ATI HD4890 XFire Performance Review

SonDa5 said:
IMO If I don't see a 30,000 3dMark06 score from 2 HD4890s within the next 30 days from now I'll consider the HD4890 as a product that wasn't needed.
Click to expand...

You do realize that 3dmark06 is EXTREMELY cpu dependent. It's a very flawed benchmark for stictly measuring the gaming prowess of GPUs. Try your own cards with a dual core chip at stock. You'll lose thousands of points. Does that make your 4850's any less potent?

Speaking of which...

I'm still waiting to see your screen shot of your 24K 06 score on your 4850's. Even if you did score that high, (which I have my serious doubts) you have to have a ridiculous OC on your 4850 to achieve that. Show us your 3dmark score at bone stock for you 4850's. I doubt you even break 20k.

Point is, you can't compare systems with 20-30% OCs on the GPUs to benchmarks with systems running at bone stock settings. You know what, I bet crossfired 4890's running at 1100 mhz on the core and 4800mhz on the memory coupled with a 5ghz i7 chip would easily give you the kind of numbers you're "expecting" with the 4890's. That would be a much more appropriate comparison.

Either way your statement is pretty silly because in the end we don't "play" 3dmark. Real game performance is all that matters.
 
Johnny78 said:
You do realize that 3dmark06 is EXTREMELY cpu dependent. It's a very flawed benchmark for stictly measuring the gaming prowess of GPUs. Try your own cards with a dual core chip at stock. You'll lose thousands of points. Does that make your 4850's any less potent?

Speaking of which...
Click to expand...

I know how 3dMark06 works and I expect to see 2 HD4890s crossfire with at least 30,000 points. It should be able to do this. There are people who know how to tune the 4800s GPUS.



I said "just under 24,000".

I hope to break 24,000 soon.

Also I don't have any bios or voltage mods on my HD4850s. I have a heat problem with my Toxic HD4850 that I'm working on. Once I fix it i should be able to break 24,000.


I'll post some "e-pen" screen shots ect. of my best 3dMark06 if i can find the results.
 
My 3dMark06 score with HD4850 crossfire at stock and core 2 duo @3.22 gave me a little over 15K. I believe 3dMark05 is more GPU dependent. What do you get there SonDa5?
 
F.E.A.R. said:
My 3dMark06 score with HD4850 crossfire at stock and core 2 duo @3.22 gave me a little over 15K. I believe 3dMark05 is more GPU dependent. What do you get there SonDa5?
Click to expand...

I've never tried 3dMark05.

After I get my Toxic HD4850 heat problem fixed I'll try it out.
 
F.E.A.R. said:
My 3dMark06 score with HD4850 crossfire at stock and core 2 duo @3.22 gave me a little over 15K. I believe 3dMark05 is more GPU dependent. What do you get there SonDa5?
Click to expand...


I downloaded it and it didn't like the 9.4 drivers.

I had to tweak a few things to get it to work.

With my [email protected] GHZ I scored almost 22,000 with my Xfire Hd4850 setup.

If I put my [email protected] will 3dMark05 score increase?

With some work I think I could score 24,000 on 3dMark05 as well.
 
SonDa5 said:
I downloaded it and it didn't like the 9.4 drivers.

I had to tweak a few things to get it to work.

With my [email protected] GHZ I scored almost 22,000 with my Xfire Hd4850 setup.

If I put my [email protected] will 3dMark05 score increase?

With some work I think I could score 24,000 on 3dMark05 as well.
Click to expand...

yes but probably would only add 1000 points tops in 3dmark05 and about 2500-3000 points in 3dmark06.. but ya never know..

also 3dmark06 is a load of crap as a benchmark.. its so friggin cpu dependent.. about the only one that ive seen even mildly realistic is 3dmark vantage..
 
sirmonkey1985 said:
yes but probably would only add 1000 points tops in 3dmark05 and about 2500-3000 points in 3dmark06.. but ya never know..

also 3dmark06 is a load of crap as a benchmark.. its so friggin cpu dependent.. about the only one that ive seen even mildly realistic is 3dmark vantage..
Click to expand...


With the same settings that I ran the 3dmark05 test 3dMark06 scoreda little over 22,000 on my machine.

My experience with HD4850 and CPU speed has been that performance in 3dMark06 greatly increases when increasing CPU speed.
 
Johnny78 said:
24K with 2 4850s in regular crossfire not quadfire? Wow, sounds awful high. Are you sure you're running 06 and not 05?

Not saying I don't believe you but I'd like to see a screen shot of that. You must be running your CPU close to 5ghz.
Click to expand...



I found my best score info.

23,462
2 HD4850s in Xfire with Q9550 CPU.

almostthere3dmark06Q9550.jpg


Link to ORB validation with more details of build(same as in my sig):

http://service.futuremark.com/compare?3dm06=10427141

GPU-Z Validation
http://www.techpowerup.com/gpuz/a8cm6/

CPU-Z Validation
http://valid.canardpc.com/show_oc.php?id=533487


i think I can break 24,000. Just need to get stuff a little cooler.
 
