Can Someone Help Me With My RAM Timings

[supastar1568]

ok, I currently just brought my CPU up to 2.2GHz. As you know, with 4 sticks of RAM because of the processor, it likes running at 166Mhz.

With "Auto" on in the BIOS, and the memory divider set to 2/1.66 I could get my memory to 185 Mhz, and the timings would be 2.5-3-3-7.

But I can set my divider to 2/1.83 to achieve 200mhz on the memory (as this is what it is rated at). When I do this with "Auto" it gives me timings of 2.5-4-4-9. I manually set them to 2.5-3-3-8.

Can I be safe with this as this is what the stock timings should be???

Also, should I be running 1T or 2T???

Thanks

 

[Bobthelost]

If you can run at 1T then run at 1T.

If you can run the RAM at 2.5-3-3-8 then run them at that speed.

Trying to run your RAM too fast won't hurt it, if it fails memtest or Prime 95 then it's being pushed too hard and you'll need to relax the settings, but untill it fails memtest then it's fine.

(You may well be stuck with 2T, the older AMD 939s and motherboards definitly were, but some of the newer ones can do 1T for 4 sticks if memory serves)

RAM isn't worth crying tears of blood over on A64 systems.

 

[supastar1568]

yea, i just tried changing it to 1T in BIOS, it didnt like it at all, as it booted from the second BIOS and loaded all of the defaults.


My question now is, when I get my RAM to 200Mhz, do I leave it at 2.5-4-4-9 (which it suggests with AUTO), or do I change it manually to 2.5-3-3-8 ???

Either way Im gonna have to run 2T

 

[Pyrokinetic]

With four sticks, you are definitely stuck with 2T, since the DFI Crossfire board is the only one that can run 4 sticks at 1T. The timings with auto are probably the default memory settings, however, if it will pass Memtest86 at 2.5-3-3-8, all the better. Since you are already at the 2T command rate, I don't think it will make all that much difference running 2.5-4-4-9 or 2.5-3-3-8. The latter might be a smidgen faster, and it will not hurt the memory or the system, so go for it. Just remember to run Memtest86.

 

[supastar1568]

ok, im gonna try memtest.

But ive tried to install it a couple of times now and have no clue how to do it.

Ive tried burning the ISO file to a CD and also tried booting from a floppy.

could someone tell me how to do it correctly?

 

[Pyrokinetic]

Been awhile since I made my floppy but you have to first make a bootable floppy from your operating system. After the bootable floppy is done, you need to copy the unzipped files from http://www.memtest86.com/ (I think I used the 3.2 windows version) to the floppy. Then just boot your system from the floppy -- just make sure your boot priority in BIOS has the floppy listed first.

P.S. -- I never got a bootable CD to work either.

 

[supastar1568]

ok thanks.

im gonna try it with 2.5-3-3-8 timings

 

[supastar1568]

I have another question:

Does the RAM frequency have a precedence over the timings?

Like am I better off with 200Mhz at 2.5-4-4-9 2T then 185Mhz at 2.5-3-3-7 2T

or even 190Mhz at 2.5-3-3-8 2T?

My guess is that the higher frequency the better, then comes the timings, but Im not 100% sure.


Im going to try 200Mhz at 2.5-3-3-8 though.


So is RAM frequency a bit more important???

 

[supastar1568]

bump to my last question

 

[Pyrokinetic]

Depends. The ideal would be to have both. The AMD chips, with the onchip memory controller, do better with faster timings. Much past say 3-4-4-8, I am not really sure you will see better performance from more bandwidth. Since most chips are not limited by DDR400, what benefit is DDR500 or DDR600? For AMD chips, the wiser action has been to get faster timings as close to DDR400 (or above) as you possibly can.

As for questions "Like am I better off with 200Mhz at 2.5-4-4-9 2T then 185Mhz at 2.5-3-3-7 2T" the answer would probably be that they are so close that you probably wouldn't notice the difference. Timings usually only will effect performance in the 1% to 3% range given the same bandwidth. Small stuff. I think you would notice a difference between fast DDR400 ram (2-2-2-5) and ram that is running DDR400 at 2.5-4-4-9. But not a huge difference.

Yet, if you can push the CPU speed far enough, it might even overcome any slow timings; provided you use some elaborate watercooling setup or phase-change. In such systems, CPU performance is king.

I know this really doesn't answer your question, but in truth, the only way to find out is to run "real world" applications on your system. Like how much time to load a particular game or to burn some media to disk.

On my system, I have a San Diego 3700+ (2.2Ghz) overclocked to a stable 2.8Ghz. Since it is a single-core chip, and I not a huge multi-tasker or "power-user", I knew I would probably never bump the ceiling with a DDR400 bandwidth. Thus, I opted to overclock the CPU but not the memory, which is why I use the 166 divider to keep my memory at DDR400 with the overclock. (255x11=2805); 166 divider = 2805/14 = 200.35 or DDR400.

Does ram frequency have a precedence over the timings? For AMD, most of the time no; for Intel systems, yes. On the AMD side, you will find advocates for both positions, but most people will tell you to go for the faster timings. I, for one, am solidly in the faster timings camp.

AMD's own designs testify to this fact. While Intel went to DDR2 (much higher bandwidth) some time ago, AMD remained with DDR and a standard of DDR400. Why? Because their own testing confirmed that DDR400 was absolutely sufficient. Higher clock speeds on chips has not changed that recommendation. The FX series of processors do just fine with DDR400. In the end however, you'll have to convice yourself.

 

[Pyrokinetic]

To be fair, I will also refer you to this Anandtech article about memory on the Athlon 64 platform:http://www.anandtech.com/memory/showdoc.aspx?i=2226&p=1
There, the test did show some performance benefit to higher bandwidth (on certain sticks like the OCZ) yet the difference was still fairly small in my opinion. Most of the memory got their very best scores around the DDR500 point, so you would have to do trial and error using Memtest86 and Sisoft Sandra for stability and benchmarking. I am not convinced it really would be worth the trouble though. Good luck.

 

[supastar1568]

ok, thanks for the replies.

Yea, I just did a 24 hour memtest without a failure at 2.5-3-3-8, so I think Im good with the memory.