Memory help...

noiseordinance

Hey there, wanted to see if anyone had any advice on memory timing. Here's my specs:

ASUS A7V8X-X
AMD 2800 Xp (Barton)
Kingston Value RAM 1GB DDR400
Samsung 256 x 2 DDR400
(Total RAM 1.5GB)
Gainward GeForce 5900XT Golden Sample AGP (128mb)

I'm running the processor at stock speed (12.5 x 166) and have no clue where I should start with memory timings.

The problem is, my system doesn't benchmark like it's supposed to. I see some people getting twice the framerates as I on similar or lesser specs. I have a nice fast hard drive. My Windows is freshly reinstalled. Is there some bios tweaks or anything I should be aware of?

Any pointers would be greatly appreciated!!!

RWB

You have two different memory types and if your system supports Dual Channel memory - it ain't runningin dual channel becuase of the memory configuration. Plus it's probably running CAS latency of 3 or 2.5 vs. 2 or so it might be able to do if you had only the Kingston memory in, but no gurantee. Plus the command rate would knock down the memory bandwidth if it's running at 2t instead of 1t... again becuase of the memory config. These are my best guesses though.

It'll help if you gave more specifics like how is it not benching the way you want it. Not to mention others with siilar specs may be overclocking which you should take into consideration.

edcentric

Try running with just one stick of each type of memory and see what the default spd settings are. My guess is that you will run faster with 2x256 only.

noiseordinance

Well, I originally had the 2 sticks of Samsung. Problem is that the game would hang up in spots and my friend said having 1GB would help a ton. Well, he was right, when I popped in the stick of 1GB it totally helped a lot. That's what made me look at othe people's benchmarks. People with the same video card, same processor and 1GB of RAM were getting 30fps at 1600x1200 with the antiscopic filtering turned all the way up. When I try it at 1280x1024 my frames are anywhere from 10-20. I just took out the two sticks of Samsung and it seems to be a slight bit better. I'm just wondering if I'd benefit a lot from a better mobo. This one I have is notorious for memory problems. I'm lucky the thing even works. It only holds 4 banks of RAM. The Kingston is two-sided and the Samsung is one-sided. All together, I maxed it out but it seems like it it better off with the solo stick of Kingston. I'm just sorta lost. CPUZ reports my RAM at 2.5, 5, 4, 9 even though the BIOS is set to manual for 2.5-3-3-7. Guess I'll just have to settle with being happy that the thing works? lol

RWB

Try the CAS at 2, you may need to up the voltage of the DDR... to 2.8v or so, it may be at 2.5, it should be just fine at 2.8... but it maywork at 2.5. GO FIND OUT lol

I noticed it's a single stick of 1GB DDR, you probably may need to up the DDR voltage, but maybe not... does CPU-Z report proper CPU speed and such?

noiseordinance

Ok, so I'm a dummy. I went into the bios and even though it displayed 2.5-3-3-7, it was set on auto detect by speed. So I set it to manual and left it at 2.5-3-3-7 and now according to CPUZ my speed is 2.5-3-4-7. Is that a good thing? LoL. I'm wondering if I should plug both of the other RAM sticks in there. I'd rather put them to use than have them sit idle. What would be the best way to tell if I benefit from all three at once? Are we talking about incredibly subtle differences?

As far as what CPUZ reports, it says it's running at 166.6MHz, a 1:1 ratio.

Sorry I'm sounding like a total rookie here!

noiseordinance

Ok so I'm totally lost here. I've tested the memory speeds with both Samsung memory and the single Kingston by itself.

When I manually set the Kingston to 2.5-3-3-7, cpu-z reported 2.5-3-4-7.
When I set the Kingston to Auto Detect by speed, cpu-z report 2.5-5-4-9.

When I manually set the 2x256 Samsung to 2.5-3-3-7, cpu-z reported 2.5-3-4-7.
When I set the 2x256 Samsung to Auto Detect by speed, cpu-z report 2.5-5-4-9.

When I manually set all 3 RAM sticks to 2.5-3-3-7, cpu-z reported 2.5-3-4-7.
When I set all 3 RAM sticks to Auto Detect by speed, cpu-z report 2.5-5-4-9.

In other words, they all appear to have the same speed and make no difference. Am I understanding this correctly? Would you just leave all 3 in?

Lastly, I'm thinking about upgrading to an Athlon 64 board and am wondering if this Kingston 1GB DDR400 stick can be used or if I need to switch to DDR2?

Thanks, and sorry for being clueless here...

muddocktor

Since you aren't overclocking your processor or running the fsb speed over spec, you might be able to tighten your memory timings further, since that is PC3200 memory you are using. I would leave your memory speed running as it is presently at 1:1, since that generally gives better performance with AXP. But since it's only running at 166 MHz instead of it's rated 200 MHz, you might be able to tighten your timings to 2-3-3-7 or so. It might even be able to run at 2-2-2-7 timings too. But be sure to test with memtest86 when tightening your timings past the rated timings programmed into the spd on the ram.

As far as an A64 upgrade, yes you can do that but if ypu want to go with socket 939 you will need to get another matching stick of the Kingston stuff so that you are running the memory in dual channel mode.

noiseordinance

I set everything to 2-3-3-7 and have been running memtest in DOS for an hour and a half. It's ran through all the tests once and is 10% into the second cycle. It says Pass: 1, Fail:0. How long should I let this thing to run? When after running, it doesn't prove to have any failures, should I take it to 2-2-2-7 like you mentioned?

Thanks!

muddocktor

If you make 2 complete passes without an error, go ahead and try setting your memory timings tighter and run memtest again. If you can make 2-3 passes without an error, your ram should be good to go.