Sign in with
Sign up | Sign in
Your question

4850 3DMark06 Scores - disappointing?

Last response: in Graphics & Displays
Share
June 26, 2008 8:21:56 AM

Just popped in my new 4850 and ran a 3DMark06 run in XP on DX9. I'm kind of disappointed with the results:

TOTAL SCORE: 10067
SM 2.0 Score: 4340
SM 3.0 Score: 5035
CPU Score: 2238

I have the following specs:

AMD Athlon 64 X2 5000+BE @ 3.0GHz
MSI K9A2 Platinum 790FX
Corsair HX520W
Visiontek HD4850
4GB OCZ SLI-Edition DDR2 800

I expected something more around 12,000 or so with such an amazing videocard. On the Return to Proxycon test, frames dipped into the 20's on several occasions. The CPU tests have never been good. What I found baffling regarding the CPU tests is that my old Athlon 3700+ Sandy @ 3.2 scored around the same as my OC'd 5000+ BE. To me that seems odd.

I can't even beat a stock E8400 and an 8800GT according to Futuremark bench results. That can't be right....can it??

Temps for my 4850 idle at 71C and reach 85C under load, which seems in line with everyone else. I just really expected higher scores. I'm gonna run Vantage in Vista later to see how that fairs.
June 26, 2008 8:24:27 AM

Those scores looks very low to me, lower than what the card can do...
June 26, 2008 8:45:54 AM

Yeah...I agree. Not really sure what to do here though. Any chances the card isn't getting enough power from my Corsair HX520W? The PSU has three rails, each at 18amps.
Related resources
June 26, 2008 8:51:41 AM

CPU bound more or less, on my 8800GT i get 11k with an e2180 at 2.8ghz
7500 at the stock 2ghz.
June 26, 2008 8:53:57 AM

Yep, I agree with the Forum Resident.
What about FPS in games?
June 26, 2008 9:02:11 AM

Who cares? It's 3dmark, the world's most irrelevant benchmark. Try the Crysis GPU test.
June 26, 2008 9:23:43 AM

Hehe...
Stabgotham, this is a DX 10.1 card. It's meant to be used under Vista and DX 10.1 not XP and DX 9.
The shaders are optimized for DX 10. Try on Vista and see the results there. :) 
On the contrary with the previous DX versions, DX10 is NOT backwards compatible with DX 9. Because of this reason, the cards performance DX 9 is limited.
June 26, 2008 9:50:51 AM

Who cares about benchmarks. Your CPU is holding back the score. My E4600 @ stock and 8800GT @ stock don't break 10k. Just a couple hundred below it.
a b U Graphics card
a b à CPUs
June 26, 2008 11:02:00 AM

My scores with an E6600 @2.7 and 9600GT moderately overclocked to 750MHz on Vista:

3DMark score: 10227
SM2.0 score: 4863
SM3.0 score: 4585
CPU score: 2325

Notice that my SM2.0 score is alot higher than yours. I would think that the CPU would be more of a bottleneck in the SM2.0 tests because they are less stressful, and my CPU is probably better than yours at 3DMark. Also notice that you scored better in the SM3.0 score than your SM2.0 score... alot better. This is quite unusual.
a b U Graphics card
a b à CPUs
June 26, 2008 11:03:19 AM

duzcizgi said:
Because of this reason, the cards performance DX 9 is limited.

That's why DX10 is so much faster than DX9 eh? :sarcastic: 
June 26, 2008 11:23:42 AM

duzcizgi said:
Hehe...
Stabgotham, this is a DX 10.1 card. It's meant to be used under Vista and DX 10.1 not XP and DX 9.
The shaders are optimized for DX 10. Try on Vista and see the results there. :) 
On the contrary with the previous DX versions, DX10 is NOT backwards compatible with DX 9. Because of this reason, the cards performance DX 9 is limited.


LOL. And that really explains why 3dmark scores are higher on xp
a b U Graphics card
a b à CPUs
June 26, 2008 11:32:48 AM

DeafB4Dishonor said:
LOL. And that really explains why 3dmark scores are higher on xp

Funny thing is DX10.x doesn't have anything to do with 3DMark06 :lol:  Now maybe if it was DX10 Vista would pwn XP! :kaola: 
June 26, 2008 11:50:07 AM

Short story:
Check your BIOS version

Long story:
Had a similar problem with MSI K9A (580X based) 5200+ (@3GHz) and an X1950XT.
I had consistent results around 6K with Vista and 6300 with XP.
Then suddenly they fell to 4K.
I removed (Cleaned) Catalyst, installed previous versions, newer, Omega's etc but nothing.
I fell back to XP from Vista and from XP to Vista but nothing changed.
Then I thought that before this happened I had flashed the BIOS. I flashed an older version and problem was solved.
Of course I communicated the problem to MSI and as far as I know all further versions for my board are good.
June 26, 2008 12:07:22 PM

