Question about A64 939 and running RAM Asynch

How much of an effect does running your RAM asynchronously have for an A64 939 system?

In fact the RAM will always run asynchronously on an A64-rig.

By default the memory speed is the same as the HTT speed (when set to 200 (or 1:1)). This doesn't mean they are related to each other in any other way - cause they aren't. The only connection between the HTT and memory, is that the CPU speed is derived off of the HTT, and the memory speed in turn is derived off of the CPU speed.

Instead of the FSB:MEM-ratios known from other platforms, with A64s this is actually replaced by a CPU:MEM-divider that the BIOS won't give you direct access to.

If the max stable HTT-speed is higher than the memory's speed - you'd have to use a lower "ratio" - which isn't what it looks like in the BIOS. Running the memory at lower speed than the HTT could be beneficial in the sense that it allows you to run @ tighter timings. Many OCers prefer tight timings over high memory clock, but this one has to figure out through experimenting and benching. Plus if the memory is holding you back, slowing down the memory would allow you to increase the HTT further - thus yielding higher CPU-OC.

The HT-bus interacts with the peripherals - the memory doesn't use this bus at all. So HTT is not the same as FSB. There is no FSB in the A64-world.

Ok, thanks, that was basically my understand of it. So, there's no performance hit due to using the non-1:1 dividers?

This might make me decide to get a 3000+ instead of a 3200+... if my RAM can't go above 280 or so then I could always just up the HTT, run the RAM on a divider, and still use the X9 multiplier.

Gregory_WE said:

How much of an effect does running your RAM asynchronously have for an A64 939 system?

Click to expand...

Zero. RAM speed is dependant on the CPU speed and not the chipset. Thus A64s take no penalty going asynch other than the lower dram speed.

Ok, sounds good

running async takes a hit because of lower bandwith. I dont know why people keep insisting there is no performance hit, if that was true, why not just buy pc2100 ram and run the ram at 100 divider instead of 200? You need about 150MHz cpu to make up for going from 200 to 166 divider due to loss of ram bandwith

Overclocker550 said:

running async takes a hit because of lower bandwith. I dont know why people keep insisting there is no performance hit, if that was true, why not just buy pc2100 ram and run the ram at 100 divider instead of 200? You need about 150MHz cpu to make up for going from 200 to 166 divider due to loss of ram bandwith

Click to expand...

I fully agree with OC on this one...

Now running async might not hurt the CPU performance...but it does hurt the overall system performance from the lower memory bandwidth...so any real world app or such that needs/wants more memory bandwidth then CPU Ghz/Mhz is going to perform less then running at 1:1 with a lower FSB/HT with better memory bandwidth...

i don't know what to say on this topic, everything is so blurry, i would just do a test when you get the stuff in, 1st run it like this
9 X 200 = 1800 @ stock 1:1
then run it @ 8 X 225 and set a divider on it like 5:6 so your ram is @ 180. Now bench it and see if you're loosing any performance (which you will, but by how much)

There's actually no accurate way to test this as we know that things just don't move in simple arithmatic, but with a little math, you can actually calculate the # to be a better test than the one i've just posted.

Overclocker550 said:

running async takes a hit because of lower bandwith. I dont know why people keep insisting there is no performance hit, if that was true, why not just buy pc2100 ram and run the ram at 100 divider instead of 200? You need about 150MHz cpu to make up for going from 200 to 166 divider due to loss of ram bandwith

Click to expand...

glock19owner said:

I fully agree with OC on this one...

Now running async might not hurt the CPU performance...but it does hurt the overall system performance from the lower memory bandwidth...so any real world app or such that needs/wants more memory bandwidth then CPU Ghz/Mhz is going to perform less then running at 1:1 with a lower FSB/HT with better memory bandwidth...

Click to expand...

I'm surprised, OC550 and glock19owner. You guys are way off. Period.

Sure, you can agree that the earth is flat if you want...

Now listen:

For example, with nForce2, performance takes a hit just from running out of sync, even if the memory is running faster that the FSB. This is all about the latency hit caused by the memory and bus not being able to communicate as efficiently as if they were in sync. On nForce2, both the memory and FSB has to run at the speed of the slowest component, which is hardly an advantage.

