PDA

View Full Version : FX 5900 Performance Issues


Eck
09-04-03, 10:36 PM
And the story begins:

Around February of 2003, I decided to upgrade my GeForce 2 Pro to a Radeon 9700 Pro. Once I had all the drivers and whatnot installed, I went to test it out. I was getting around 12,000 in 3DMARK01 SE and around 5,200 in 3DMARK03. Game performance was mediocre, should've been better. I tried everything to no avail. I eventually sold the card on ebay, thinking it was ATI's fault.

Well now, I've got an FX 5900, exact same problem. It's obviously something in my system, most likely my motherboard. I've elminated IRQ issues, sound card, tried all kinds of tweaking, blah blah, you name it I've done it. I have the latest VIA hyperions, which are 4.49. I doubt anyone has a fix for my problem, or ever will, since it's probably just the VIA chipset/Soyo motherboard that suck @$$, but I thought I'd throw it out there one last time...

Only other thing I could think of would be my power supply, which is 300 watts (the minimum). It's got 15A on the +12V line though, so I've always kind of ruled that out.

All ideas are welcome.

digitalwanderer
09-04-03, 10:40 PM
Question: Could you please describe what is wrong with your 5900? I can sense you're not pleased, but why exactly? What is it not doing for you?

Eck
09-04-03, 10:43 PM
My game performance isn't what it should be. I get around 30-50 FPS in just about any game regardless of settings. My 3DMark benches should be much higher, also. It's really not the cards fault though, it has to be something else. I thought I'd just throw my problem out there to see if anyone has had the same thing and fixed it.

archaeo.
09-05-03, 01:12 AM
I was getting around 12,000 in 3DMARK01 SE and around 5,200 in 3DMARK03. while I don't generally care for the tendency in 3dmark03 towards a purely videocard dependent benchmark, in this case we see that it has it's uses.

your 03 score is quite in keeping w/ what you would expect, thus showing that your 9700pro was probably fine.

however, only 12K in 01 is pretty low for a fairly sweet system. To me, this indicates most likely a memory bus problem (if the proc. really is running at it's stated speed). Have you run any troubleshooting benchmarks, like Sisoft Sandra?

I'd look especially to mem bandwidth results, esp given as to how sensitive the P4 is to that.

ricercar
09-05-03, 01:14 AM
My 5800 Ultra gets a boost of 2,000 3DMark2001 SE (15k->17K) moving from a 300 Watt PSU to a 350W PSU. I haven't specifically duplicated the tests on my 5900 Ultras yet, but have no problem imagining a mere 300 W PSU would hold back a 9700 and 5900.

Eck
09-05-03, 09:04 AM
Yeah, I've got Sanra. My CPU IS running at 2.4 GHz. Here are some of my benches: http://dellepro.dellhost.com/eck/sandra.

These are the performance tips under Motherboard Information:

Performance Tips
Tip 2530 : MPS 1.4 or later is recommended for MP systems. Check BIOS version.
Notice 224 : SMBIOS/DMI information may be inaccurate.
Tip 2511 : Some memory slots are free so the memory can be easily upgraded.
Warning 2545 : Large memory modules should be Registered/Buffered.
Tip 2546 : Large memory modules should be ECC/Parity.
Tip 2 : Double-click tip or press Enter while a tip is selected for more information about the tip.

I really don't know what the first two tips mean or if that has anything to do with my problem.

Also, the memory bench never has been where it should. I've tried setting the CMOS to both 400 and 333 for my DDR, but they both bench right around 2400MB/s.

RM. Andersson
09-06-03, 07:53 AM
I also think there might be a problem with your motherboard. If you are prepared to spend some money to solve your problem you could buy a MoBo with i875/i865 chipset. Perhaps it will help. It will be a much better MoBo and a great upgrade anyway.

Regards!

Uttar
09-06-03, 08:21 AM
Originally posted by ricercar
My 5800 Ultra gets a boost of 2,000 3DMark2001 SE (15k->17K) moving from a 300 Watt PSU to a 350W PSU. I haven't specifically duplicated the tests on my 5900 Ultras yet, but have no problem imagining a mere 300 W PSU would hold back a 9700 and 5900.

Agreed, but I doubt the boost would be as big as with your 15->17 result. Remember the FX5900U does need less juice than the FX5800U ( DDR1 instead of DDR2, lower core clock, and lower per-clock requirements for the core I believe )


Uttar

Rytr
09-06-03, 09:16 AM
I use the 3DMk programs for testing stability and your results apparently do not show any stability issues. If you have instability issues with your system I would first look at the 300W psu.

You said, "I get around 30-50 FPS in just about any game regardless of settings." This is interesting since you have checked about everything I can think of so it seems the mobo is at fault, except, maybe trying less ram if you have not tried that. Anyway, something with the bus may be at fault although your 3DMk scores do not indicate any problem.

I think your evaluation that it is the motherboard may be correct. The 875P boards, while very good, are not necesssary. An 865PE, if you have a 'C' processor or even a good 845 mobo should provide excellent gaming results matched up with the components you now have. There are some very good prices on 845 boards and even some 865PE boards are <$100.

Eck
09-06-03, 09:05 PM
Thanks for the replies guys, it seems it must be the motherboard. As far as which one to get, I'll most likely get an 865PE Asus or MSI board, but not for a while -- gotta spread out the upgrades because of the parents, and I really, really hate formatting.

