Unigine Valley Benchmark Scores Thread + SUPERPOSITION ***Over 1000 Submissions!***

[Jumper118]

4 hours ago, Simon771 said:

Any problems with running 8 core CPU at 4,0GHz on that motherboard? It doesn't have the best VRM, and just wondering if it was stable or if you had any problems.

Also what voltages did you use on R7 1700, and what temps?

 

Sorry for so many questions

It was fine tbh. Was using a Noctua low profile cooler as well. It was stable at 4ghz on 1.375v and I never looked at temps but it ran loads of multi thread benchmarks at 4ghz fine and for wprime 32m and super pi it did 4.025ghz. Vrm was cool all the time. I will post a picture of it. 

 

 

[Simon771]

4 minutes ago, Jumper118 said:

It was fine tbh. Was using a Noctua low profile cooler as well. It was stable at 4ghz on 1.375v and I never looked at temps but it ran loads of multi thread benchmarks at 4ghz fine and for wprime 32m and super pi it did 4.025ghz. Vrm was cool all the time. I will post a picture of it. 

 

Sounds great

I'm just a bit worried since I'm going with B350M mortar, which have pretty much the same VRM.

But I'm only interested in running it at around 3,8GHz, since I heard that going from 3,8 to 3,9 increases temps by around 12°C + much higher voltage.

[Jumper118]

2 hours ago, Simon771 said:

Sounds great

I'm just a bit worried since I'm going with B350M mortar, which have pretty much the same VRM.

But I'm only interested in running it at around 3,8GHz, since I heard that going from 3,8 to 3,9 increases temps by around 12°C + much higher voltage.

It depends on how good the cpu is and howuch voltage it needs. 1.35-1.375 can be done on the stock cooler

[Simon771]

18 minutes ago, Jumper118 said:

It depends on how good the cpu is and howuch voltage it needs. 1.35-1.375 can be done on the stock cooler

Cooling it shouldn't be the problem since I'm going to use Dark Rock 3. 2 intake fans at the front: 1 for providing air to the GPU and another one for the CPU. And then 1 exaust fan at the back to move hot air out of the case faster.

I'm pretty sure I could cool down even 1,45V on that CPU, but temps would get to 80°C and above, so I will try to stay under 1,4V.

[Cheddle]

third card on air. using low-bandwidth bridge. cpu @ 4.25ghz

[Cheddle]

and 4k and 8k. 8k was a stutter fest as shown in the 8k screen shot I have highlighted the frametimes. the blue box is 4k and the red box is 8k

 

probably bad due to my 8x/8x/9x pcie lanes combined with a non-HB bridge for 3-way Pascal SLI.

[Jumper118]

1 hour ago, Cheddle said:

third card on air. using low-bandwidth bridge. cpu @ 4.25ghz

Nice to see a 3 way setup

[done12many2]

I replaced my last pair of 1080 Tis with a new pair.  No shunt mod on the new pair, but they would definitely benefit from it as they are hitting power limits during Superposition 4k and 8k tests.  With that said, the new cards are still faster with no shunt mod then the old ones were with it. 

 

1080p Extreme

CPU:  7700k @ 5.4 GHz 
GPU(s):  GTX 1080 Ti SLI
Average FPS:  95.72 FPS
Score:  12,798

 

4k Optimized

CPU:  7700k @ 5.4 GHz 
GPU(s):  GTX 1080 Ti SLI
Average FPS:  157.66 FPS
Score:  21,078

 

8k Optimized

CPU:  7700k @ 5.4 GHz 
GPU(s):  GTX 1080 Ti SLI
Average FPS:  72.02 FPS
Score:  9,628

 

 

 

[flibberdipper]

6 hours ago, Cheddle said:

and 4k and 8k. 8k was a stutter fest as shown in the 8k screen shot I have highlighted the frametimes. the blue box is 4k and the red box is 8k

 

probably bad due to my 8x/8x/9x pcie lanes combined with a non-HB bridge for 3-way Pascal SLI.

