Poor returns

[Bobsnail]

I've been overclocking since the pentium 75 (upto 120 on air...nice ) but i have never really had good results... ive always put it down to cheap components (usually RAM,or poor CPU steppings)

For once i decided to spec myself a decent system. As standard i get around 6800 in 3DMark 2005. When overclocked (significantly) i still only get around 6950.

Opteron 146 CABNE 0545BPMW
Corsair TWINX2048-3200C2PT
Asus A8N-E
Gigabyte 7800GT
300 Gig Maxtor Diamond Max 10 + few other drives
Soundblaster LIVE is the only PCI card
Nothing else important

I followed the overclocking thread on here and overclocked to 2.8 with relative ease, i even overclocked *slightly* my video card. Ran 3D mark and saw on a few percent increase in performance only (just over 2 percent)

Does that sound about right in 3DMARK 2005?
Any ideas what the matter could be?

I think my most successfull benchmark was with 310 * 9 and .66 memory divider (133). This clockd my Ram and about defult speeds, and i tightened the memory timings to 2-3-3-8 and 1T

The performance differance was so negligable that im currently running my rig 'nomrally' to keep it cooler. Will have another play this weekend though, with anything you guys can suggest.

Many thanks in advance

 

[sandorski]

3DMark2k5 is more a Video Card benchmark than a CPU benchmarkk. Use more CPU intensive tasks and you'llnotice a much better improvement.

 

[lopri]

3DMark05 also does CPU test. Did you compare the scores?
Yah your increase seems a bit low (assuming you OC'ed from 2.0GHz to 2.8GHz), but not totally out of the round. I thought a 100MHz increase in CPU speed (AMD) would yield about 50~100 more points in 3DMark05.

 

[Bobsnail]

Yeah thats going from 2.0 to 2.8 Gig. There was a slight improvement on the CPU benches, but it certainly wasnt noticeable to the eye. Hovered around 1- 2 fps on the canyon one and i think between 5 and 8 fps on the space hanger one.

I will try some more CPU intensive benchmarks over the weekend then.

Just i remember seeing a guy with an FX-57 (clocked to 3 Gig) get 12k in 3Dmark 2005...(no link sorry, though i think i can find it again if your interested). He did have a 7800GTX... i guess it may have been SLI. He provided all the usual screen prints etc.. in fact i have them on my desktop as inpiriation lol

Just annoying i cant hit 7000 despite starting at 6800 (i had hoped to hit 8000ish)

Cheers for you replies

 

[TSS]

Try to OC your vidcard more, maybe get a better cooler for it if you havent already. that should get you a little more 3Dmarks if you want those

to see how much your CPU benifits, boot up a recent game you have, run a benchmark 800x600 no AA no AF @ 2.0Ghz and then @ 2.8Ghz. that should show you some definate improvements in framerate since at those settings it'll be CPU limited.

else run encoding benchmarks. getting 800mhz exta will definatly show you a increase.

 

[Wentelteefje]

Yeah, overclocking the GT will give you much better results than overclocking the CPU... In games this is the same... They hit quite high...

 

[Bobsnail]

Im Aircooled and trying to run at stock volts. I have an artic cooler which is just huge. Must admit i seem to have a heat problem tho. My CPU is currently at 40 Degrees idle.

That said my Mobo temp is 42 Degree so i guess CPU isnt doing to bad. Ive just thrown the PC together.. i will re-route cable and power up the 3 case fans soon... Hopefully reduced the temp in the case and so the CPU temp.

Does voltage requirement go up with heat? (I KNOW that heat goes up with voltage btw)... I say that cos im on auto volts at the moment. And the hotter the CPU gets the more volts it seems to draw... interesting.

I will check out the vid thread for hint on overclocking the Gigabyte 7800GT cheers. To be honest ive not had much luck there either. Ive take core up quite a bit... but the memory seems to be quite sensitive.

Cheers again all

 

[Wentelteefje]

Which AC cooler? Yeah, connect those case fans and temps should lower quite a bit... 40°C is indeed high... Your "Voltage up with heat" theory seems strange to me, but I do know that the comp might supply more volts when under load...

Here you have some reviews on the GigaByte 7800GT... While YMMV of course, they should be pretty good guidelines... Stock is of course 400/1000MHz...

AMDZone: 440/1050MHz
Bjorn3d: 455/1075MHz
HotHardware: 473/1130MHz
PlanetAMD64: 450/1200MHz
OCWorkBench: "any core clock speed over 470MHz hangs..."

 

[GuitarDaddy]

As others have said 3dmark is all about the video card, O/C the video card and you will see some nice improvements.