Well guys, my 2600 XT also showed that strange behavior on my computer. :)  Maybe it's because of my computer's configuration:
2GB DDR2 800
Athlon x2 4800+ (2.5 GHz)
HD2600XT 512 MB

I have both Vista and XP installations. On XP, my graphics card shows about 10% less performance than on Vista. So I have had the conclusion that it doesn't support DX 9 as well as DX 10. :) 
Please don't kill me OK?
a b U Graphics card
a b à CPUs
June 26, 2008 12:13:20 PM

In what applications? I highly doubt the applications you were using were DX10 anyway, or they would have crippled your card.
June 26, 2008 12:21:29 PM

:D  In fact it's Crysis :p 
It works terrible on XP (non playable) with medium settings, but works pretty fine (playable) on Vista with the same settings.
Resolution for both are 1280x1024.
Sorry don't ask me framerates, I didn't measure.
Just on XP I wasn't able to aim the gun before being killed by Koreans and on Vista, I'm about to finish the game.
I'm not too much of a gaming enthusiast. :)  It's my first graphics card which I can consider "up to the time". :p 
June 26, 2008 2:08:46 PM

Honestly i think it's your processor.
June 26, 2008 2:17:48 PM

Did you do a full reinstall of the OS when you put in the dual core processor?
June 26, 2008 4:49:59 PM

I dont know why he would do any of that, Crysis is optimized like summer tires on snow. crytek even said that themself, i dont know why people keep going "Check crysis (insert random stuff here)" Sure, the game requires a good comp.. but its not optimized at all
Anyways, theres nothing wrong with that score, for an AMD dualcore thats actually pretty good.
June 26, 2008 5:38:22 PM

xenios said:
Short story:
Check your BIOS version

Long story:
Had a similar problem with MSI K9A (580X based) 5200+ (@3GHz) and an X1950XT.
I had consistent results around 6K with Vista and 6300 with XP.
Then suddenly they fell to 4K.
I removed (Cleaned) Catalyst, installed previous versions, newer, Omega's etc but nothing.
I fell back to XP from Vista and from XP to Vista but nothing changed.
Then I thought that before this happened I had flashed the BIOS. I flashed an older version and problem was solved.
Of course I communicated the problem to MSI and as far as I know all further versions for my board are good.


I'm currently running the latest BIOS v1.4.
June 26, 2008 5:40:11 PM

monst0r said:
Did you do a full reinstall of the OS when you put in the dual core processor?


Absolutely!
June 26, 2008 5:42:18 PM

xx12amanxx said:
Honestly i think it's your processor.


I think you may be right, but what really bothers me is that it's a dual-core processor that has been overclocked to a nice number. The processor was released how long ago...what October of last year, right? When was 3DMark06 released? 2006. How come my dual-core processor cannot score any better than my single core processor?
June 26, 2008 6:08:06 PM

Quote:
Does seem a bit low.
Ive got my cpu at 2.8ghz right now because of the summer heat. This is what I score with my 256mb 8800GT.

3Dmark 9870
SM 2.0 4550
SM 3.0 4286
CPU 2362

Vista 64


I re-ran again after posting that first post and my scores dropped by 500. Down to around 9500 which is below your score....wtf?
June 26, 2008 6:10:44 PM



Thats just the Demo,btw
June 26, 2008 6:14:18 PM

Here is my 3D mark Q9450 @ 3.9Ghz w/ 8800GT SLI @700/1750/2000

hope it helps



I'll run one with a single 8800GT enabled for you guys give me a few minutes.
June 26, 2008 6:20:31 PM

See, that's what has me concerned. Why are 8800GT's, which are good cards, but not as good as 4850's out scoring my 4850?
a b U Graphics card
a b à CPUs
June 26, 2008 6:21:13 PM

With the gaming system in my Signature (but with a 2900XT OC'ed to 820) I scored almost exactly 10k with a CPU score around 2.1k. Now, I did a lot of experimenting with 3DMark a little while ago, and found it is much more dependant on your CPU and ram than most games are. On my 780G system, OCing the GPU from 500MHz to 900MHz had little effect on 3dMark yet a huge effect on games (as youcan see in the Tom's OC test of it). Thus, I am sure your 4850 is fine ad will be awesome in games, just don't expect to break any 3dMark records without a new processor. If I ever get my 4850 back from Newegg RMA (stupid UPS Ground) I can share my results and see how they compare.
June 26, 2008 6:27:50 PM

