Radeon HD 4830 3DMark06 scores are low...

[alexruiz]

I am not a big gamer, but anyways I went ahead and gave myself an upgrade to a Sapphire Radeon HD4830. I am very very pleased with the card, it is potent during actual gameplay, the fan is decently quiet at full load AND it throttles down in idle or low load.

Just for the sake of curiosity, I ran the standard 3dmark06 test. I was surprised at the result as it wasn't that much higher than the Gigabyte Radeon HD4670 it replaced. Both in the same machine:

Gigabyte GA-MA780GM-S2H motherboard (ATI 780g, SB700, running in native AHCI)
AMD Athlon 64 X2 5000+ BE @ 2.8GHz
2 GB Corsair DDR2-800

The integrated video scores ~1300 in the 3dmark06 default test. A Radeon HD4650 scores ~5100, while the HD4670 scores ~7800. These scores are well in line with the averages I have seen on the web, even considering the web benchmarks are run on C2Quads.
However, the 4830 scores ~9400, well short of the ~11000 I have seen for the 4830. Is my CPU bottlenecking the videocard? Was the 4670 a better match for my 5000+?

Curiously, in game the videocard performs like a champ. The game we play (Universe at War: Earth Assault by Petroglyph, uses Alamo engine) is very very demanding in the graphics, but the game engine is also very flexible. You can play with integrated GeForce 6100 @ 800 x 600 with everything down, and go all the way up from there. The 780g allowed some medium details at 800 x 600. The 4650 allowed native resolution with most 50 / 50 of settings med / high. The 4670 allowed everything high with only 2 settings in med. the 4830 allows everthing in hign and only one setting as med-high. Reading from other players, that is normal as even GTX260 with core i7 struggle with eveything at max.

So, just as curiosity, why the low score of the videocard in 3dmark06?

Thanks



Alex

 

[chizow]

Originally posted by: alexruiz
Is my CPU bottlenecking the videocard? Was the 4670 a better match for my 5000+?

Yes, 3DMark06 is heavily CPU bottlenecked to the point there's very little difference in score beyond a G80/R600 class GPU, only major differences from CPU clockspeeds and number of cores. I wouldn't say the 4670 is a better match for the 5000+ unless you were happy with that level of performance along with your comments about noticeable improvements in games.

 

[alcoholbob]

Only thing that should matter is gameplay. Looking at synthetic benchmark numbers can be misleading because they often stress one part of hardware more than the other (or render something obscure which some hardware has an advantage with...and that most games don't use).

 

[VirtualLarry]

Originally posted by: alexruiz
However, the 4830 scores ~9400, well short of the ~11000 I have seen for the 4830. Is my CPU bottlenecking the videocard? Was the 4670 a better match for my 5000+?

I think that you are wrong, and the 3dmark06 score is right.

My E2140 @ 3.2Ghz, with a VisionTek 4850 at stock speeds, scores 11,323. I don't think that a 4830 is equal to a 4850, as much as you might like that to be true.

 

[FalseChristian]

That's a good score with that underpowered CPU you have. The 4830 is about equal to one of my 8800GT's.:thumbsup:

 

[cusideabelincoln]

The Athlon X2 is a good match for the HD4830. I have an HD3850 (similar to the HD4670) and wish I had more GPU power for games like Far Cry 2. Of course in other games I do miss more CPU power, such as in TF2, but it still plays perfectly fine.

Oh and I am able to get up to 9900 in 3DMark06, so I'd say the CPU is holding back your score because of those CPU-only tests. What are the framerates that you get in the actual game tests? I'll post mine in a minute...

Graphics Tests
1 - Return to Proxycon 34.8 FPS
2 - Firefly Forest 38.68 FPS
CPU Tests
CPU1 - Red Valley 0.69 FPS
CPU2 - Red Valley 1.12 FPS
HDR Tests
1 - Canyon Flight (SM 3.0) 43.82 FPS
2 - Deep Freeze (SM 3.0) 52.17 FPS

 

[alexruiz]

Thanks for the replies guys.

