077

an interesting graph there is that out of 37 similar systems, yours is right in the middle of the pack along with the bulk of 24 others who range from 12600-14699pts.

In fact there are only 4 scores above 15K and if you look at the lowest over 15k you'll see rather low clocks on the cards 1051/1753. If you look at the highest of the sub 15k scores, you'll see them clocked at 1259/1928. It looks as if 3 of the top 4 scores are invalid as they are playing around with textures. Ie they are not running at the same detail level you are. The guy at 15825 looks like he just insanely clocked his cards.

Yeah if you look at the total score its normal but its the gpu score i am Walking about. Parhaps there is something with the gigabyte cards in sli and firestrike atm.

Btw i have plugget in the 4 pin molex from the mobo to psu ev en tho i have a 6 and a 8 pin connector to both cards. Is this right or should i skip that?

 

[buildzoid]

BINGOOOOOOOOOOOO
Is that a PCI-E Sound Blaster?

If you have an additional PCI-E card, it will take x4 lanes from your x8/x8 GPU's, making one of them run at x4. Try running the benchmark without the PCI card installed. I could be totally wrong, but these things can happen.

Not the case Nvidia SLI doesn't work on 4x slots. The control panel wouldn't let him enable SLI.

 

[Aquinus]

You're complaining about an issue where there is none. The score you're getting is just like everyone else's and as others have said, your CPU can, will, and does impact your GPU score. If you adjust your CPU's clocks, you'll see the score change. Don't take my word for it, do it yourself and find out.

 

[rtwjunkie]

I've ALWAYS come up with that as a theoretical problem, but nobody has ever been in the situation where it becomes an issue. If it is the case, and it works, I can finally justify telling people to watch out for extra PCI devices with their multi GPU setups without sounding like a nagging wench.

If not, well, I'm a wench I suppose.

Soooo, basically then, I can infer that having other devices in PCI/PCIe slots doesn't hurt performance on a PCIe 3.0 board IF you use a single GPU setup, because there's plenty to share? It's only on SLI/CF setups?

 

[Aquinus]

Soooo, basically then, I can infer that having other devices in PCI/PCIe slots doesn't hurt performance on a PCIe 3.0 board IF you use a single GPU setup, because there's plenty to share? It's only on SLI/CF setups?

It depends if they're switched lanes or not. If you have PCI-E 1x slots and PCI slots, they won't impact any other lanes as these tend to have dedicated PCI-E lanes, usually off the PCH/IOH/Chipset. When you have slots that are like 16/DC/DC or 8/8/DC or 8/4/4, then the slot being populated will impact the number of lanes available to the device that is sharing its lanes.

Lanes aren't "shared" unless your board has a PLX chip in it for switching the PCI-E packets. Lanes are switched when ports become populated which can take lanes away from another device depending on how the motherboard is setup.

 

[mull54]

You're complaining about an issue where there is none. The score you're getting is just like everyone else's and as others have said, your CPU can, will, and does impact your GPU score. If you adjust your CPU's clocks, you'll see the score change. Don't take my word for it, do it yourself and find out.

But come on. Look at the GPU and physics scores. The cpu score are almost the same as mine and he gets 24k on GPU.

http://www.3dmark.com/fs/2832646
http://www.3dmark.com/fs/2807735
http://www.3dmark.com/fs/2806068
http://www.3dmark.com/fs/2817211
http://www.3dmark.com/fs/2808450
http://www.3dmark.com/fs/2816970
http://www.3dmark.com/fs/2784943

And i get 17k and 17900 when i clocked the core to 1325 mhz. You can clearly se that something is wrong.

 

[rtwjunkie]

It depends if they're switched lanes or not. If you have PCI-E 1x slots and PCI slots, they won't impact any other lanes as these tend to have dedicated PCI-E lanes, usually off the PCH/IOH/Chipset. When you have slots that are like 16/DC/DC or 8/8/DC or 8/4/4, then the slot being populated will impact the number of lanes available to the device that is sharing its lanes.

Lanes aren't "shared" unless your board has a PLX chip in it for switching the PCI-E packets. Lanes are switched when ports become populated which can take lanes away from another device depending on how the motherboard is setup.

Clear as mud...I think, LOL. Thanks for the explanation!

 

[mull54]

It depends if they're switched lanes or not. If you have PCI-E 1x slots and PCI slots, they won't impact any other lanes as these tend to have dedicated PCI-E lanes, usually off the PCH/IOH/Chipset. When you have slots that are like 16/DC/DC or 8/8/DC or 8/4/4, then the slot being populated will impact the number of lanes available to the device that is sharing its lanes.

Lanes aren't "shared" unless your board has a PLX chip in it for switching the PCI-E packets. Lanes are switched when ports become populated which can take lanes away from another device depending on how the motherboard is setup.

wait a min. My mobo got a PLX chip:

1. "Integrated PLX PEX8608 chip onboard to offer sufficient PCI-E lanes"

http://www.asrock.com/mb/Intel/P67 Extreme4/

Is that the problem?

 

[Frag_Maniac]

It is the PCIE2.0 vs PCIE 3.0 thing too. I have read something with SLI/CFX on 2.0 motherboards with this generation is not optimal performance. Although bandwidth should be more than enough.

The bandwith itself is def not an issue.

http://www.pugetsystems.com/labs/articles/Impact-of-PCI-E-Speed-on-Gaming-Performance-518/

 

[yogurt_21]

But come on. Look at the GPU and physics scores. The cpu score are almost the same as mine and he gets 24k on GPU.

http://www.3dmark.com/fs/2832646
http://www.3dmark.com/fs/2807735
http://www.3dmark.com/fs/2806068
http://www.3dmark.com/fs/2817211
http://www.3dmark.com/fs/2808450
http://www.3dmark.com/fs/2816970
http://www.3dmark.com/fs/2784943

And i get 17k and 17900 when i clocked the core to 1325 mhz. You can clearly se that something is wrong.

almost the same? the ones with the 3770k show 2000 more pts on the physx test and 800-1000 pts more on the combined.

That's a variance of 15-20%

the 4770k physx and combined scores are even higher, more like the 23-28% variance range.

On top of that some of those are invalid as I said before because they are playing around with textures in the NV control panel. This makes it so that their scores are higher simply because the detail level is lower than the benchmark design.



Then you have the old disclaimer "results are not guaranteed even if you have the exact same hardware"

3dmark these day does a piss poor job at actually evaluating gaming performance. That's why W1zzard stopped using it in his reviews. What about actual gaming performance? Any stuttering, slowness? Are you able to set details to the highest level at 1080p? or if you have a bigger monitor does performance meet expextations at 1440p/1600p, 4k etc?