On the platform in question though, one has the option of adjusting the memory speed and HTT independently of each other, without losing performance due to latencies, as the RAM doesn't use the HTT. I'm not saying one should lower the memory speed just for fun. But if the memory cannot run as fast as the HTT, it can be adjusted so that it still runs stable, without losing performance due to the latency-issues that are seen on platforms that have an FSB. Thus allowing higher CPU-clock (on a locked CPU). And again, there is no such thing as "running in sync" with A64s.

they are not way off. i also a agree with them because ive tried it and i KNOW for a FACT that theres alittle bit of a proformance hit theres not an incrediable amount but theres some

a performance hit doesn't make sense, solely from the divider. what does make sense is a performance hit from slower ram.

if your ram ends up slower than it was before, all other things equal, your pc will be slower. duh. to make up for this, you'd have to clock your cpu a bit higher. in some cases, slower ram will enable a faster cpu, giving you better overall performance. would you give up 5mhz of ram speed for 200mhz of cpu speed? of course. would you give up 20mhz of ram speed for 100mhz of cpu speed? that's a harder choice.

it's all about balance. some tasks are cache-bound, others are io-bound, all that matters is finding the right mix for whatever your task is. blindly reaching for 1:1 all the time is slowing down computers more often than not.

my 3000 on air .. so far

Steven4563 said:

they are not way off. i also a agree with them because ive tried it and i KNOW for a FACT that theres alittle bit of a proformance hit theres not an incrediable amount but theres some

Click to expand...

do you know how much of a hit it would be? is it as big as 200mhz? because if it is then it's pointless to hit high speed cuzz lower speed with ram sync will result in lower votage, lower temp, more stablity, and all that stuff.

i agree with above.. in fact the dividers work towards your benefit..

if you can only run your ram so high you can still push your cpu .. and vise versa or verse visa...

but whatever winchester you get .. nice ram will be key to great bandwidth..

the hardest part though is setting it up in the bios .. lol

Overclocker550 said:

running async takes a hit because of lower bandwith. I dont know why people keep insisting there is no performance hit, if that was true, why not just buy pc2100 ram and run the ram at 100 divider instead of 200? You need about 150MHz cpu to make up for going from 200 to 166 divider due to loss of ram bandwith

Click to expand...

Simple way to prove it is to get someone to test a 3500+ run it at 229 x 11 at 1:1 and 280 x 9 at 5:6 that way the cpu and ram frequencies are virtually identical and if running async does give a performance hit then it would show up in benchmarks.

Miguita said:

On the platform in question though, one has the option of adjusting the memory speed and HTT independently of each other, without losing performance due to latencies, as the RAM doesn't use the HTT. I'm not saying one should lower the memory speed just for fun. But if the memory cannot run as fast as the HTT, it can be adjusted so that it still runs stable, without losing performance due to the latency-issues that are seen on platforms that have an FSB. Thus allowing higher CPU-clock (on a locked CPU). And again, there is no such thing as "running in sync" with A64s.

Click to expand...

So Based on what you just stated...how is running at say, 320FSB/HT with 5:6 divider and 3x HT be faster/equal then running like my setup at 275 1:1 with 4x HT and tighter timings...

OC Dect:

If you know of a good way to run a 3200 like you stated with the 3500 above...let me know and I will rum some benchmarks with latency test included...

warlock110 said:

do you know how much of a hit it would be? is it as big as 200mhz? because if it is then it's pointless to hit high speed cuzz lower speed with ram sync will result in lower votage, lower temp, more stablity, and all that stuff.

Click to expand...

when i had it at 220 1:1 and cpu @ 2.2ghz its about 1000 points in 3dmark2001 slower than how my rig is now see sig and also my Boinc WU's have dropped about 10mins roughly

What I'm gathering is if your RAM speed is still the same synch and asynch, the performance would be virtually the same - which is what I was specifically wondering about. Obviously, a system at 300 1:1 will be faster than a system at 300 5:6.

your cpu ghz/mhz plays a huge roll in this .. alot more of a jump going from 2.0ghz to 2.4ghz in bandwidth when comparing socket A to 939..

Sorry, but anyone who thinks that there is no performance penalty for using dividers is flat out wrong. While it is nowhere near as significant as on an Athlon XP system, performance differences do in fact exist. Here's a good article that should clear everything up:

http://www.insanetek.com/index.php?page=overclocka64

deception``