I had the exact same problem. your CPU would be bottlenecking it in 3Dmark06, but If you ran real games on it I doubt it wouldn't make any difference at all from your previous card. From what I observe these cards are not very good with 3dmark06, but they rock in games. how is 3dmark06 with sm2.0 and 3.0 supposed to use 800shaders? plus the fact that 2.0 relys on high GPU clocks.
June 26, 2008 6:34:19 PM

single 8800GT @ 690/1730/1898, Q9450 @ 3.9Ghz



SM 2.0 score is more dependent on CPU clock. my SM 3 score may be boosted a little bit because of the Q9450 but an AMD black isn't good for comparing graphics benchmarks you want the same CPU @ the same speed ideally.
June 26, 2008 7:07:01 PM

Y-o-u a-r-e c-p-u l-i-m-i-t-e-d
June 26, 2008 7:17:16 PM

Your all wrong. I can get higher scores than him with a 8800GT and an 65nm athlon x2.
June 26, 2008 7:22:42 PM

3dmark scores irrelevant but the data isn't. Your SM3 score is higher than mine with slower CPU which tells us there's nothing wrong with your card. Your CPU might be bottlenecking some and the new 48x0 isn't better for 3dmark2k6 but it's much better for vantage.

My scores

10906
4811 SM3

E6300 3.05ghz
2gigs at the time
8800gs 702/1728/1044 at the time
June 26, 2008 7:35:24 PM

I think if I (E8400@3.8 ghz,single 4850) got around 300 less than the guy with the Q9450@ 3.9 ghz and single 8800GT,I think I would call the impressive,considering hes got just about twice the processor.

Not to mention mine is stock speeds. My SM3.0 score is better anyway :/ 
June 26, 2008 7:36:25 PM

Id like to see that screenshot and the clock of that x2
June 26, 2008 8:31:30 PM

I'm going to give it a go on Vista Dx10 this evening. Will be in about 3 hours from now. I'm also going to increase the OC on my processor to 3.2GHz which i have reached stable easily before. I just set it back to 3.0 for the fact that I can lower the voltage.
a b U Graphics card
a b à CPUs
June 27, 2008 1:21:57 AM

Out of curiosity, run the VirtualMark "benchmark" (you may need to be logged out of the ORB). I just want to know what it "thinks" you should be getting. From some ORB searches, your CPU is getting about what it should be getting in the CPU tests. There are some strange anomalies though. Compare the pair:

http://service.futuremark.com/compare?3dm06=6377106

Similar score to you, moderately higher SM2.0 score, slightly lower SM3.0 score, slightly higher CPU score (within margin of error though). Note that he is using a 9800GTX, I couldn't find any HD4850 results, this is as close as I could come. CPU score is comparable though. Now for the second one.

http://service.futuremark.com/compare?3dm06=7012975

MUCH higher SM3.0 and SM2.0 scores, marginally higher CPU score due to the CPU speed being 3.1GHz. Your CPU is certainly operating correctly in the CPU tests, but the SM2/3 tests are being held back by something, as in the case of the first link I provided. I don't think a driver change could make that much difference.
June 27, 2008 3:11:22 AM

considering that i get 9991 max on 3dmark06 under vista with a x2 5600+ 3Ghz and 2900pro at 827/1900Mhz.(crap gigabite mb doesnt help) i would think something is off. i know the x2 5000+ be is capable of 3.1-3.2 and the 4850 smokes my card. i would think you should be atleast 11k+ with a good overclock on that cpu and gpu. now grant it 3dmark is for bragging rights only(come on we all like to brag about it) i would be more worried about real world performance. personally when i get new equipment i like to start fresh with a complete reinstall. perhaps your having a software conflict. if your cpu does have a decent overclock it shouldnt be holding your gpu back. http://www.tomshardware.com/reviews/cpu-gpu-upgrade,192... "The speed of the CPU should lie somewhere between 2600 and 3000 MHz; any lower, and the new graphics chips lose considerable performance."
anyways good luck and good gaming!
a b U Graphics card
a b à CPUs
June 27, 2008 3:33:39 AM

3DMark means nothing, but an abnormally low score could also reflect a problem that could affect "real world" performance. "Real world" performance in't always everything, some people care more for 3DMark than any game.
June 27, 2008 3:38:59 AM

No seriously, he is CPU limited, a AMD x2 dosent perform at good as a core2 clock for clock. this is why he is not at said scores.
And to the guy who says he can get over that easy with the same stuff, i urge you to tell him how you did that cause obviously you have miracle hands.
on my 8800GT + e2180 @3Ghz(stock voltage) I get 11k 3dmark 06, where my cpu is holding me back.
Youre perfectly fine on your score there as is, if you want it higher you need to OC your CPU, That is what is holding you back, nothing else.
June 27, 2008 3:41:13 AM

stabgotham said:
Yeah...I agree. Not really sure what to do here though. Any chances the card isn't getting enough power from my Corsair HX520W? The PSU has three rails, each at 18amps.


