GPU Benchmark Thread

I thought everyone only had two i's? Christ do I have an impairment lol

@thelonewanderer

Here's your 8800 Ultra benchmark.

GPU-Z is reporting the hardware wrong though. Core config should be 24/32 @ 80nm.

1 Like

Heh, considerably faster than a 9800+.
I still lust for a 8800 ultra, or three.
DX11 rendering????

Also 8800ultra/gtx are 90nm and not 80.

Yeah I just noticed I typed 80nm. I was very tired.

Also the 9800+ suffers from 512MB on a 256-bit bus. Heaven murders 512MB cards.

1 Like

I'm pretty sure you would get more or less the same results with a 1gb version.
8800gtx/ultra was always better than the 9800+.

This is how I know it wouldn't be the same.

9600 GT 1GB just about beats the 9800+ 512MB. Heaven uses 740MB average at these settings, the 512MB bottlenecks the card like crazy.

1 Like

Hmm, true.
The 9600 gt is ridiculously strong for what it was nonetheless, a 9800 gtx cut in half more or less.
64 shaders, hm.

It is an incredible card for the time. Beating out the outstanding 8800 GT (with the help of a massive 150MHz overclock compared to a generic 9600).

I feel if this were a standard 512MB card it would be scoring in the 260 range below the 8800 GTS.

Well I remembered that I have a GTS 250, which is just a rebadged 9800 GTX+ with twice the VRAM. This is excellent, since that was the perceived bottleneck.

Despite that it doesn't overclock as well as the EVGA 9800 GTX+ with 512MB, it still gave a very definitive result for WHY the 9800 can't keep up with the 8800 Ultra.

So here we go:


This makes the 8800 Ultra look really good. The 9800 GTX+ (GTS 250) with a healthy overclock can hardly close the gap trailing by 62 points, but it does score significantly better than the 512MB version, clearing it by 59 points. The extra RAM essentially puts the performance straight down the middle.

This leaves me wondering what to blame for the performance deficiency. Since memory capacity is accounted for, how about memory speed? Let's drop the memory clock to the 9800 GT's level and see if there's a significant impact. Seeing as the 9800 GTX+ trails the 8800 Ultra's throughput by around 32GB/s, this should show if the performance is lost in that difference.


And there it is. The performance tanks. The 8800 Ultra is definitely faster clock-for-clock, but the extra memory bandwidth really seals the deal when it comes down to outclassing the 9800 GTX+.

1 Like

Well yes, the wider bus is really what made the G80 outlast the G92.
In games at the time 9800 gtx seemed better because it wasn't memory bottlenecked.

G94 matching a G92 chip when running at the same memory speeds puts the blame on the memory entirely.

Indeed. But when you strip away that advantage the 8800 Ultra does still come out on top, by the incredible amount of 4.8%.



So there you have it. G92 despite bringing many improvements, ends up slower as a result. Of course this is only one test, and to get a definitive answer I would need to run a whole suite of tests... Which I really don't want to do right now.

What I do want to do is attempt to overclock this 8800 Ultra now that it's back in my PC. Not much thermal headroom, but I think I can get another 40MHz out of it.

As long as you don't change the voltages you're fine, they're all set on a hardware level anyway.
Changing the frequency alone is never going to hurt it, and G80 seemed to me to be a pretty stout OCer.

My old 8800 gtx with a zalman cooler did 640 core 1050 memory and a blistering 1750 on the shaders.

Meh. Started stuttering and pushing into the 78C range. I know this card crashes at 81C. Maybe I'll get it under water to play with it some more later on.

Locked the fan at 100%?

Yep. Card idles at 52C, just needs some TLC and a bit more cooling strapped to it. This is pretty much in line with what people were getting when they were new, so I'm not complaining. I feel the gains don't really reflect the OC much.

10.3% Core OC, 14.28% Shader OC, and 7.5% Memory OC for a 6.5% increase in average FPS. I had to backdate drivers to 197.45 to get RivaTuner active, but I did test the original overclock and the score had only changed by 2 points. Dunno, maybe it comes back to a memory bottleneck.

(Those are over stock 8800 Ultra values, not PNY's factory OC.)

Got my first numbers with my new Skylake build today, quick and dirty overclock has it running stable and nice. I'll update once I have the second card in and more solid tinkering.

6700k with an MSI 980 TI and MSI Z170A M7, 16 gigs of Corsair RAM.

I went back and spent a little more time with it, got a better result that ended up being ultimately more stable in real world situations.

XMP Enabled
4.6 on the processor.
NO Base Clock tampering until I've got both cards in (maybe not then), I imagine my settings will be different with two cards.

Video Settings in Afterburner:
Core Voltage: +56
Power Limit: 109
Temp Limit: 91 (topped out at 84)
Core Clock: +170
Memory Clock: +600
Fan Speed: Auto

1 Like

Awesome overclock! Very impressive memory OC.

I'm actually putting you above my QuadFire setup since your max FPS ended up higher, congrats on having the highest single-GPU score on the chart!

@SoulFallen has been beaten. :D

That's it, it's settled, next purchase is a water block.