Running a-sync = poor performance for AMD?... maybe not

[Jeff7181]

Lately I hear a lot of people saying running your memory a-sync hurts performance, and I agree with that half the time. I agree that running your memory faster than your FSB is pointless because Athlons don't need that extra memory bandwidth. However... if you can't increase your multiplier past 12.5, and you have PC2700 RAM... if you want to overclock much, running a-sync is your only option to get that extra clock speed.

While running in sync provides better performance at equal clock speeds, a 2.2 Ghz Athlon XP with a 220 Mhz FSB, and 166 Mhz RAM will perform better than a 2.0 Ghz Athlon XP with a 166 Mhz FSB and 166 Mhz RAM.
The Athlon XP isn't starved for memory bandwidth like the P4 is, so I see no problem running things a-sync in order to achieve a higher clock speed.

I've tested this myself... and even at similar clock speeds, the faster FSB provides better performance than keeping it in sync with the memory.

Of course, if you are able to run things in sync, you should do it, but if you're stuck with slower RAM that doesn't overclock, running a-sync is your only option, and in my opinion and experience, it's not as detrimental to performance as some people have been eluding to.

 

[DAPUNISHER]

Where are the comparative benchies at? Thugsrook and the gang have spoiled us, we like to see the results For me I prefer screenshots of everything but I understand many are still on dial up so numbers are always good too. You have me half tempted to perform the comparison myself but I'll gladly except your results instead Also you need to qualify what chipset you're refering to because some are actualy using the NF2 IGP and no amount of higher clockspeed will compensate for the memory bandwidth the IGP requires to perform it's best.

 

[Jeff7181]

Originally posted by: DAPUNISHER
Where are the comparative benchies at? Thugsrook and the gang have spoiled us, we like to see the results For me I prefer screenshots of everything but I understand many are still on dial up so numbers are always good too. You have me half tempted to perform the comparison myself but I'll gladly except your results instead Also you need to qualify what chipset you're refering to because some are actualy using the NF2 IGP and no amount of higher clockspeed will compensate for the memory bandwidth the IGP requires to perform it's best.

When I get my new RAM I'll be doing some benchmarks so I'll show you then... but the last time I ran them to compare for myself I never saved any screen shots. Although I do remember the numbers... my current RAM and CPU score is in my profile... drop the CPU down to 2 Ghz, it score 6500 and 5500, drop it down to 2 Ghz with a 166 FSB, CPU score was 6200 and ram was 5200. I'll provide evidence when I get my new RAM.

 

[jjyiz28]

Originally posted by: Jeff7181
Lately I hear a lot of people saying running your memory a-sync hurts performance, and I agree with that half the time. I agree that running your memory faster than your FSB is pointless because Athlons don't need that extra memory bandwidth. However... if you can't increase your multiplier past 12.5, and you have PC2700 RAM... if you want to overclock much, running a-sync is your only option to get that extra clock speed.

While running in sync provides better performance at equal clock speeds, a 2.2 Ghz Athlon XP with a 220 Mhz FSB, and 166 Mhz RAM will perform better than a 2.0 Ghz Athlon XP with a 166 Mhz FSB and 166 Mhz RAM.
The Athlon XP isn't starved for memory bandwidth like the P4 is, so I see no problem running things a-sync in order to achieve a higher clock speed.

I've tested this myself... and even at similar clock speeds, the faster FSB provides better performance than keeping it in sync with the memory.

Of course, if you are able to run things in sync, you should do it, but if you're stuck with slower RAM that doesn't overclock, running a-sync is your only option, and in my opinion and experience, it's not as detrimental to performance as some people have been eluding to.

well of course, a 2.2 will perform better async than a 2.0 sync. think about the pentiums, with 533 bus, and running lets say a ddr333. does anyone in there right mind think dropping their cpu fsb to 333 to match their memory will be better performance than keeping cpu bus at 533 and running memory at 333? you're stating the obvious.

