• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Memory Frequency Percentage

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

John Malachi

Registered
Joined
Oct 16, 2003
My BIOS offers the ability to set the memory bus speed to a percentage-multiple of the FSB. It offers a range from 20% up through 100% to 200%.

How do I use this during overclocking? If I leave it on Auto, it clocks down the mem bus when I take the FSB over 150MHz. If I set it to 100%, it forces the mem bus to stay in sync with the FSB. I feel this setting could almost certainly be used to do interesting things. I'm just too new to overclocking to tell what, exactly.

Currently running:

System as listed in sig, except current settings...
9.5 X 155MHz
and memory at 5, 2, 2, 2 without so much as a whisper of difficulty.
 
To run it in sync with your FSB run it at 100%. Basically the % are that of your FSB. So 100% of your FSB would be the same.


So say you want a 200 FSB and 200 memory, you would select 200 for the FSB and 100% for the Memory to get 400/400 for both. AKA in sync.

The math for it is 200x100%
200x1.00=200

Just move the decimal over 2 spots to the left and multiply the FSB and you will get your memory MHZ
 
Uh...

That much is apparent to me, but thank you. :)

Query: Is it possible for the mem bus speed to be the weak link, thus permitting a slightly higher FSB speed if the memory is run at a slower figure (e.g. 80%)? Exempli Gratia: System maxes out at FSB 170 MHz. Drop the memory down to 80% of that (136MHz), and could the system bus continue to be increased until the memory is once again at or near 170 (FSB would then be about 212MHz).

Would the inverse be true? Might it be possible, at maximum sustainable FSB, to get a bit more juice out of the memory by using a slightly increasing the percentage multiplier?
 
This is mainly seen on Intel Boards AFAIK to allow the bus to be overclocked higher than the memory will allow. The memory may stop being able to handle speeds higher than 170MHz (let's say), while the bus speed could be pumped all the way up to 200MHz. In this case, on a system without memory dividors, your RAM would hold you back from a super high OC.

Enabling the memory to run asynchornusly from the FSB will allow you to make the most usage out of your memory (by running it at ~170MHz) AND the CPU/bus (by running it at ~200MHz).

However, running the memory asynchronusly results in reduced performance when compared to a synchronus speed. Really, you need to benchmark you system in both configurations (highest synchronus speed, and highest asynchonus speed) and see which is the fastest.

JigPu
 
Oh I see what you mean now.........

As far as a test I would try it, but for performance I dont think you will gain anything. I know when I would try the inverse of that 266/333 I gained nothing, however if I ran 266/266 I had faster speeds on my AMD setup.

I can only imagine that running the FSB higher than the memory you will lose even more performance, although it may be a good test to max out your FSB without worrying if your memory is the weak link. JUst make sure the PCI and AGPs are locked at 33 and 66.
 
On most motherboards, you see it as a ratio rather than a percentage, which is why you'll see people talking about "1:1" (100%), or "5:4" (80%) or "6:5" (uh, 83.3%, I guess) memory configurations. That's the ratio of bus to RAM speed. You normally wouldn't run the RAM faster than the bus, because the bus can only talk to the RAM once during each of its cycles. It might increase transfer speed very slightly, though.
 
Ah HAH!

I thought they might be the same thing! Thank you Mathias.

In other news, 162 X 12.5 boots all the way to windows, but fails Prime95 within a minute or two. Rounding error, expected less than 4, got 5 (or approximately 5). Whence cometh this error? Is it mem related, FSB specific, or just too much total speed for my chip?
 
It's either the RAM or the CPU crapping out at that speed. If it'll run with the ram async (use the above multiplier to put the RAM at a lower speed), then the RAM won't OC that high. If it still fails with the RAM at a known good speed, then the CPU just won't go that high stably.

You might also want to try running with each stick of RAM individually if you have more than one in there now -- sometimes one of them won't OC as high as the others, or the memory controller can't OC as high with multiple sticks in.
 
Again, thank you. I'll try those. So far no luck getting the RAM to run async, though, unless I let the Mobo do it with the "Auto" setting (which is different from "by SPD"). We'll see what happens. Thanks for your help. I'm just pleased to see it POSTing. This is the first time I've ever overclocked a computer. Getting this little XP2200+ all the way to Windows at 2GHz gives me a nice, warm fuzzy feeling.
 
Most BIOSes are really bad at reading SPD settings, so "by SPD" is a crapshoot. Most people find that the best performance comes from running the mem in sync with the FSB (at least on AMD platforms). 72C under load is really, really warm. I wouldn't be suprised if it was causing instability. Try runniung the same fsb with a lower multiplier, so that you can run your processor at stock voltage and speed. This should help isolate whether it is a RAM or CPU problem.
 
Warm, he says.

Yeah, 72 is just a bit warm for my tastes, too. :) Right now I'm running the computer at only slightly above stock settings for the sake of temperature. For now, I'm just playing around at the higher settings, seeing what it will handle and still remain stable. When I get the cooling issue worked out, then I'll overclock it for keeps.

In the mean time, messing around with it is a good learning experience. :)

Thank you for your advice, Gnufsh. I've tried several things, and it seems to be pretty much a processor limitation. We'll see what happens if I ever get it cooled off.

Thanks again.
 
I know that most people talk about synchronous being the fastest, but it doesn't make too much sense to me. Are you guys saying that in two otherwise identical systems, one running 200fsb and 150mem is slower than one running 150fsb and 150mem (we're assuming that cpu speed is same for both as well by manipulating multipliers)? How about the opposite? If your mobo isn't letting you hit high fsbs and lets say its limited to 200, yet you have some crazy ram, would increasing ram over 200 really decrease performance?

It just seems to me that if you have the option to increase performance at any level, you should take it. What am I not understanding?

Mark
 
If we're talking about AMD systems, yes. The bandwidth between the northbridge and the processodoesn't have any room for the increased bandwidth. Some things will get a little better, but this is mostly counteracted by how inefficient the chipsets are. Best to sync up, and go for tighter timings, then run faster Async, I believe.
 
Back