The first "proper" Kepler news Fri 17th Feb?

[CAT-THE-FIFTH]

Have you tried overclocking a little more as i am sure you could get a higher score with that setup. I am pretty sure that you are cpu limited though.

My DDR4 system RAM says no!!

Let me guess you are still using Win7 or 8?

If so you should get a more modern PC, seriously..its 2012!!!!!

+1

 

[Dan2kx]

Lol, for a second I thought someone has broken out the LN2 on an i3 lol! I was in shock!!

 

[ALXAndy]

windows 9? 8.1ghz i3?? I'm skipping bits of the thread but is this right?

haha it seems you can edit the results as they are HTML files. IE you could score 20000000000 with a ten million ghz CPU.

I just ran the test copying the settings and accidentally forgot to save the bloody results.

Any way, I scored around 1600 with my card at 1125 and stock I7 950.

However, my min FPS was 9. I'm telling you guys, CPU definitely plays a part. Max FPS was 172.

 

[Gregster]

I ran the benchmark again on my card again!

Why did you not pay the extra tenner and get the 128bit Win 9?

 

[CAT-THE-FIFTH]

snip

You need to run it at 1600X900.

 

[ALXAndy]

Okay I have no idea why my score is so low, but here it is. -__-

Not sure if this is a great comparison really.

Again, this is comparing a single stock GTX680 to a pair of stock GTX470's.

Your res is wrong mate. That pleb with the shop used 1600x900.

 

[CAT-THE-FIFTH]

Why did you not pay the extra tenner and get the 128bit Win 9?

I spent all my pocket money on the LN2!!

 

[thezakarum]

Your res is wrong mate. That pleb with the shop used 1600x900.

Damn it!

Here we go again then...

 

[ALXAndy]

Damn it!

Here we go again then...

LOOL I'm done mate. All that bloody fan revving and heat.

 

[Duff-Man]

However, my min FPS was 9. I'm telling you guys, CPU definitely plays a part. Max FPS was 172.

When you run the benchmark the first time through, there are some places where the data pages in. You will get small (~0.1s) pauses here, which leads to unrepresentative minimum framerates (i.e. ~10fps).

If you let the demo run through one complete cycle before starting the benchmark, you will get more realistic minimums (probably ~25fps).

 

[CAT-THE-FIFTH]

Damn it!

Here we go again then...

...on the Kepler merry-go-round!!

 

[ALXAndy]

Yeah that used to work OK. This new one though doesn't just jump back to the start and begin like the older Heaven used to. I used to do what you said, though. Wait for it to stutter and then smooth out before beginning the run.

Now though it stops and then loads. Gits.

 

[Cleeecooo]

haha it seems you can edit the results as they are HTML files. IE you could score 20000000000 with a ten million ghz CPU.

I just ran the test copying the settings and accidentally forgot to save the bloody results.

Any way, I scored around 1600 with my card at 1125 and stock I7 950.

However, my min FPS was 9. I'm telling you guys, CPU definitely plays a part. Max FPS was 172.

I'll help settle this next week. I have two pcs. The only real difference is one has a 2500K an the othe a 3930K

I am willing to vet there won't be a huge FPS difference. Everything else (bar case and PSU) shall remain te same.

 

[ALXAndy]

Good lad Cleeecoo. There probably won't be a huge difference in avg and max, but it's minimums I reakon you'll score well with.

 

[Cleeecooo]

Good lad Cleeecoo. There probably won't be a huge difference in avg and max, but it's minimums I reakon you'll score well with.

Any particular control variables needed?

So far I can think of:
RAM
GPU
SSD
heaven settings

 

[Duff-Man]

Yeah that used to work OK. This new one though doesn't just jump back to the start and begin like the older Heaven used to. I used to do what you said, though. Wait for it to stutter and then smooth out before beginning the run.

Now though it stops and then loads. Gits.

That's... rather crap

Still, if the data is fully loaded into VRAM during the run it should have the same effect. I'm not on a PC that I could test heaven 3.0 with right now though, so I can't check...

 

[thezakarum]

Okay so first pic is stock gtx470's.

Second pic is overclocked gtx470's (800mhz).

And this is the gtx680.

Either my 470's are broken or somthings fishy with the 680...

 

[ALXAndy]

Any particular control variables needed?

So far I can think of:
RAM
GPU
SSD
heaven settings

I run crap memory mate. 6gb 1333mhz Elpida. I used to run 1600 Gskill but I only had 3gb (basically I bought my board and ram etc when the memory prices flew up and 3gb cost me close to £60 !).

Mate of mine gave me the Elpida a couple of months back and I figured it was best to sacrifice the speed for the larger amount given I now run X64 Ultimate.

Specs..

I7 950 on a MSI Alienware X58 micro.
6gb ram
120gb Corsair Force SSD

Will upgrade my ram soon, but it looks like I'll have to buy four sticks and chuck one in the bin.

Okay so first pic is stock gtx470's.


Either my 470's are broken or somthings fishy with the 680...

Nothing fishy. Right now it seems that a 680 is 20% faster in Heaven than a 7970. Don't forget though there's AA running. And the 680 if anything like Nvidia's previous cards does AA better than AMD.

You need to remember that your cards are getting on now. Infact, IIRC on the default settings @ 1080p my 470 used to score around 1100 points. My 7970 at the same settings bashes out 2900.

 

[Duff-Man]

Either my 470's are broken or somthings fishy with the 680...

Why?



edit: Check out the performance of a single 470 vs a single 7970 (click me). As you can see there's a huge gulf in performance...

The 680 appears to be a fair chunk faster than the 7970 at the heaven benchmark, so it shouldn't come as a surprise that the 680 gets a higher score than SLI 470s.

 

[CAT-THE-FIFTH]

So the GK104 is the tessellation MASTER??

With the launch on the 22nd I need to get my final rumours list updated.