• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

4870 massively underperforming

n1kunj

n1kunj

n1kunj

Associate
Joined
9 Aug 2005
Posts
621
Location
Surrey
[email protected]
2GB RAM @938MHz
4870@790/1080@16xPCI-E

And I only get 12k in 3dmark06!

Here's an equivalent system

CPU score on mine is mildly lower than theirs as I've got slightly less cache, but that sure as hell doesn't explain why there's a 3k+ gap! My SM2 and 3 scores are 1k and 3k! lower as well.

I've also tried the 8.7s, same deal. I've formatted since getting the card, pre format it was the same as post format.

Is there anything blindingly obvious I've missed? I'm using Cat 8.8s, no hotfixes or anything - is there one I should know about?
 
Same proccy speed though - only difference is the cache - take a look at the CPU scores, only a slight difference.
 
No vista yet, unfortunately so I can't run vantage - however look at the equivalent system, same card and proccy speed on 06 but massively lower performance for me.

Could it actually be me running XP instead of Vista that's causing the issues?
 
n1kunj said:
No vista yet, unfortunately so I can't run vantage - however look at the equivalent system, same card and proccy speed on 06 but massively lower performance for me.

Could it actually be me running XP instead of Vista that's causing the issues?
Click to expand...

lol good lord no. XP is still the best system for gaming. check to see what other processes you have running aso what psu your running. if its less than 450w that could lower the score
 
Something is deffo wrong, because the the SM3.0 mark shouldn't really be affected that much by extra CPU cache.

Personally, I'd put everything back to stock and reinstall windows.

EDIT: OOps nevermind, you've already tried that.

Could just be some bug with 3dmark06. Have you tried anything else. You could try Crysis and compare them with other people (in the Crysis benchmark thread floating around somewhere).
 
Last edited:
Brand new PSU yesterday as my old one went boom. Tis a 600W OCZ StealthXstream - my old was a 2 1/2 year old Hiper type R 580W so it's not that. Same performance before and after.

Process number at bootup is 29 and nothing too straining, just stock windows processes, ultramon, CCC and antivir - I'll try disabling that next and see if avira hates me.
 
3Dmark06 is terrible for HD4xxx series. I get less score with 4870 than 8800GTX and we all know 4870 destroys it FPS wise.

3DM06 favours Nvidia cards & Quad Cores so pointless trying to use it to benchmark. Try some games and use AA + vsync then you will see the big difference.

COD4, Bioshock or Assassins Creed all have big gains with 4870.
 
Why are the driver versions so different? Also look at the core and memory clocks, they're bizarrely wrong (for both scores), but also different. I would guess your card is suck in 2D mode, possibly due to out-of-date drivers that don't support the 4870.
 
Last edited:
Did a test, not my antivirus being evil, 3d clocks are kicking in fine according to GPU-Z, drivers are the latest available ones (8.8 cats). 8.7 cats have the same performance.

As with the drivers displayed in 3dmark, it's just being silly and not reading them right - I was using the 8.522.0.0 display driver according to device manager when the benchmark was run this morning(which is higher than the 8.512.0.0 aka cat 8.7s are) and still am.

I'll try looking for some other benchmarks and see if it's just 3dmark06 being silly on the whole.
 
n1kunj said:
Same proccy speed though - only difference is the cache - take a look at the CPU scores, only a slight difference.
Click to expand...

Its 15% faster clock for clock, it has 6MB Cache.

I only got like 300 3dmarks more than him and my 4870 was at 850/975 with a 4Ghz E8500, my SM3.0 and 2.0 scores where lower, CPU higher, 3dmark is useless for comparing anymore.
 
Last edited:
Your looking at this the wrong way. His hd4870 must be heavily overclocked and theres no way his e8400 is only running at 3.0ghz.