 

[Jeff7181]

Originally posted by: jjyiz28

Originally posted by: Jeff7181
Lately I hear a lot of people saying running your memory a-sync hurts performance, and I agree with that half the time. I agree that running your memory faster than your FSB is pointless because Athlons don't need that extra memory bandwidth. However... if you can't increase your multiplier past 12.5, and you have PC2700 RAM... if you want to overclock much, running a-sync is your only option to get that extra clock speed.

While running in sync provides better performance at equal clock speeds, a 2.2 Ghz Athlon XP with a 220 Mhz FSB, and 166 Mhz RAM will perform better than a 2.0 Ghz Athlon XP with a 166 Mhz FSB and 166 Mhz RAM.
The Athlon XP isn't starved for memory bandwidth like the P4 is, so I see no problem running things a-sync in order to achieve a higher clock speed.

I've tested this myself... and even at similar clock speeds, the faster FSB provides better performance than keeping it in sync with the memory.

Of course, if you are able to run things in sync, you should do it, but if you're stuck with slower RAM that doesn't overclock, running a-sync is your only option, and in my opinion and experience, it's not as detrimental to performance as some people have been eluding to.

well of course, a 2.2 will perform better async than a 2.0 sync. think about the pentiums, with 533 bus, and running lets say a ddr333. does anyone in there right mind think dropping their cpu fsb to 333 to match their memory will be better performance than keeping cpu bus at 533 and running memory at 333? you're stating the obvious.

Actually, it's not truely 533, that's 133 quad pumped, which is effectively 533, but it's not the same. You're comparing Intel technology to my post about AMD technology... you can't do that... AMD and Intel use completely different ways to achieve performance, so what's true for one, is not necessarily true for the other.

 

[DAPUNISHER]

Originally posted by: Jeff7181

Originally posted by: jjyiz28

Originally posted by: Jeff7181
Lately I hear a lot of people saying running your memory a-sync hurts performance, and I agree with that half the time. I agree that running your memory faster than your FSB is pointless because Athlons don't need that extra memory bandwidth. However... if you can't increase your multiplier past 12.5, and you have PC2700 RAM... if you want to overclock much, running a-sync is your only option to get that extra clock speed.

While running in sync provides better performance at equal clock speeds, a 2.2 Ghz Athlon XP with a 220 Mhz FSB, and 166 Mhz RAM will perform better than a 2.0 Ghz Athlon XP with a 166 Mhz FSB and 166 Mhz RAM.
The Athlon XP isn't starved for memory bandwidth like the P4 is, so I see no problem running things a-sync in order to achieve a higher clock speed.

I've tested this myself... and even at similar clock speeds, the faster FSB provides better performance than keeping it in sync with the memory.

Of course, if you are able to run things in sync, you should do it, but if you're stuck with slower RAM that doesn't overclock, running a-sync is your only option, and in my opinion and experience, it's not as detrimental to performance as some people have been eluding to.

well of course, a 2.2 will perform better async than a 2.0 sync. think about the pentiums, with 533 bus, and running lets say a ddr333. does anyone in there right mind think dropping their cpu fsb to 333 to match their memory will be better performance than keeping cpu bus at 533 and running memory at 333? you're stating the obvious.

Actually, it's not truely 533, that's 133 quad pumped, which is effectively 533, but it's not the same. You're comparing Intel technology to my post about AMD technology... you can't do that... AMD and Intel use completely different ways to achieve performance, so what's true for one, is not necessarily true for the other.

The NF2 platform in particular is the chipset this should be concerning as generally running synch results in better performance due to the latency hit the NF2 takes running asynch. What Jeff is saying is that enough clockspeed can overcome the benefits of synch mode given the circumstances he describes in his first post. While I can accept his assertion I just wanted to point out that this would not apply to those who do rely on the NF2 IGP for graphics because it needs the DC-DDR mode and all the bandwidth it can get so higher clockspeed is less important to it's performance than it appears to be for the SPP chipset.

