Upgraded memory, less performance?

[bokster]

Hey guys, I just got a 2gb kit for my current system:
Gigabyte GA-7vt600-P
Athlon XP 2600+ Barton
ATI AIW 9800pro

I originally had 2x512mb Kingston CAS3 ram. Everything stock (fsb 166mhz, memory running 1:1) , I get a a 3Dmark '03 score of 5528. I just installed 2x1024mb Mushkin XP4000 Black Series, and I'm scoring 2880 on average. I've tried simply tightening the timings, but no improvement. I don't want to overclock or anything, I am just wondering why the new memory is underperforming? Is it because it's rated at pc4000, and running it at pc2700 is holding it back?

I am going to use this memory on my newer system, but decided to test it out on the old one, just to see if doubling the memory will increase performance in synthetic and real world benchmarks. And it seems to be doing really bad. Can someone please explain? Looking for quick simple answers or links to guides that might help, cuz I suck at searching. THanks.

 

[0roo0roo]

maybe its the mb
the ram either crashes your system if its bad or runs as fast as spec. its not the ram.

 

[Jiggz]

Can you post the setting on the ram timings? This could be a simple ram timing setting in the bios. As for your suggestion that it is rated at PC4000 and running a PC2700 will not have an adverse affect on performance. Instead you should be able to keep a tighter or faster timing on the rams. At PC4000 you might be able to ran the rams at CAS2.0 or at least CAS 2.5 if you're only running them at PC 2700.

 

[bokster]

I set the timings on 2-2-2-6 at 1T. It boots fine, everything runs fine, but i yield the same results as 3-4-2-7 at 2T timings.

 

[imported_hopeless]

May want to take out the old ram & put the new ram in the first two slots and try it with just the 2 new sticks of pc4000.

 

[bokster]

Yup, their all alone. The thing is my old 2x512 ram was actually mixed with a 256mb pc2700 stick. Performance was actually better with the extra 256mb. When I installed the new ram, I used all three slots for 2.5 gigs, but my system was unstable, so I took the 512mb out, and everything runs fine, except my performance has halved. Could this be the result of faulty memory? Or a single memory stick bad?

 

[Matthias99]

Originally posted by: bokster
Yup, their all alone. The thing is my old 2x512 ram was actually mixed with a 256mb pc2700 stick. Performance was actually better with the extra 256mb. When I installed the new ram, I used all three slots for 2.5 gigs, but my system was unstable, so I took the 512mb out, and everything runs fine, except my performance has halved. Could this be the result of faulty memory? Or a single memory stick bad?

Something's out of whack; slow timings wouldn't HALVE your overall gaming performance. Even running in single-channel mode wouldn't do that. Running memory at a lower speed than it is rated for doesn't hurt performance (although running out of sync on an AthlonXP platform will; make sure you're actually at DDR333).

Are you sure you're installing the memory in the right slots? Some boards require you to use particular slots with two DIMMs installed; consult your manual. You could also run memtest86 and Prime95 to see if your memory is running stably; bad memory could cause a lot of different issues (though you'd usually see crashes rather than just bad performance). Also make sure you didn't do something stupid like unplugging the GPU's power feed or nudging it partially out of the socket while you had the case open.

 

[furballi]

Try one module. Also run a quick check with 1M digits Super Pi using the old and new RAMs. Does CPUz properly detect the RAMs?

 

[bokster]

The mother board has only three slots, no dual channel. The sticks are in dimms 1&2 and nothing in 3. The manual states that I should be using the dimms in consecutive order. I did have to remove the gpu to instal the memory but I made sure it is correctly in place, and that all the memory sticks are secure as well. I could run the memtest progs but performance is my problem not stability. Even real world performance in games has dropped significantly, but everything else runs fine. Computer boots and loads just as fast or a little faster actually, every application loads within seconds. It's pretty much the games that are suffering and don't know why!

I'm at work right now so I can't really check. I was pretty tired and sleepy when I was working on this, I'm sure everything is in order, but I'll double check later when I get home to make sure.

 

[Fox5]

I'd say make sure your fsb and memory are at the same speed. (166/166) That is the only possible explanation I can think of. Your system probably dropped from 1T timing to 2T timing when you added more ram, but that wouldn't give anywhere near the performance decrease you're seeing.

 

[m1ldslide1]

Originally posted by: bokster
Yup, their all alone. The thing is my old 2x512 ram was actually mixed with a 256mb pc2700 stick. Performance was actually better with the extra 256mb. When I installed the new ram, I used all three slots for 2.5 gigs, but my system was unstable, so I took the 512mb out, and everything runs fine, except my performance has halved. Could this be the result of faulty memory? Or a single memory stick bad?

If you're even considering faulty memory to be an option, then you need to run memtest86 and prime95, one stick at a time FIRST and both sticks together SECOND.

My mushkin pc3200 is running at 2.75 vdimm right now (OC'd)-- whats yours at?

 

[bokster]

Originally posted by: Fox5
I'd say make sure your fsb and memory are at the same speed. (166/166) That is the only possible explanation I can think of. Your system probably dropped from 1T timing to 2T timing when you added more ram, but that wouldn't give anywhere near the performance decrease you're seeing.

Yup, they are at the same speed. Timing has always been 2T. The wierd thing is i've tried lots of combinations and all of them yield exactly the same performance on 3dmark 03. I've tried 166mhz with tight timings (2-2-2-6 1T & 2T) and loose timings (3-4-3-7 1T & 2T). I've also tried those timings at 200mhz, but 2-2-2-6 was unstalbe so its out of the picture. All pretty much the same score in 3dmark03 and same performance in halflife2 and doom3 & Fear. With my one gig of memory, all those games played a lot smoother at the same settings I've always had them at. I haven't tried increasing the fsb or something because that will go into overclocking and complicate things more. I just don't understand why with everything stock, the 2x512 CAS3 (running at 3-4-3-8 2T 166mhz) kingston performs much better then 2x1024 XP4000 black (running at 2-2-2-6 2T 166mhz) mushkins.

 

[furballi]

Not sure why you cannot run at 1T with one module. I've had a similar problem with 2 x 512 Kingston VR stuck at 2T on the ECS NF3 board. Switched to Corsair VS and was able to run at 2.5-3-3-5-1T using DIMM1 and DIMM2.

 

[bokster]

Originally posted by: furballi
Not sure why you cannot run at 1T with one module. I've had a similar problem with 2 x 512 Kingston VR stuck at 2T on the ECS NF3 board. Switched to Corsair VS and was able to run at 2.5-3-3-5-1T using DIMM1 and DIMM2.

Yes I can run 1T with one module, I can run 1T with both, but it doesn't change anything, 3dmark 03 score still around 2800-2900.

 

[bokster]

I've figured it out!!! Correct me if I'm wrong. I just stumbled across Mushkin's motherboard-memory compatibilty list, which I've been trying to find for a long time now, and according to the list, my motherboard is compatible only with the Green EM modules and Blue HP modules, so I've come to the conclusion that although the XP4000's works 100% with my computer, the slow fsb and older chipset cannot use the full potential of the memory. Probably the only way I can actually use that potential is by overclocking, so my next system, with a lanparty ultra-d mobo will surely be able to accomodate the speed and size of the memory, plus get even more through overclocking. Thanks for the suggestions!

 

[Fox5]

Umm, as long as the memory is running at the correct speed, I don't think there should be a performance problem.