 

 

How the sam hell did you manage to get 3 way Pascal working? I thot Novideo limited Pascal to 2?

[done12many2]

14 minutes ago, tmcclelland455 said:

How the sam hell did you manage to get 3 way Pascal working? I thot Novideo limited Pascal to 2?

 

You can run it.  Nvidia just doesn't support it.  You are also forced to use low bandwidth bridges as HB bridges only work for 2 way.

[flibberdipper]

1 hour ago, done12many2 said:

 

You can run it.  Nvidia just doesn't support it.  You are also forced to use low bandwidth bridges as HB bridges only work for 2 way.

Huh.

[Jumper118]

you can run 4 way as well  

[PCGuy_5960]

4 hours ago, done12many2 said:

1080p Extreme

CPU:  7700k @ 5.4 GHz 
GPU(s):  GTX 1080 SLI
Average FPS:  95.72 FPS
Score:  12,798

 

4k Optimized

CPU:  7700k @ 5.4 GHz 
GPU(s):  GTX 1080 SLI
Average FPS:  157.66 FPS
Score:  21,078

 

8k Optimized

CPU:  7700k @ 5.4 GHz 
GPU(s):  GTX 1080 SLI
Average FPS:  72.02 FPS
Score:  9,628

I am sure that you mean 1080 Ti SLI, not 1080 SLI

[PCGuy_5960]

LMAO, this guy has 100k subs and doesn't understand that core count and performance don't scale linearly  And on top that, Vega doesn't have double the cores of a 480, it has only 77% more cores, so at best Vega would be 50-60% faster than a 480.

[done12many2]

1 minute ago, PCGuy_5960 said:

LMAO, this guy has 100k subs and doesn't understand that core count and performance don't scale linearly  And on top that, Vega doesn't have double the cores of a 480, it has only 77% more cores, so at best Vega would be 50-60% faster than a 480.

 

Same guy who produced those Ryzen initial benchmark videos that contradicted everyone else's.  He got some free chips from AMD and got stupid.  

[PCGuy_5960]

Just now, done12many2 said:

Same guy who produced those Ryzen initial benchmark videos that contradicted everyone else's.  He got some free chips from AMD and got stupid.  

Yeah, I bet that he is going to benchmark the 1080 Ti with the fan speed set to 5% and watercool and overclock the Vega card so that he can "prove" that Vega is faster than a 1080 Ti

[Jumper118]

9 minutes ago, PCGuy_5960 said:

LMAO, this guy has 100k subs and doesn't understand that core count and performance don't scale linearly  And on top that, Vega doesn't have double the cores of a 480, it has only 77% more cores, so at best Vega would be 50-60% faster than a 480.

doesnt quite work like that  it has hbm 2.0 memory as well as the extra cores. it will be 2x480 +/-10% not 100% sure they are exact same cores as rx480/580 either, i supect they will be a slightly more efficient version. like the 290X vs fury. both 28nm, but a a fury uses less power than a 290x even with all the extra cores.  

 

but yes he is called the joker and when Bryan went to stay with him in the us, all he wanted to do was get 420'ed and smash  

[PCGuy_5960]

Just now, Jumper118 said:

doesnt quite work like that  it has hbm 2.0 memory as well as the extra cores. it will be 2x480 +/-10% not 100% sure they are exact same cores as rx480/580 either, i supect they will be a slightly more efficient version. like the 290X vs fury. both 28nm, but a a fury uses less power than a 290x even with all the extra cores.  

Yes, but it almost definitely won't be 2x480. For example the 1080 has GDDR5X and double the core count of a 1060, but it performs 60-65% better than a 1060. The Vega card has 77% more cores than a 480 and HBM2. The 1060 and 480 trade blows and because the Vega card will have more efficient cores than a 480, it will probably be as fast as a 1080 or a bit slower than a 1080 (or 480 CF) but it won't be as fast as a 1080 Ti....