 

[jjyiz28]

Originally posted by: Jeff7181

Originally posted by: jjyiz28

Originally posted by: Jeff7181
Lately I hear a lot of people saying running your memory a-sync hurts performance, and I agree with that half the time. I agree that running your memory faster than your FSB is pointless because Athlons don't need that extra memory bandwidth. However... if you can't increase your multiplier past 12.5, and you have PC2700 RAM... if you want to overclock much, running a-sync is your only option to get that extra clock speed.

While running in sync provides better performance at equal clock speeds, a 2.2 Ghz Athlon XP with a 220 Mhz FSB, and 166 Mhz RAM will perform better than a 2.0 Ghz Athlon XP with a 166 Mhz FSB and 166 Mhz RAM.
The Athlon XP isn't starved for memory bandwidth like the P4 is, so I see no problem running things a-sync in order to achieve a higher clock speed.

I've tested this myself... and even at similar clock speeds, the faster FSB provides better performance than keeping it in sync with the memory.

Of course, if you are able to run things in sync, you should do it, but if you're stuck with slower RAM that doesn't overclock, running a-sync is your only option, and in my opinion and experience, it's not as detrimental to performance as some people have been eluding to.

well of course, a 2.2 will perform better async than a 2.0 sync. think about the pentiums, with 533 bus, and running lets say a ddr333. does anyone in there right mind think dropping their cpu fsb to 333 to match their memory will be better performance than keeping cpu bus at 533 and running memory at 333? you're stating the obvious.

Actually, it's not truely 533, that's 133 quad pumped, which is effectively 533, but it's not the same. You're comparing Intel technology to my post about AMD technology... you can't do that... AMD and Intel use completely different ways to achieve performance, so what's true for one, is not necessarily true for the other.

ok, ill keep it to amd chips. lets say i have a barton 3000+ running 333 fsb, and have ddr226. you are saying that running the chip at 333 fsb rather than underclocking it to run in sync with memory at 266 will be faster. no sh!t. i am agreeing with you. just that you are stating the obvious.

 

[Jeff7181]

Everybody keeps saying "run it in sync" ... and they seem to ignore the fact that sometimes running things in sync means running things slower. So yes, I am stating the obvious, but some people need to hear the obvious. As DAPUNISHER said, I'm stressing that if you run things a-sync to allow you to overclock higher, a higher overclock will outweigh the benefits of running in sync.

Oh, by the way DAPUNISHER, here's some benchmark results from my new RAM. I didn't feel like doing assloads of testing, so I posted the relevant info.

 

[Cerb]

Well of course 220/166 is faster than 166/166!
The idea is for you to run it at 220/220!
Most people are running the memory faster than the FSB in async, which often does hurt performance--depending on the chipset and benchmark being used.

 

[pspada]

What's with the 220Mhz? :frown:

 

[Jeff7181]

Originally posted by: Cerb
Well of course 220/166 is faster than 166/166!
The idea is for you to run it at 220/220!
Most people are running the memory faster than the FSB in async, which often does hurt performance--depending on the chipset and benchmark being used.

As jjyiz28 said... no sh!t... geeze... you people love to try to proove somebody wrong... go read my post again.

 

[pspada]

Originally posted by: pspada
What's with the 220Mhz? :frown:

I'm serious, where'd the 220Mhz FSB come from? I can see comparing 200fsb to 166fsb, but did you pull the 220 out of your a$$?

 

[amoralist]

for programs that don't use 100% of cpu, sync is faster than async. that is 220/166 isn't any faster than 166/166.
in fact 166/166 is faster... that is if you are not doing winrar or divx encoding (but are more concerned with
boot up times and program startup) then sync with slower proc is better...

