MSI Kombustor- odd results

mojo1990

New member
I have a Gainward GTX 580 Phantom...

When I first set the card to its default Factory OC using MSI afterburner, my results on the MSI Kombustor benchmark software were very good. However when I increased the Voltage of the card, and subsequently increased the Clock speed, my results showed deterioration instead of gain...

Default Voltage: 1000, clock speed: 800Mhz and memory 2010 (I think)...

I will post some benchmark pics later (I am using my laptop at the moment) for comparison sake.

I understand that increase in voltage would result in higher temps, but would that also affect the performance under heavy stress tools like Kombustor?

Any help on this would be appreciated.
 
shouldnt see a devrease off the bat unless something is bottlenecking it. I noticed it with my 560Ti after I went past 1026MHz on the core I started getting a diminished benchmark score. Which stood to reason my cpu oc needed more. Which unfortunatly I had already hit that wall. Might be a bottleneck but hard to say without knowing more about ur rigs oc.
 
shouldnt see a devrease off the bat unless something is bottlenecking it. I noticed it with my 560Ti after I went past 1026MHz on the core I started getting a diminished benchmark score. Which stood to reason my cpu oc needed more. Which unfortunatly I had already hit that wall. Might be a bottleneck but hard to say without knowing more about ur rigs oc.

Interesting

I have a 2500k clocked at 4.2 GHz I simply put the multi at 42 with the default voltage. I have been told my Gigabyte UD7 can handle that
cool.gif
without any need to alter voltages...

However in regards to my GPU, well I have been told the higher voltage + higher clock, the better the performance in terms of FPS.

Default Factory settings

msidefualygpuclock.jpg


Results

msikomnormalpresetdefau.jpg


Over clocked Settings

msioverclocked.jpg


Results

msikomnormalpresetoverc.jpg


My first score was 5775 and my second score after having applied a over clock was 5740...
sad.gif
 
The 580 has protection against programs such as furmark/combustor etc. as they call them thermal virus' or something, therefore it'll never increase it'll just stay at the threshold where the protection kicks in!

I had a 480 previous which didn't have protection and the temps were scary. The power draw from the socket was also scary
smile.gif


Use another program for benchmarking, combustor is no good.

Out of interst what voltage you using to get 1000 core on your phantom?
 
The 5 series don't just down throttle for Kombustor or Furmark but everything.

Over the past few years it has been made mandatory for makers of electronic equipment to use ROHS lead free solder. Lead free solder is nowhere near as soft and flexible as real leaded solder.

That, coupled with lots of heat have led to a very high failure rate in GPUs, simply due to the heat they generate and the flexing that occurs on the PCB of a GPU. The PCB changes shape and size as it heats and contracts and eventually breaks the solder loose, causing artefacts and lines. This can then be bodged by oven baking but it never lasts.

It was this failure rate that put BFG out of business during the wait for Fermi. All they were doing for nearly a year was replacing video cards with stock they had. Nvidia had stopped making the 200 series cards and during that patch it put immense pressure on Nvidia only card makers. This is why XFX switched to AMD and have stayed there.

Any way yes, there isn't much point in overclocking a 5 series GPU. Nvidia have made sure that their returns will be far lower by throttling the card and I can't say I blame them really. AMD have done the same also with the 6 series. /old git mode on. I remember back in the day the Radeon 9 series had a fuse onboard. If you overclocked it the fuse would blow and they wouldn't replace your card if it failed.
 
The 5 series don't just down throttle for Kombustor or Furmark but everything.

Over the past few years it has been made mandatory for makers of electronic equipment to use ROHS lead free solder. Lead free solder is nowhere near as soft and flexible as real leaded solder.

That, coupled with lots of heat have led to a very high failure rate in GPUs, simply due to the heat they generate and the flexing that occurs on the PCB of a GPU. The PCB changes shape and size as it heats and contracts and eventually breaks the solder loose, causing artefacts and lines. This can then be bodged by oven baking but it never lasts.

It was this failure rate that put BFG out of business during the wait for Fermi. All they were doing for nearly a year was replacing video cards with stock they had. Nvidia had stopped making the 200 series cards and during that patch it put immense pressure on Nvidia only card makers. This is why XFX switched to AMD and have stayed there.

Any way yes, there isn't much point in overclocking a 5 series GPU. Nvidia have made sure that their returns will be far lower by throttling the card and I can't say I blame them really. AMD have done the same also with the 6 series. /old git mode on. I remember back in the day the Radeon 9 series had a fuse onboard. If you overclocked it the fuse would blow and they wouldn't replace your card if it failed.

So there is no real point in Over clocking then (on air)? Hmm I was considering water cooling maybe that is a better solution to stop heat from causing the solder to break
rolleyes.gif
 
The 580 has protection against programs such as furmark/combustor etc. as they call them thermal virus' or something, therefore it'll never increase it'll just stay at the threshold where the protection kicks in!

I had a 480 previous which didn't have protection and the temps were scary. The power draw from the socket was also scary
smile.gif


Use another program for benchmarking, combustor is no good.

Out of interst what voltage you using to get 1000 core on your phantom?

I set the core Voltage to 1000 using MSI. The default factory settings (by gainward) are the same as the first pic... I was told by an Online guide that 1138 mV is the practical maximum you can go when benchmarking without bottleneck...
 
Back
Top