[Jumper118]

4 minutes ago, PCGuy_5960 said:

Yes, but it almost definitely won't be 2x480. For example the 1080 has GDDR5X and double the core count of a 1060, but it performs 60-65% better than a 1060. The Vega card has 77% more cores than a 480 and HBM2. The 1060 and 480 trade blows and because the Vega card will have more efficient cores than a 480, it will probably be as fast as a 1080 or a bit slower than a 1080 (or 480 CF) but it won't be as fast as a 1080 Ti....

i was under the impression that 2x480 was = gtx1080p performance. 

[PCGuy_5960]

Just now, Jumper118 said:

i was under the impression that 2x480 was = gtx1080p performance. 

No. 1 480+70-80% 480=1080. Crossfire and SLI don't scale 100% 

[Jumper118]

Just now, PCGuy_5960 said:

No. 1 480+70-80% 480=1080. Crossfire and SLI don't scale 100% 

ok when i say 2x480 i mean 480cf  in which case we are agreed but misunderstanded

[PCGuy_5960]

5 minutes ago, Jumper118 said:

in which case we are agreed but misunderstanded

Yeah, exactly

[VagabondWraith]

4 hours ago, PCGuy_5960 said:

Yeah, I bet that he is going to benchmark the 1080 Ti with the fan speed set to 5% and watercool and overclock the Vega card so that he can "prove" that Vega is faster than a 1080 Ti

You can't take that guy seriously. He's as bad as Logan from Tek Syndicate and his infamous video comparing the 8350 to the 3570K. It's painfully obvious they were paid and/or rewarded immensely for their shilling er I mean work.

 

This video is an all-time classic!

 

 

[Cheddle]

10 hours ago, tmcclelland455 said:

How the sam hell did you manage to get 3 way Pascal working? I thot Novideo limited Pascal to 2?

its only 'supported' for 'benchmark' applications, literally a list of titles Nvidia have approved as 4 way allowed. and when running games it only loads up two cards (out of the box any way  )

10 hours ago, done12many2 said:

 

You can run it.  Nvidia just doesn't support it.  You are also forced to use low bandwidth bridges as HB bridges only work for 2 way.

Technically there is also a way to 'trick' the cards into thinking they have a 'HB' bridge and run the bridges at the higher clock rate. Either you can chop up a few bridge and 'make' your own - or you need to use a bridge with the metal grounding connectors - the newer MSI and EVGA v2 bridges both detect as a HB bridge - a 3 and 4 way HB bridge (something Nvidia never made)

 

8 hours ago, Jumper118 said:

you can run 4 way as well  

 

Indeed!

the magic trick to running 3 and 4 way in games is to modify the SLI profile to tell the drivers that you are running a benchmark - thats easy - but making the rest of the profile 'compatable' with 3/4 way SLI is not easy and takes a lot of trial/erorr and looking at how other games profiles are set to run... Usually you can just use another game with the same engine as a starting point (unreal, unity, frostbite ect...) - I can DEFINITLY apprecaite why Nvidia dropped 3 and 4 way SLI support - its NOT an easy job getting even the driver references correct, let alone writing the actual code that they reference to make the games work propperly...

 

so far ive managed to get Metro Last light, superposition, project cars - all working with good scaling. I cannot get any Frostbite engine game scaling well and im still working on rise of the tomb raider - I am 99% sure this not scaling well becaues I currently am using an old school non-LED bridge and I have a 5820k (8x8x8x) - im trying to get my hands on a 40-lane CPU to test out 16x/16x/8x to see if the scaling improves and I have the V2 bridge on its way to me now.

FWIW im seeing 70% CPU usage im timespy graphics tests lol! I didnt think my 5820k would become a bottleneck for a LONG time! haah

In superposition I mostly got good scaling. Except at 8k where it was a stutter fest.

[flibberdipper]

52 minutes ago, Cheddle said:

its only 'supported' for 'benchmark' applications, literally a list of titles Nvidia have approved as 4 way allowed. and when running games it only loads up two cards (out of the box any way  )

Huh. Again back to the whole the more you know meme thing. I actually didn't know this.