Sign in with
Sign up | Sign in
Your question

low 1900xtx 3dmark05 results on new system

Last response: in Graphics & Displays
Share
March 20, 2006 10:32:50 PM

hi guys,

I just built a new box. I used to be really into pc's a few years ago but spent a couple of years out of it so am a little rusty.

athlon 4800+ (not overclocked)
4 x 1gb oc-z
2 x 300gb seagate sata2 drives (not in a raid config)
dfi rdx-200 motherboard
gecube ati 1900xtx
win xp sp2 32bit edition.

It only benchmarked 8943 3dmarks in 3dmark05 and i see other guys with similar specs getting 11k+

In some games it runs well - i.e. call of duty 2 I can play 1920x1200 with 4x AA and it's smooth as butter. In other games like battle for midle earth II it lags a little but I'm not sure whether to blame the video card, or my hard drives which seem a bit slow for some reason.

with the hard drives, I ran the sisoft sandra (2005 SR3) filesystem benchmark and it did 41mb/s I don't know if thats good or not. i doubt it is.

running latest ati drivers.

any tips would be greatly appreciated. cheers
March 21, 2006 1:55:48 PM

The only thing I can think of in terms of slow hard drive performance is that fact that you've got 2 monster drives that aren't raided, as opposed to something like a 74GB Raptor at 10K RPM.

I'd suggest RAIDing those 2 or getting a raptor for your applications/games, and using the 300GB drives for storage.

-Ryu
March 21, 2006 2:25:25 PM

It seems to be a common problem - my 3dmark05 is also in the 8000's - same for some other people. Dunno why though, but the rest of my games are on par...
Related resources
March 21, 2006 2:29:04 PM

hmm yes, those scores are lowish. Try run 3dmark06

i have a x1900xt with amd 4400+ and 2gigs of RAM and i score about 10k on 3dmark05.

Both my hard drives are rubbish and are not raided btw.
March 21, 2006 2:39:05 PM

Low yes, but numerous people have reported lowish scores with decent setups on 05 and I'd really like to know why. Especially a person with 4800+ CPU should be hitting 10k regardless of HDD or RAM (bar under 1gig) - my 06 scores are quite good though....
March 24, 2006 2:03:08 PM

I'm hitting 8K with stock x1900xt and a 3.6GHz P4, 2GB DDR2.

I also have programs running while testing.

-Ryu
March 26, 2006 1:15:28 AM

I get not as great of marks as others, but excellent as far as I'm concerned with my ati x1900xtx. there are other issues that bottleneck these tests.

your card is really "beauty in the eye of the beholder". mine is great for my needs and am not fussing over the benchmarks. it is the games and apps that matter for me.

3dmark5 - 9,175 (avg)
3dmark6 - 6,062 (avg)
March 26, 2006 8:33:23 PM

well anyway moving on, I got rid of my DFI motherboard and bought the new asus motherboard, the A8R32MVP. I was reading on some forums and saw similar complaints with the DFI (though i know it is a good overclocker) so wanted to swap it out, and in addition the DFI had really poor USB2 performance which annoyed me (lots of external drives).

Anyway, just doing that m/b swap, and changing nothing else and I'm now getting 11264 3dmarks in 3dmark05.
March 30, 2006 12:04:10 AM

It's the FPS during gameplay that really matters. As long as you get good and consistant FPS without lags, is good enough.

My system hit the 9k mark. I just can't break the 10K mark even if I oc to 4.11GHz. :( 

-----------------------------------
3.8 GHz P4, x1900xtx,
1GB DDR2, 160gig SATA2,
XI-FI Fatal1ty sound card
March 30, 2006 9:05:41 PM

Have you tried taking 2 of the 1gb sticks out and just running 2gb? Just a thought. You might have problems running the full 4 gigs. If it runs better...i will take those 2 sticks off your hands :D 
March 31, 2006 9:43:51 PM

Quote:
I don't see the point of 4gb RAM...


Especially with a 32-bit OS. I know this might seem crazy, but try taking out 2GB, and then see your results.
March 31, 2006 10:50:50 PM

The best comparison is game for game.

Two thirds of the 3DMark scores are extreme overclocks, that yes, while they do 'pass' the test and generate a score... the test was just garbled 3D images, and hypercolour textures, with registry hacks to lower texture (and possibly 'new' FSAA pattern) quality so low the whole polygon is just one blur of colour.

Until they build an artifact tester into 3DMark, it will be less useful for comparisons using the Online Result Browser.

Even now websites only use it for 'system to system' comparisons and do not use the results in ORB.

Is the system slow in the games you play ?, or do people these days have a 3DMark 05/06 server they join so they can sit around 'benchmarking' [cough Futuremark cough] while not engaging in tactical combat ? Razz

It is for these reasons Futuremark has shamed ORB, while 'MadOnion' tried to make it something useful.

Razz Also you forgot to set minimum texture quality, force no FSAA, minimum mipmap quality, etc in drivers..... this is how people get +15% higher on their 3DMark scores, then overclock another +25% or so (until you get massive artifacts but the test just passes) to get near +44% higher scores. (1.15 x 1.25 = 1.4375). Razz - That is how 3DMark ORB was left in the state it is in now. No Artifact testing enforced during testing makes it useless for comparisons. Seriously.... Rolling Eyes
March 31, 2006 11:35:48 PM

Just my opinion:

I wouldn't blame a company for not getting good marks on a benchmarks. Hey, if I scored 15 on '06 and could play F.E.A.R. at all settings high with 60+ FPS, I'd be fine.

