OC FSB apart from CPU and RAM ?

I notice that alot of a64 users are pushing the FSB to the level that the RAM can handle with nice timings... and using dividers and multiplier just to set the cpu at its max level afterwards...

Is there any advantage to running a higher FSB apart from the increase you might get in your RAM and CPU clock speed ? In other words, if the cpu speed and ram speed stay the same, and you crank up your mobo's FSB only, by using a multiplier and divider to keep the cpu and RAM at same levels, will this increase speed significantly?

If so does this mean that an overclocker should determine what the top FSB of the mobo is before determining what the RAM and cpu can handle?

Thanks,

Having a higher FSB will make a difference, but in real apps and games I don't think you will see it. You will see a higher frame rate in your games benchmark. but seeing the extra speed with your eyes? nah.
Higher the FSB, the better.

With A64's, I believe the hyper transport bus is determined by (HT-bus X HT Multi, normaly 3x or 4x). So having higher HT-bus speed will give higher performance, but not by much at all. With Intel and AXP's, the FSB was how fast the cpu talked to the northbridge, but since the 64's have most of that integrated, you might as well leave it how it is.

Short answer, for Intel, it helps a lot. For XP's, running the ram and fsb async loses performance. For A64's, you wont see much difference at all.

I see... thanks for the explanations.

Silent Frog said:

Having a higher FSB will make a difference, but in real apps and games I don't think you will see it. You will see a higher frame rate in your games benchmark. but seeing the extra speed with your eyes? nah.
Higher the FSB, the better.

Click to expand...

I've seen a 5-10 fps increase in farcry going from 180 fsb to 217 fsb, during testing (also changing the CPU multi so cpu speed stays similar).

Think about it... if your games benchmark runs faster, games would similarly run faster... obviously 5 fps isn't a huge difference and wouldn't be visually obvious to most, but to fps h00rs, it would certainly be an improvement.

Yuriman said:

For XP's, running the ram and fsb async loses performance. For A64's, you wont see much difference at all.

Click to expand...

XP + Async = *shutter*

the cpu clock is determined by the fsb clock through the cpu multiplier

fsb x multiplier = cpu clock

your ram runs at double your fsb (ddr)

by increasing your fsb you overclock your ram and cpu simultaneously.

at a certain point either your ram or your cpu won't be able to handle the higher fsb.

in the case of your cpu maxing out you can't gao any higher.

if your ram maxes out you can use the ratio in order to have bragging rights on a fast cpu ( there is virtually no performance gain (sometimes there is a loss) from using the ratio).

using the ratio can also be called using a divider

using the ratio is running your cpu and ram async

the only really good ratio is 1:1 (sync) this is the only way you'll keep seeing performance gain from overclocking. if you reach a point where you know your ram is bottlenecking you you can buy faster ram ie pc 3500, 3700, 4000 etc.

But none of the stuff you just said applies to A64's...

yes it still does

when you overclock an a64 you overclock the memory bus and the ram as well as the htt

the htt only handles the comunication with the northbridge (ram is not connected to this on a64)

instead the ram runs through a traditional bus that leads to the on die memory controller.

http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/24659.PDF

thats the features list of a64's it notes the ondie memory controller and hypertransport links separatly

this is because the memory bus (fsb) is 64 bits wide and runs at 200 mhz stock

940 and 939 socket processors feature a 128bit wide bus that handles dual channel ram

the htt is 16 bits wide and runs at 200 x the htt multiplier = either 600, 800 or soon 1066 (because they feature ddr technology it comes out to 1200, 1600, 2132)

the reason the htt bus never gets saturated is because it has all of that bandwith 1200 x 2 = 2.4 GB/s bandwith unburdoned by ram. Which compared to an intel bus its 800 x 8 = 6.4GB/s that also has to carry the ram which has a maximum theoretical speed of 6.4GB/s.

so when your overclocking an a64 out of sync you don't lose performance but you don't gain any either.

I know a64 motherboards feature the ability to use a divider to set a maximum mem clock compared to the fsb 6/5, 5/4

because if the clock generater is still on the northbridge and sends the cloc from their into the a64 then the ram goes through a divider to get 200

htt x multiplier = processors clock

divide that by the memory divider and you get your ram clock

but, how about we just look at it as, fsb x multiplier = processors clock

because really your fsb and htt are always going to be the same unless you change the divider (or for symplisity we could call it a ratio because thats all it really is)

basicly overclocking an a64 is the same as overclocking an axp from the users perspective ie. you don't benifit from running your ram async, you can use something simlar to a ratio to continue overclocking the processor

You gain performance by runing "async" as you can get your CPU to run faster. Are you trying to tell me that if I ran my proc at 2.3GHz "in sync" it would be quicker than 2.5GHz with a CPU/11 divider? Because that's quite obviously not true.

And the reason I put sync in quotes is because you're technically always running asynchronously with the A64- async with respect to the CPU speed, which is what the memory is derived off of. There is no "6/5" or "5/4" "ratios." No such thing exists in reality with the A64's.

Since the HyperTransport and memory buses are independent, you don't really have to care about the HyperTransport. Just get the memory running as quickly as necessary, and keep the HT as close to stock as possible.