3000 3DMarks with Radeon 9000 P4 1.6GHz HELP!

[toolfan]

I just installed a Sapphire Radeon 9000 128MB card, replacing my ATI AIW 7500. On the 7500 I was getting about 4500 in 3DMark. With the 9000 I can't get above 3000. This is with all default settings, no AA or anything. I have no idea what is wrong, I went so far as to format and reinstall WinXP but that didn't help. I've tried both the Catalyst 2.3 drivers and the ones that were on the Sapphire CD. Before the reinstall of Windows I tried installing both the Intel INF utility and the Application Accelerator, to no avail. Could it possibly be a bad card or something hardware related? This is on a Epox 4G4A mboard (latest BIOS), P4 1.6A, 512MB PC2100, WinXP. Thanks in advance for any input.

 

[Deeko]

*sigh*

Have you played any games yet? Run any REAL benchmarks? If you do, and they run poorly, then come back here for help. If they run well, don't freakin worry about it.

 

[toolfan]

The only game I've tried with the new card so far is UT2k3. It was pretty choppy during gameplay before, and is no better now. Deeko, do you not think it's abnormal to be posting lower 3DMark scores than an AIW 7500? Or to have scores about half that of other systems with the same CPU and a Radeon 9000? This isn't a post looking for tweaks so I can be a l33t d00d with high 3DMark scores, but using it as a general benchmark it looks like something is not working properly (and I expected UT2k3 to be a bit more playable with the upgrade - which it is not). Thanks anyway...

 

[Deeko]

Do I think its odd for a Radeon 9000 to be running at 7500 AIW speeds? Yes. Getting equivalent 3DMark scores? No. Its a waste, don't worry about where your score is. Why not run benchmark.exe in UT2K3? I would think you probably spend more time playing that then running 3DMark. (unless you enjoy watching it all day long).

 

[tapir]

What's the sense in telling him not to use benchmarks and then telling him to use a different program as a benchmark? Obviously real-world framerates are more important to most of us than benchmarks, but shouldn't UT2K3 produce similar results? And if theres a problem in one program, there'll probably be problems in others..

I dunno, it seems reasonable to me to use any program, and if possible every program as a reference point. What's so bad about 3DMARK2001SE?

Sorry to get off topic here... maybe you could try AGP drivers and such? Check with radeonator or something to see what speed your card clocks are. Also if the 9000 Pro is an 8X part (I thought it was) maybe your AGP bus is getting bugged down.

 

[Deeko]

Because 3DMark isn't a good benchmark.

I was going to suggest AGP drivers as well, if you ran a good benchmark lol

 

[rogue1979]

If that is a northwood P4 why are you not putting it on a 133MHz fsb for 2133Mhz? Use an overclocking utility and put the 9000 up to an easy 285/235 for starters.

 

[toolfan]

rogue1979, yes it is a Northwood P4 and it is normally clocked at 2.13. I brought it down to 1.6 to take out all possible variables while testing this. It is back at 2.13 now, but scores still seem pretty low...

Ran the Antalus flyby, got about 23 when CPU clocked at 1.6. Upped it to 2.13G, again got a score of about 23. UT2k3 settings are 1024x768x32, high details, normal physics. This is with the default clock speed of the Radeon at 250/200. I may overclock that down the road, but it should be doing better even at stock speeds.

As far as the AGP drivers go, as far as I know the only ones available are the Intel INF ones I have installed. If I'm forgetting something please let me know.

I have a feeling I'm missing something stupid, so if anyone has any suggestions at all let me have 'em. Thanks.

 

[88637]

250/200 you say? That sounds like the speed of just the regular 9000 not the pro. I didnt think they made a 128 meg regular 9000 though...i never really found many benchmarks for the non pro ver though so I dont know how fast it is in comparison to the pro version but it is a chunk slower (pro is to be rated at 275/275 to my knowledge)

 

[toolfan]

It is a regular 9000, not a Pro - sorry I didn't make that clear before. Got the 128MB version from Newegg. The regular 9000 should definitely be slower than the Pro, but not slower than a AIW 7500 right? Unfortunately I don't have any benches in UT2k3 from the 7500 to compare to, but it definitely doesn't play any better.

 

[88637]

It is a regular 9000, not a Pro - sorry I didn't make that clear before. Got the 128MB version from Newegg. The regular 9000 should definitely be slower than the Pro, but not slower than a AIW 7500 right? Unfortunately I don't have any benches in UT2k3 from the 7500 to compare to, but it definitely doesn't play any better.

Maybe, maybe not...
http://www.anandtech.com/showdoc.html?i=1655&p=5

9500 pro gets 45.9 fps, 7500 gets 39.1 at 1024x768
1280x1024 its 29.4 to 25.5
1600x1200 its 20.5 to 17.9

if you take 15% or so for the clockspeed diff then who knows. Although they are just testing on a 64 meg pro in that too but still. That mem clock is probably killing it though.

 

[toolfan]

Yeah, it might be pretty close if taking off 15-20% for the clock difference. But I'm scoring 23fps right now at 1024x768. That's on a 2.13 as compared to the 2.4 in that benchmark, but that CPU difference shouldn't account for the 40% difference between my 9000 and the 7500 in that test.

 

[88637]

This is true, although to be honest I dont trust benchmarks on these websites. If I were you, I would check against your own 7500 since we dont know precisely every setting they use, etc.