If you want to see the effect of your CPU overclock run SuperPI, you will see a huge difference from 2.0 to 2.8. And as said before if you make graphics benches CPU dependent by dropping to the lowest available resolution and turning off AA and AF you will also see a large difference.

The bottom line is overclocking the CPU makes very little difference in games and gaming benchmarks. However if you do alot of encoding/rendering or other CPU intense apps you will see a significant improvement from overclocking.

 

[Wentelteefje]

Originally posted by: GuitarDaddy
As others have said 3dmark is all about the video card, O/C the video card and you will see some nice improvements.

If you want to see the effect of your CPU overclock run SuperPI, you will see a huge difference from 2.0 to 2.8. And as said before if you make graphics benches CPU dependent by dropping to the lowest available resolution and turning off AA and AF you will also see a large difference.

The bottom line is overclocking the CPU makes very little difference in games and gaming benchmarks. However if you do alot of encoding/rendering or other CPU intense apps you will see a significant improvement from overclocking.

QFT :thumbsup:

 

[Bobsnail]

Cheers all

Wentelteefje good skills with the links im running core at 450 and memory at 1100.. So i guess what i thought was a modest overclock is actaully a very good one lol.

Im still a bit confused how my system is getting trounced when i have a decent overclock and good components. Im running at FX57 speeds, with corsair xtreme ram and the vid card reviews you pulled out for me say the card is very good.

Im not too bothered i just expected more is all.

Never mind...thanks for your time everyone

 

[Bobsnail]

Oh and Wentelteefje my arctic cooler is the Freezer 64 Pro. Its almost as big as my PSU lol. Its kinda copper base and pipes running through lots of aluminium fins. The fan sits on the side (as opposed to the top) blowing air towrds the back of the case. Clever idea i guess (but im easily impressed) lol

 

[Wentelteefje]

Originally posted by: Bobsnail
Oh and Wentelteefje my arctic cooler is the Freezer 64 Pro. Its almost as big as my PSU lol. Its kinda copper base and pipes running through lots of aluminium fins. The fan sits on the side (as opposed to the top) blowing air towrds the back of the case. Clever idea i guess (but im easily impressed) lol

They simply rule! I couldn't agree more... I have the non-Pro Freezer 64...

EDIT: Man, forget about 3DMark... I'm sure games simply run great with your setup...

Welcome to the forums BTW!

 

[rmed64]

Is that your 3dmark 2005 score or 3dmark2005 CPU score? An 800Mhz increase in cpu speed gave me ~600 more points in CPU test.

 

[CP5670]

I actually got a big gain in 3dmark05 from overclocking my processor. The stock 2ghz speed gave a score of 9900 while the 2.8ghz speed yields over 13000, with the video cards at the same speeds in both cases. The performance improvements in games (during actual gameplay, not benchmarks) were generally much less noticeable though. I can mainly see big differences in performance with emulators and numerical computation programs.

 

[aka1nas]

You would only see a huge gain in 3dmark05 if your CPU was slow enough to bottleneck the card at the resolution and settings that 3dmark05 tests at. You will not get much improvement in scores beyond that point.

 

[nerp]

I see a huge difference in Source games like CS:S and HL2 with my opteron 148 @ 2.8 instead of 2.2 Ghz.

 

[LumbergTech]

Originally posted by: nerp
I see a huge difference in Source games like CS:S and HL2 with my opteron 148 @ 2.8 instead of 2.2 Ghz.

same here..i oced my opteron 144 from 1.8 ghz to 2.7 ghz and it stopped the game from dipping into the 30s in firefights etc..now it never goes under 55 and generally stays at like 70 1680/1050 8xaf 2xaa

 

[BOLt]

nerp, those are amazing OC figures in your sig for the proc and gpu... and at STOCK volts? goshdarnit i need an opteron! your video card OC is amazing too, man... well done!

 

[TSS]

some games are more CPU dependant than others...

games with the source engine (HL2, CS:S, DOD:S) will see a improvement at just about any resolution and IQ because its very CPU dependant.

games based on the doom 3 engine (Doom 3, Quake 4) and FEAR will see very little improvement since they are easily GPU limited.

i'd recon the difference comes from the physics used in HL2 which will put a strain on the CPU. HL2 uses higher res texture with lower polygon models, while doom 3 goes with higher polygon models vs lower res textures. which is why HL2 benifits more from a strong CPU and has lower PC requirements overall.

 

[mrkun]

I think there's something wrong if you can't hit 7k on 3DMark05. Your system should hit that at stock settings.