What is it, I think Windows uses only 2.75GB of RAM? I would actually call that the suspect, try 2GB...Or reducing page file, tho that shouldn't be the problem.

~Ibrahim~

P.S.10k+ in '05 would not be bad, either, lol.
April 1, 2006 1:09:06 AM

The PCI devices use parts of the 32 bit address space, between 512 MB and 1280 MB, leaving only 2.75 GB to 3.5 GB 'free' to have physical memory mapped to them by Win32 Kernel.

On the same hardware, EM64T or AMD64 req, running Win x64 Kernel 'fixes' that, as it has 2^40 bit address space to map physical memory too.

eg: (Assuming you know hex):



You can see I am down 1280 MB (B000,0000:0000 to FFFF,FFFF:FFFF) there right away, and that entire range can't have memory mapped to it. The range is still in use under Win x64 kernel, however it can just map memory in other free parts of the 40 bit 'physical' address space, and 48 bit 'virtual' address space. All memory applications request is virtual anyway, and the OS kernel pages when required while trying to keep disk cache large.

Windows 'Server' editions have PAE (2^36 bit addressing) enabled in the Win32 (really Win32+PAE36) Kernel, so the 'server' editions of '32 bit' [cough 36 bit] Windows OS(es) can map the memory aswell.

The 'problem' only affects 32-bit kernels in workstation versions of Windows, such as: NT Workstation, 2K Pro, XP Pro/Home, etc.

I hope that made some sense. :p 
April 1, 2006 1:55:06 AM

Quote:
hi guys,

I just built a new box. I used to be really into pc's a few years ago but spent a couple of years out of it so am a little rusty.

athlon 4800+ (not overclocked)
4 x 1gb oc-z
2 x 300gb seagate sata2 drives (not in a raid config)
dfi rdx-200 motherboard
gecube ati 1900xtx
win xp sp2 32bit edition.

It only benchmarked 8943 3dmarks in 3dmark05 and i see other guys with similar specs getting 11k+

In some games it runs well - i.e. call of duty 2 I can play 1920x1200 with 4x AA and it's smooth as butter. In other games like battle for midle earth II it lags a little but I'm not sure whether to blame the video card, or my hard drives which seem a bit slow for some reason.

with the hard drives, I ran the sisoft sandra (2005 SR3) filesystem benchmark and it did 41mb/s I don't know if thats good or not. i doubt it is.

running latest ati drivers.

any tips would be greatly appreciated. cheers


Did you turn off all the eye candy? You need to do this to see where you compare with others (of course, to see if there are any hardware/software problems). The thing is, everyone cheats. Its sad but that's what everyone does so you cant really change the system.

Funny cos just minutes ago I ran a loop of 3dmark05. Got 12,092 with my Dual Core Opteron 170 @ 2.71 Ghz and x1900xtx stock
April 1, 2006 2:20:54 AM

Seems very low as i get on a XFX 7900GT@520 and an opty 165 8900. I wonder if it isnt the ram that might be the problem also as windows has problems addressing it right and that its in T2 over T1.
April 1, 2006 3:15:08 AM

I agree about the RAM, 4gb is unnecessary and may actually give performance loss. But something new I have to bring to the table that I noticed when I set all of the image quality setting to highest in catalyst control center with my X800pro, I lose a couple thousand 3dmarks. But when I have it all set on auto and "let the application decide", I get better marks. So you may want to look into all of that, and to give you an idea here is my score with settings on auto:
http://service.futuremark.com/orb/projectdetails.jsp?pr...
Now on high:
http://service.futuremark.com/orb/projectdetails.jsp?pr...

See so its about a 1300 point difference, with 5700 compared to 4400. Hope this helps, and btw your fps in games really is the only thing that matters, the thing in catalyst control center doesnt seem to affect fps in games at all.
April 2, 2006 11:28:25 PM

just a reminder guys, i posted this a few posts back. this is with the same video settings, and same 4gb ram. It was the motherboard, or at least the settings on the motherboard. I run 4gb ram as I'm a wedding photographer and have lots of raw images loaded at one time when I'm processing a wedding. Yes I only get 3.25gb ram but when I move to XP64 or vista that should use the whole 4gb.

Quote:
well anyway moving on, I got rid of my DFI motherboard and bought the new asus motherboard, the A8R32MVP. I was reading on some forums and saw similar complaints with the DFI (though i know it is a good overclocker) so wanted to swap it out, and in addition the DFI had really poor USB2 performance which annoyed me (lots of external drives).

Anyway, just doing that m/b swap, and changing nothing else and I'm now getting 11264 3dmarks in 3dmark05.
April 4, 2006 7:41:39 AM

Okay, this has been irritating me for a while now. Whenever people post there 3DMark scrores they never post the settings they're running it at. I can also set it to run at 800x600 an get a higher score than someone running it on 1600x1200. Drivers also make a big difference, so when people post they should post there score, settings, and drivers. So just check that the people you are comparing your card to are running on the same settings as you before you think that your pc is the problem :D 

I'm waiting for my X1900xtx, currently have a 9800pro.
April 4, 2006 3:21:08 PM

i believe most people run their mark scores at the default settings for that particular version in order to maintain continuity with with scoring. it is always best when posting to have run your marks with the defaults.

though posting clock speeds is very helpful too.
April 5, 2006 8:45:36 AM

yeah, got my X1900xtx yesterday :D 

I ran 3DMark05 on 1600X1200, everything else default. Default clock speeds, using the omega drivers based on the catalyst 6.3 and I got 8860 3DMarks.
!