 

[toolfan]

I tried OC'ing the card and it didn't go over too well. Upped the memory to 235, and the core to 275. That accounted for a 2fps increase and a wealth of texture corruption. Played around with different speeds but the performance increase was negligible. I may have to put the 7500 back in and bench that in UT2k3. But that will have to wait until tommorrow, as I've had enough of reinstalling drivers and redoing my custom res'es in Powerstrip for one day. Fortunately this isn't my main gaming rig (that one hits 98 in Antalus), so I won't lose too much sleep tonight. Still would be nice to know if the performance of the regular 9000 is supposed to be like this...

 

[88637]

Well if it isnt vid drivers, agp drivers, a conflict of some kind perhaps, only other thing I can think of at this time of night is maybe a bios flash of the vid card (its on my mind since i just flashed mine this afternoon...) maybe it shipped with some old crap bios or something. Other than that, maybe it is just a pos...

 

[rogue1979]

I have three Radeon 9000's running in the house. One of them replaced a Geforce2 Pro @240/475 which is getting close to Radeon 7500 territory. The Radeon 9000 is much faster in everything.

 

[Mingon]

my pci radeon beats that 5331 , I wouldnt agree with people who say 3dmark is useless - its a good tool for telling you when something is not quite right such as with the problems your gettting. Each time I have think my machine has a perf problem I run 3dmark as its a simple way to check how things are.

 

[AnAndAustin]

My goodness guys the Rad9000 sucks eggs, it's 20% slower than the Rad9000PRO and if you put aside DX8 hw and a couple of slight enhancements it is no better than a Rad7500. Even the full Rad9000PRO is still a good 15%+ slower than the Rad8500LE, sure the Rad9000 series are better buys than the GF4MX but while Rad8500's are still around they are by far the best cards for under $100.

toolfan, if only you'd checked before buying a Rad9000 you'd have seen it is NOT an upgrade to a Rad7500, even the Rad9000PRO isn't a huge leap. Anyway here's a few articles.

Tech Report Rad9000PRO review (to see how it fairs against the other cards)

Digit Life Rad9000 (non-PRO) (p)review (Rad7500, 8500LE-64, Rad8500LE-128, GF4MX440 and GF4MX460)

AnAnd CPU Scaling article 1 - many games and gfx cards

AnAnd CPU Scaling 2 article 2 - many gfx cards but focusing on UT2003

 

[Mingon]

My goodness guys the Rad9000 sucks eggs

Bit harsh - most people still use 1024 x 768 as their main desktop size - the 9000 is more than capable of running all games at this resolution. To say it sucks just shows youve never used one. For the price their is no comparison, yes the 8500 might be better but unless your running with AA & AF running the 9000 will do just fine.

 

[toolfan]

Thanks for the replies everyone. Instead of debating whether or not it would have been better to go with a 8500LE instead of the 9000, I am more concerned about the drop in performance from the AIW 7500. Getting 1500 less in 3DMark and about 15fps less in UT2k3 Antalus flyby doesn't seem right. Anyone have any other suggestions before I RMA this thing and try something else?

 

[AnAndAustin]

I didn't mean to sound harsh, I dislike nVidia's GF4MX naming even more than ATI's Rad9000 naming. Anyway lets consider the prices:

GF4MX440 $59 (faster in pretty much anything than a Rad9000 but only DX7)
GF3TI200 $69 (faster than Rad9000PRO and also sports DX8 too)
GF4MX460 $93 (faster than MX440, still DX7)
GF4TI4200 $111 (much faster than GF3TI200 and of course DX8)

Rad7500 $52 (about the same as Rad9000 but only DX7)
Rad9000 $65 (HERE IT IS, at least it does have DX8)
Rad9000PRO $82 (20% faster than Rad9000, DX8)
Rad8500LE $80 (15%+ faster than Rad9000PRO, DX8)
Rad8500 $81 (10% faster than Rad8500LE and comparable to GF3TI500, DX8)

Then on to perf:

DIGIT LIFE: A P4 2.2ghz and Rambus can only manage 32-34FPS at 1024x768 on Rad9000 and that's without AA or AF on in both UT2003 and Commanche4. 60FPS is what most people consider to be playable.

Now bearing in mind a Rad8500LE is a good 15% faster than the Rad9000PRO which in turn is 20% faster than the Rad9000 ...

ANAND CPU1 (RAD8500LE which is significantly faster than Rad9000):
UT2003 build918 1024x768 800mhz=19FPS XP2100+=20FPS
SerSam2 1024x768 800mhz=44FPS XP2100+=51FPS
JK2 JO 1024x768 800mhz=60FPS XP2100+=100FPS
Commanche4 1024x768 800mhz=18FPS XP2100+=35FPS
Q3A 1024x768 800mhz=150FPS XP2100+=210FPS
RTCW 1024x768 800mhz=97FPS XP2100+=150FPS

 

[Mingon]

try booting into safemode then uninstalling the card, restart, then reinstall with newest drivers. Perhaps some registry settings are different between the cards as the have quite different architectures

 

[toolfan]

try booting into safemode then uninstalling the card, restart, then reinstall with newest drivers.

I've formatted and reinstalled Windows since swapping cards - no change. I think I'm about convinced to return this card and get an 8500 instead.

 

[fluxquantum]

AnAndAustin, nice job. thanks for the info.

 

[AnAndAustin]

I wouldn't play around with it. I've seen a wealth of info which all agree that the Rad9000 is equivilent to a Rad7500 (ie is the same speed) but has the added advantage of DX8 hw and slightly improved image quality, AA, AF etc. Rad8500LE-128MB is the least you should consider and has hugely better perf than the Rad9000 and Rad7500. Do check the clocks and dual display if that's important, clocks should be about 250/500 (full 8500 275/550 o/c usually possible) as they do vary by manu and also retail vs oem/bulk. A Rad8500LE-128MB @ 250/500 is preferable to a Rad8500-64MB @ 275/550.