1070ftw 2100mhz - did I win the silicone lottery?

[Bender Blues]

After watching Jays video today on the new AMD Radeon 5700XT, he reached a max OC of 2144mhz before crashing, it got me thinking, did I get lucky with this 1070ftw card at 2100mhz stable, cool, no artifacts for a week now?

The card runs at 2100, the 2088mhz was just from a reset and re-test. Running at a max temp of 56c. NO VOLTAGE ADJUST. And also the GPU doesn't run at 99% GPU usage, that was a stress test. It usually runs in the 50%-60% range.

So did I get lucky with this card?

Oh and one other thing - the red arrow.  Is there some reason why its max usage was 101% when I have it set to run at 120%? 

In the Nvidia control, I have gpu power set to "prefer maximum performance" - globally.

 

Thank you!

 

[Mark Kaine]

"here's a screen shot of my record breaking gpu - not actually showing the records it's breaking" 

 

Ok?  

 

PS: you can't compare two completely different kind of cards like that at all. 

 

 

[Praesi]

If your GPU runs at 50-60%, somethings isnt working like it should. You want the GPU be at 99%

[Bender Blues]

Just now, Praesi said:

If your GPU runs at 50-60%, somethings isnt working like it should. You want the GPU be at 99%

It depends on the program or game doesn't it? Maybe the 50-60 was from bad self memory. Gonna try a heavy game load again and see.

 

[Mark Kaine]

6 minutes ago, Praesi said:

If your GPU runs at 50-60%, somethings isnt working like it should. You want the GPU be at 99%

Depends,  if I play a 2003 game on my gtx 1060 it's also just using ~40% or less and I think that's totally normal. 

 

In any case OP should run firestrike and post the results - because they seem to think it's stable but didn't actually test it. 

 

 

[Jurrunio]

1. Power % is not usage, it's power the card pulls from the PCIe power connector(s) and PCIe slot. 100% power means 185W.

 

2. I'd say it's above average but not god-like for a card this overbuilt. I am interested in knowing what will happen if you max out the power slider and voltage slider though

[Fasauceome]

2100 is quite good but not amazing. Most 1070 models will peak around 2000 just from GPU boost, Pascal cards that can push 2200 are the real winners.

[Bender Blues]

6 minutes ago, Jurrunio said:

1. Power % is not usage, it's power the card pulls from the PCIe power connector(s) and PCIe slot. 100% power means 185W.

 

2. I'd say it's above average but not god-like for a card this overbuilt. I am interested in knowing what will happen if you max out the power slider and voltage slider though

Oh and please let me clarify, by NONE, zero stretch of ones imagination do I think this is some monumental thing or braggin thread. Its far far far from it. I know theres thousands of more users that can bang out far more here.  

 

Right, I know the difference, just a horrid choice of words. Power is how much the card is pulling, usage is how much of the card is being utilized.

 

And to your message, I did and have had the power slider always at 120%. But theres no way (I think) I want to start touching minuscule voltage changes heh

ty!  

[Bender Blues]

27 minutes ago, Mark Kaine said:

Depends,  if I play a 2003 game on my gtx 1060 it's also just using ~40% or less and I think that's totally normal. 

 

In any case OP should run firestrike and post the results - because they seem to think it's stable but didn't actually test it. 

 

 

Ran firestrike and Superposition, no crash or artifacts. Umm, I know its stable and cool, its been a week now.

Also ran Aida64 Extreme system stability/GPU test. No issues.'

 

So heres Assetto Corsa max settings and with 20 cars, more cars, the usage goes up.

So the poster saying it needs to be at 99% is kinda wrong. If it were 99% that would mean the GPU is capped out (now in bottleneck territory).

 

 

[Praesi]

4 minutes ago, Bender Blues said:

Ran firestrike and Superposition, no crash or artifacts. Umm, I know its stable and cool, its been a week now.

Also ran Aida64 Extreme system stability/GPU test. No issues.'

 

So heres Assetto Corsa max settings and with 20 cars, more cars, the usage goes up.

So the poster saying it needs to be at 99% is kinda wrong. If it were 99% that would mean the GPU is capped out (now in bottleneck territory).

 

 

Thats not how this Works.

[Bender Blues]

Just now, Praesi said:

Thats not how this Works.

Well then, care to explain?

[vK 3 1 RON]

18 minutes ago, Bender Blues said:

Well then, care to explain?

Umm.. I might be mis-interpreting this whole thread, I kinda skimmed through it, and it's like 1AM for me.

 

But, if you're saying that a GPU can be sub-99% (100%) utilisation and be considered 'stable' then I have some bad news for you... It's not a bottleneck if the GPU is at 100% utilisation, it means that the GPU is drawing the maximum amount of wattage that it can and boosting as high as it can to drive the absolute apex-performance that it can. When a GPU draws this extra power, the stability can decrease, for example, my CPU can be stable @4.7GHz at 1.3V in desktop, but when I start running an intensive benchmark or game, it crashes windows. Because the overclock has become unstable. At least, that's how I've kind of interpreted this, feel free to correct me @Anyoneelse 

 

Edit: I understand you referenced firestrike and something else, as well as Assetto Corsa using not 100%, that could be for a lot of reasons why the GPU isn't going 100% into it, but again, I might be wrong - and I'm very very tired haha  

[Harry Voyager]

What is the CPU it's paired with, what's the CPU usage looking like and are you using v-sync during benchmarks? 

 

As I recall, Assetto Corsa tends to be CPU limited rather than GPU limited. When I was testing my performance in IL-2, when I had my 1080 to paired to a 4770K, the GPU was running at about 60% loaded, and the primary thread was running one CPU core at about 100%. Now that I've got it paired with a 3800X and a Reverb, the GPU generally runs at 100% loaded unless to many bombers are in combat and start flooding the CPU. 

 

In Fire strike, however, the system is generally running the GPU at 100%

[Bender Blues]

I'll reply to you both above, thank you. 

 

But what the heck am I seeing here?

This test is with Aida64 Extreme GPU only stress test.

 

As you can see, the GPU usage shoots to 99% on the start.

But what is going on with GPU power draw? Shouldn't that be pegged to???

 

[xAcid9]

Just run Firestrike and post your result link. Overclocking is a waste of time if there is no performance boost. 

 

Also you can't really compare clock speed between 1070 and 5700XT, my 980 Ti is only clock at 1400mhz but it almost as fast as your 1070. 

[Bender Blues]

2 hours ago, xAcid9 said:

 

Also you can't really compare clock speed between 1070 and 5700XT, my 980 Ti is only clock at 1400mhz but it almost as fast as your 1070. 

I wasn't comparing, please read my original post again. I was watching Jays vid on the 5700 and it got me thinking about clocks.

And was wondering if 2100mhz is decent on a 1070 - which it is. Not that its better than a 5700.

 

Well 1400mhz compared to 2100 mhz isn't almost. Are you implying that a 980ti is almost as fast as a 1070ftw card? lol

[xAcid9]

35 minutes ago, Bender Blues said:

I wasn't comparing, please read my original post again. I was watching Jays vid on the 5700 and it got me thinking about clocks.

And was wondering if 2100mhz is decent on a 1070 - which it is. Not that its better than a 5700.

I mean ~2100mhz with light load on 1070 is not silicon lottery. 

Also i've seen many cases where user claimed they get silicon lottery with high clock speed but when run benchmark the performance actually decreased over stock clock which mean the overclock is not stable. 

 

35 minutes ago, Bender Blues said:

Well 1400mhz compared to 2100 mhz isn't almost. Are you implying that a 980ti is almost as fast as a 1070ftw card? lol

Yes.