I agree with everyone who sees your problem as CPU limited. Keep in mind the benchmark sites often use Intel quads, or at least much faster Intel dual cores like Wolfdales.

My 3870x2 does fine in games, but gets "only" 9437 in 3DMark06. I've noted that benchmark sites using Intel quads get 12000 or so. Note that my CPU's an Athlon X2 4600+ @ stock 2.4, and when I do the CPU tests, it's a slideshow of 1 fps.

I haven't tried Vantage yet, even though I use Vista and DX10 modes in games, because I'm waiting till the fall when I can get a 95 watt Phenom 9850 (or at the very least an 8750) and overclock.

When you upgrade your CPU, you'll get better 3DMarks, not that it matters all that much. If you use a monitor that's at least 24", you'll avoid CPU limitations with the 4850, as it seems to be just a tad faster than my trusty 4 1/2 month old card.
June 27, 2008 5:52:35 AM

randomizer said:
Out of curiosity, run the VirtualMark "benchmark" (you may need to be logged out of the ORB). I just want to know what it "thinks" you should be getting. From some ORB searches, your CPU is getting about what it should be getting in the CPU tests. There are some strange anomalies though. Compare the pair:

http://service.futuremark.com/compare?3dm06=6377106

Similar score to you, moderately higher SM2.0 score, slightly lower SM3.0 score, slightly higher CPU score (within margin of error though). Note that he is using a 9800GTX, I couldn't find any HD4850 results, this is as close as I could come. CPU score is comparable though. Now for the second one.

http://service.futuremark.com/compare?3dm06=7012975

MUCH higher SM3.0 and SM2.0 scores, marginally higher CPU score due to the CPU speed being 3.1GHz. Your CPU is certainly operating correctly in the CPU tests, but the SM2/3 tests are being held back by something, as in the case of the first link I provided. I don't think a driver change could make that much difference.


The VirtualMark would not give me a score. It said there was only 1 other configuration similar to mine which was insufficient data.
a b U Graphics card
a b à CPUs
June 27, 2008 5:53:13 AM

I don't believe in free AA.
a b U Graphics card
a b à CPUs
June 27, 2008 5:54:02 AM

stabgotham said:
The VirtualMark would not give me a score. It said there was only 1 other configuration similar to mine which was insufficient data.

I swear it always says that no matter what :sarcastic: 
June 27, 2008 5:55:43 AM

I understand that the benchmarking sites always use the top of line CPU's, and I don't expect to get 20,000 or 30,000. I just thought with the equipment I have, I should get higher than 9K. I've run it 4 times now and cannot break 10K again. That first run must have been an anomaly.

I ran the Crysis benchmark (High Settings: 3 loops / 9 time period):

1024x768/No AA - AVG. FPS=41.56

1024x768/4x AA - AVG. FPS=37.2

1280x1024/No AA - AVG. FPS=40.805

1280x1024/4x AA - AVG. FPS=31.65

My FPS took a major hit with 4x AA at 1280x1024 which is surprising since everything we have read about the 4800 series is that you practically get 4x AA for free. Apparently, I do not.
June 27, 2008 6:03:21 AM

From Futuremark's product description:

"VirtualMark is based on the Futuremark industry standard benchmark, 3DMark. With our vast database of over 15 million results, VirtualMark is able to accurately predict the performance of almost any imaginable computer configuration."

Wow...apparently not Futuremark...
a b U Graphics card
a b à CPUs
June 27, 2008 6:04:34 AM

Well you are using a new card. Give it a month or so and there should be a few hundred to compare to.
June 27, 2008 6:07:57 AM

Looking at Anandtech's review of the 4850 and the Crysis benchmark specifically, it appears I'm actually getting better FPS then they are at 1280x1024. Maybe it's all in my head....

I wonder if I should pick up a second for CF....
June 27, 2008 6:29:07 AM

Anand is worse than toms so that is to be expected.
a b U Graphics card
a b à CPUs
June 27, 2008 6:34:04 AM

I won't take sides, but you will find that comment quite controversial. :lol: 
!