I'm going to Xfire a Diamond OC Hd4890 with a VisionTek HD4850.

I'll have some 3dMark06 scores up soon.

Performance should be much better than my HD4850 Xfire set up.

If i don't post scores it's because my PSU (Corsair TX650) isn't up for the job.
 
Last edited:
New Xfire config for my RIG.


HD4850xFireHD4890small.JPG

Diamond OC edition HD4890 with VisionTek HD4850.

Not sure if 2 HD4890s would fit in Xfire.

Still running Corsair TX650W PSU. Hope it holds up.
 
kadozer said:
Beautiful, 4890x2 is going to be a beast. 2gb usable and at least 1k clocked..Bye bye gtx295. p
Click to expand...

I think it's very foolish for AMD to waste the time of their engineers and scientists on a 4890x2 when DX11 is right around the corner...

But they are doing it.
 
Dozoy said:
I think it's very foolish for AMD to waste the time of their engineers and scientists on a 4890x2 when DX11 is right around the corner...

But they are doing it.
Click to expand...


I think 4890 is compatible with DX11.
 
4890 is NOT Direc3D 11 compatible. It's Direct3D 10.1 compliant but not 11.
I agree with Dozoy though, it's a waste to design a 4890X2 when they could be put to work making the 5xx0 series better.
 
I was under the impression a Hd4800 card would be able to become DirectX11 compliant by means of a driver update.



Here is a shot at my HD4890+HD4850 Xfire set up.


XfiretempsHD4850withHD4890.JPG
 
SonDa5 said:
I was under the impression a Hd4800 card would be able to become DirectX11 compliant by means of a driver update.
Click to expand...
What are you smoking son? If the 48xx cards were able to be updated to Dx11 nobody would buy an Nvidia card, NOBODY. Buying a first gen Dx11 card is foolish. Remember the FX 5xxx that Nvidia produced as first gen Dx9 cards? The 6 series came out less than 18 months later and completely wiped the floor with the FX 5 series.

By the way, why did you purchase a 4890 to match up with a 4850? You do know that the higher card will scale to the lower card right? Going to buy another 4890 down the road?
 
Dozoy said:
By the way, why did you purchase a 4890 to match up with a 4850? You do know that the higher card will scale to the lower card right? Going to buy another 4890 down the road?
Click to expand...

I got a good deal on one. If it doesn't work out I'll go back to my HD4850 Xfire setup which does better on at least 3dMark06.

HD4890 by itself rocks for 3dmark06. Scored over 20,000 with single HD4890 in my RIG.

Add a HD4850 in Xfire and everything goes south. :(

terriblematchHD4890andHD4850.JPG



Time to play some UT3. Hope that works out better.
 
Both my HD4850 Toxic and VisionTek HD4850 are 512mb models.



The HD4890 with my Hd4850 work great for game play.
Playing UT3 right now. Awesome 3d Eye candy and performance. :)
 
SonDa5 said:
Both my HD4850 Toxic and VisionTek HD4850 are 512mb models.



The HD4890 with my Hd4850 work great for game play.
Playing UT3 right now. Awesome 3d Eye candy and performance. :)
Click to expand...
I cant understand why you would stick a faster 4890 1gb with a 4850 512mb. now you only have 512mb to use plus it defaults to the slower card. whats the point in getting a 4890 to do that with it??
 
cannondale06 said:
I cant understand why you would stick a faster 4890 1gb with a 4850 512mb. now you only have 512mb to use plus it defaults to the slower card. whats the point in getting a 4890 to do that with it??
Click to expand...


Both of them in crossfire are more powerful than a single HD4890.
Still not the best match. HD4890 with HD4890 should be alot better. Even still my RIG is scoring close to 23,000 in 3dMark06. Almost as much as both of my HD4850s in Xfire.

The 3D settings play better with the HD4890. Despite the 3dMark06 score being lower.

My MB DFI P45 JR T2RS mATX would suffocate 2 HD4890s in Crossfire.
So going 2xHD4890 really isn't an option for my system. Not only that but I'm not sure if my Corsair TX650W has got enough juice to
power 2 HD4890s in crossfire.
 
My scores are as follows:

3dMark Vantage (Extreme Profile): 9920
3dMark 06: 25860

My cards are clocked at 950/1100.

I recently bought two GTX 275s to test them in SLI.

3dMark Vantage (Extreme Profile): 11420
3dMark 06: 26980

Kept the ATI cards because they absolutely kicked the GTX a$$ in all of my games. Synthetic benchmarks are worthless in my opinion...
 
Vantage uses PhysX so Nvidia cards almost always walk away with the "win." 3Dmark06 is extremely CPU dependent so it's not much use either. All in all, 3DMark is worthless. That's not news.
 
silent-circuit said:
Vantage uses PhysX so Nvidia cards almost always walk away with the "win." 3Dmark06 is extremely CPU dependent so it's not much use either. All in all, 3DMark is worthless. That's not news.
Click to expand...


Don't agree.

It's only a part of the peformance measurement. Not all. It is something though.
 
You must log in or register to reply here.
Back
Top