Thanks again,

Eck

Eck
09-23-03, 04:20 PM
Well, the new board is great. All of the noticable video lag is gone (where it would stutter, but the fps stays high). For some reason, Sandra is only giving me around 3,000 MB/s in the memory benchmark while theoretically I should be getting around 4,000 MB/s. The 3D Mark 01 score is still low, right around 12,500. OK, so it's not the video card, nor the motherboard. Sound isn't involved, and it's not my harddrive. That leaves 3 things: cpu, ram, and the power supply. I'm thinking of getting an Antec True430. Any opinions?

Rytr
09-23-03, 05:17 PM
The Antec 430W should meet all your needs, excellent choice on motherboard. :)

All your components look to be good choices. I recently with an Abit IC7 board with the 2.4C processor. My first Intel chipset board in a long time and it performs very well. Usually Samsung ram is very good so I don't think you have gone wrong there. You might download one of the memory testing programs to test it if you have not already done so. Are you running dual channel or single? If you are running DC and the memory sticks are not matched that could be a problem. If you have two sticks of ram have you tried running the test in Sandra with one?

12.5K seems low for the FX5900 as does the 3M bandwidth result in Sandra. I just run the memory bandwidth test and scored 5300+. I am running PC3700 in dual channel at 225FSB. At 200FSB it runs about 4.7, IIRC. Right now I have a FX5900 Ultra and I score 17K in 3DMK2001se.

Sazar
09-23-03, 05:18 PM
Originally posted by Eck
Well, the new board is great. All of the noticable video lag is gone (where it would stutter, but the fps stays high). For some reason, Sandra is only giving me around 3,000 MB/s in the memory benchmark while theoretically I should be getting around 4,000 MB/s. The 3D Mark 01 score is still low, right around 12,500. OK, so it's not the video card, nor the motherboard. Sound isn't involved, and it's not my harddrive. That leaves 3 things: cpu, ram, and the power supply. I'm thinking of getting an Antec True430. Any opinions?

yooohooo Eck... :)

Eck
09-24-03, 08:54 AM
Yes I'm running in dual channel mode, and from what I know when I ordered my RAM online, they are both sticks of 512MB Samsung DDR 400 cas 2.5. Sandra does see them as slightly different, though. Have a look:

Memory Module(s)
Memory Module 1 : Wintec 512MB 16x(32Mx8) DDR-SDRAM PC3200U-344-650 (CL3 up to 200MHz)
Memory Module 3 : Kingston 00143D05 512MB 16x(32Mx8) DDR-SDRAM PC2700U-2533-700 (CL2.5 up to 167MHz) (CL2 up to 133MHz)

I think I'm just gonna go for the power supply soon.

Rytr
09-24-03, 05:00 PM
You really need 'matched,' or at least from the same batch when buying ram for DC. Your two sticks are probably not close enough to adequately run DC, IMO. This could be your problem. Did you ever try running just one of the chips and see if performance was better? There should be a difference going by what you posted.

Also you can get the numbers off the chips on each stick (all should be the same on a given stick of ram). Then go to the Samsung site and run a check on the numbers to tell you what ram you have.

When buying Samsung I found it best to get 'genuine' Samsung. Only difference is the pcb is made by Samsung and not some other third party. Several sites actually list them differently according to whether or not the Samsung ram is mounted on Samsung pcb. Those on Samsung pcb are usually a few dollars higher.

Eck
09-24-03, 08:55 PM
Actually I did try only one stick, back when I only had one. This was with my Soyo P4X400, but that still kinda rules out the RAM. Right now they're not even running at 400, but rather 266 which coincides with my FSB. I wouldn't think it should make that much of a difference just because of a different PCB.

Rytr
09-24-03, 11:06 PM
Ram performance will be limited to that of the least capable stick of ram when running more than one.

This is my first Intel chipset board in a number of years and I am a noobie at the dual channel thing.

I have learned some, however, in the last four months from running dual channel in my NF2 chipset board. It requires two like sticks of ram to run effectively in DC. That's why OCZ, Corsair, and others offer 'matched' sets of ram for DC setups.

digitalwanderer
09-24-03, 11:17 PM
Gads! I just noticed your problem! :eek:

Why in the heck are you running a SB-live on that nice rig? :eek2:

5150 Joker
09-25-03, 01:21 AM
Originally posted by Eck
My game performance isn't what it should be. I get around 30-50 FPS in just about any game regardless of settings. My 3DMark benches should be much higher, also. It's really not the cards fault though, it has to be something else. I thought I'd just throw my problem out there to see if anyone has had the same thing and fixed it.

Maybe you have vsync turned on?

Edit: nm just noticed you upgraded. Too bad you sold your 9700 pro for a 5900..I feel for you, maybe next time you wont jump the gun so quickly and blame ATi.

Eck
09-25-03, 08:55 AM
Nah, vsync is off. Yeah it's too bad I did blame ATI, though I also had another issue with the card. When 3D stuff moved up or down the screen, I got major tearing. Vsync helped, but didn't get rid of all of it. That was another reason I sold the card. I had always had nvidia before that, and they were always good. Plus, I kinda like the FX 5900, except for it being so fricken' loud. I'll probably wait for the next gen cards, or maybe even for PCI express, and then upgrade.

Originally posted by digitalwanderer
Gads! I just noticed your problem!

Why in the heck are you running a SB-live on that nice rig?

Eh? It's a decent sound card, is it not? I MIGHT get an SB Audigy 2 if I decide to set up my old Dell and use it as a linux box and stick the SB Live in the Dell. But really, I do have good sound.