From what i've seen, your score is fine. Wouldnt be surprised if his was crossfire.

http://forums.techpowerup.com/showthread.php?t=63949

4870

Dual Core

1. 16967 - CBOT - E7200 @ 4.1ghz 4870 @ 790 Core 1100 Mem XP
2. 16105 - Oil ramasy - E8400 @ 4.0ghz 4870 @ 850 Core 1125 Mem vista
3. 15759 - Conroe - E8400 @ 4.29ghz 4870 @ 790 Core 1100 Mem vista
4. 14155 - Darknova - E8400 @ 3.6ghz 4870 @ 750 Core 900 Mem
5. 12810- Jmatt110 - Athlon x2 6000+ @ 3.4ghz 4870 @ 845 Core 1110 Mem Vista
6. 11216 - Polarman - Athlon FX-62 @ 3.0ghz 4870 @ 750 Core 900 Mem Vista
7. 11022 - paulo7 - E6300 @ 2.33Ghz 4870 @ 780 Core 1090 Mem XP
Click to expand...

Try doing some vantage tests and comparing those.
 
willhub said:
Its 15% faster clock for clock, it has 6MB Cache.
Click to expand...

Considering the difference between the comparison CPU and mine then his system is underperforming if it's 15% faster clock for clock - the actual difference is nowhere near that at 3GHz. Even if it were 15% faster that still doesn't explain the HUGE SM2.0 and 3.0 score differences,especially as mine is overclocked to near the overdrive limit.

I accept my system will have slightly lower performance than the comparison but 3k points is a little overboard - it's not a large enough difference to say the card is faulty though but not small enough to be just a difference caused by the CPUs - I know it can't be 3dmark being silly either as it would be silly for everybody elses 4870 as well!
 
I have a Q6600 overclocked to 3.4Ghz and a 4870. I get 17,000 in 06.

12k is not low for your setup imo. But I could be wrong...
 
Wasabi said:
I have a Q6600 overclocked to 3.4Ghz and a 4870. I get 17,000 in 06.

12k is not low for your setup imo. But I could be wrong...
Click to expand...

Nope you're right. The score is fine. I couldnt break 15k with a hd4850 running at 835mhz with an e8400 at 4.05ghz.

I can now, but only becasue of the quad.

If he RMA's his card, he'll just be left out of pocket.
 
I still think my card is underperforming because of the low graphics scores - and the graphics scores are completely independent of the CPU.

My SM2 score is 5108, comparison system is 6391 - a 1200 point difference
My SM3 score is 6279, comparison system is 9119 - a 2900 point difference
My CPU score is 2552, comparison system is 2764 - a 200 point difference

Same graphics card, slightly better proccy. Yet his graphics scores are monstrously higher - this cannot be explained by overclocking as it wouldn't provide such a big increase.

In fact I just ran a test with one core disabled - see this linky

My 2 core SM2 score is 5108, my 1 core score is 4950, a difference of 200 points
My 2 core SM3 score is 6279. my 1 core score is 6344, it went up! by 100 points
My 2 core CPU score is 2552, my 1 core score is 1335, a difference of 1200 points (2 cores is double the score of 1 core, obviously)

So CPU score is independent of the GFX scores - yet mine are still incredibly low compared to all the others.

As a final test to see if clock speed affected GPU score I underclocked my CPU from 3.0GHz to 2.25GHz (this is starting to border on excessive isn't it :()

My 3.0GHz SM2 score is 5108, my 2.25GHz score is 3927, a difference of 1200 points
My 3.0GHz SM3 score is 6279, my 2.25GHz score is 5286, a difference of 1000 points
My 3.0GHz CPU score is 2552, my 2.25GHz score is 1949, a difference of 600 points

Hang on. That means my original score was right!

Thank you, you wonderful and glorious people! You made me see that I'm a grade A moron and that my card is fine but my processor is bottlenecking the hell out of it! Yay for being excessive, it worked in my favour for once :) I completely ignored the fact the gfx tests were single threaded and that was making my score plummet (in all fairness, I didn't actually know about this until I did the underclocked benchies)

Roll on Nehalem :D

*Goes to overclock his processor even more*
 
Nice bit of analysis there :)

When I looked at the e2160 against the e6600 at the same clock speed (see sig) I found a bigger difference in 3d mark (albeit the earlier ones) than a lot of games. You'll probably find that although the epeen score is low, games are perfectly acceptable.
 
You must log in or register to reply here.
Back
Top Bottom