Either we are doing something wrong, or I want something of what the reviewers are smoking

A few sites showed 3dmark06 scores of the 4830 in the neighborhood of 11k:
ExtremeTech
LegitReviews
HotHardware
catch: all of them used C2Q QX9700 or higher

xbitlabs has a score of ~10,300 with a C2D, so me might be into something

I won't worry anymore, as I said, in game it performs like a champ.

 

[legcramp]

Core 2 Duo smokes the X2 series clock for clock, so your cpu @ 2.8ghz is probably equal to a stock E6600.

 

[VirtualLarry]

Originally posted by: alexruiz
A few sites showed 3dmark06 scores of the 4830 in the neighborhood of 11k:
ExtremeTech
LegitReviews
HotHardware
catch: all of them used C2Q QX9700 or higher

xbitlabs has a score of ~10,300 with a C2D, so me might be into something

Now that makes sense. I have yet to test a 4850 with my Q6600 at 3.6.

 

[Andrew1990]

Originally posted by: RIFLEMAN007
Core 2 Duo smokes the X2 series clock for clock, so your cpu @ 2.8ghz is probably equal to a stock E6600.

Actually a little less. The X2 6000+ Windsor is on par with the E6600 stock so comparing it to something such as an E6400 would be closer.

Anyway, even though the Core2s are faster, the X2s are still pretty fast.

 

[speedfreak]

it seems that there is a bottleneck. the cpu is holding back the true potential of the gpu. try overclocking it a bit more. =)

 

[crazylegs]

Just ran 3DMark 06 on my system for a test got 11,330 - thats with my E8400 running 4.01Ghz@446x9 and 9600GT with a mild factory OC.

Think your scores are fine for you CPU. Enjoy ur 4830 heard lots of good things about it. (like others said its gameplay that actually counts)

 

[AzN]

3dmark2k6 is heavily dependent on CPU. I remember breaking over 7k with a 8600gts because I overclocked my CPU to 3.5ghz. In actual games it performed exactly the same whether I was running 1.86ghz or 3.5ghz.

 

[Leyawiin]

Its your CPU. When I built my PC a few years ago it had an X2 5200+ (stock at 2.6 Ghz) with an 8800 GTX (which is maybe a little stronger than a stock HD 4830). My 3dmark06 score was around 9900. Since then I've upgraded the CPU to an X2 6400+. Stands at 11,430 now. Still, if you are happy with your gaming experience don't sweat those synthetic benchmarks. They're pretty meaningless really (I only use them to gauge if my PC is running reasonably well).

 

[cusideabelincoln]

I'd still like to know what framerate he (and everyone else) gets in the various tests, since that is more important to me than the final score.

 

[kmmatney]

I'll run it on my system to see what I get with my HD4830. I'm running the automatic CCC overclock of 690 Mhz core, and 2140 Mhz memory (it can go higher, but this seems OK for now). I'm downloading 3D mark right now, will take a while.

 

[soccerballtux]

Originally posted by: VirtualLarry

Originally posted by: alexruiz
However, the 4830 scores ~9400, well short of the ~11000 I have seen for the 4830. Is my CPU bottlenecking the videocard? Was the 4670 a better match for my 5000+?

I think that you are wrong, and the 3dmark06 score is right.

My E2140 @ 3.2Ghz, with a VisionTek 4850 at stock speeds, scores 11,323. I don't think that a 4830 is equal to a 4850, as much as you might like that to be true.

I can second this; OP your scores sound fine for your rig.

 

[imported_Asianman]

Its fine, On a phenom X4, with a HIS 4830, I get 10400, so a 1000 point increase from 2 to 4 cores seems about right. The C2Qs are much faster in most synthetic benchmarks so you have nothing to worry about. And a Athlon is fine for the 4830.

 

[kmmatney]

On my system, I get 10703 with my overclocked HD4830. I would probably get 11000 with a better cpu (and my GPU overlock was the default CCC overclock - could go higher).