 

[jjyiz28]

Originally posted by: pspada

Originally posted by: pspada
What's with the 220Mhz? :frown:

I'm serious, where'd the 220Mhz FSB come from? I can see comparing 200fsb to 166fsb, but did you pull the 220 out of your a$$?

pulled out of his a$$. just used to prove his point. =|

 

[jjyiz28]

Originally posted by: amoralist
for programs that don't use 100% of cpu, sync is faster than async. that is 220/166 isn't any faster than 166/166.
in fact 166/166 is faster... that is if you are not doing winrar or divx encoding (but are more concerned with
boot up times and program startup) then sync with slower proc is better...

what program doesn't use 100 percent of your cpu?? don't they all ? if not, there would be no performace gain going from a 300mhz cpu to 500 mhz cpu (hypothetical) which wouldn't make sense

 

[Whitedog]

Originally posted by: jjyiz28

Originally posted by: amoralist
for programs that don't use 100% of cpu, sync is faster than async. that is 220/166 isn't any faster than 166/166.
in fact 166/166 is faster... that is if you are not doing winrar or divx encoding (but are more concerned with
boot up times and program startup) then sync with slower proc is better...

what program doesn't use 100 percent of your cpu?? don't they all ? if not, there would be no performace gain going from a 300mhz cpu to 500 mhz cpu (hypothetical) which wouldn't make sense

You people are talking about stuff way over your heads here... This is the funniest thread I've read in a while.

Maybe you should talk about heat sinks or something...

 

[Jeff7181]

Originally posted by: jjyiz28

Originally posted by: pspada

Originally posted by: pspada
What's with the 220Mhz? :frown:

I'm serious, where'd the 220Mhz FSB come from? I can see comparing 200fsb to 166fsb, but did you pull the 220 out of your a$$?

pulled out of his a$$. just used to prove his point. =|

Are you talking about me? I was running my FSB at 221... I use to have that posted in my profile, but not anymore cause I'm not running at 221 now that I have my new RAM

 

[Cerb]

Give it up, Jeff7181.

 

[Jeff7181]

Originally posted by: Cerb
Give it up, Jeff7181.

What the hell are you talking about?

I swear... this is one of the best message boards I've found, but it's still full of some really strange people with HUGE egos. A lot of you need to take yourselves a little less seriously... I was pointing out that running a-sync does have it's advantages... like when you're stuck using slow RAM and can't adjust the multiplier any higher and you're left with just tweaking the FSB to overclock. Then a bunch of you get your shorts in a knot, try and say I'm stating the obvious. Geeze... I started this post with the intent to inform others who are in the same situation I was in so they don't think they're stuck using a 166 Mhz FSB if their RAM won't go higher than 166. But damn, look how your egos and attitudes have turned it into an argument over where "220" came from.

 

[WarCon]

Originally posted by: Jeff7181

Originally posted by: Cerb
Give it up, Jeff7181.

What the hell are you talking about?

I swear... this is one of the best message boards I've found, but it's still full of some really strange people with HUGE egos. A lot of you need to take yourselves a little less seriously... I was pointing out that running a-sync does have it's advantages... like when you're stuck using slow RAM and can't adjust the multiplier any higher and you're left with just tweaking the FSB to overclock. Then a bunch of you get your shorts in a knot, try and say I'm stating the obvious. Geeze... I started this post with the intent to inform others who are in the same situation I was in so they don't think they're stuck using a 166 Mhz FSB if their RAM won't go higher than 166. But damn, look how your egos and attitudes have turned it into an argument over where "220" came from.

Yeah there are alot of people with ego issues.......

Anyway, there are also alot of really well informed individuals that take the time to do the testing to show things like the benefits of certain memory timings (even if they are miniscule). They put together a plan and document it, so the rest of us don't have to go through all that to find the break even points of doing certain things like what you suggest. I am not faulting your logic, because I am sure at some point the large hit you take from not running in sync can be overcome by sheer horsepower of processor. What is probably getting you all this grief is that you aren't offering even one set of comparison data to back your assertion.

Anyway, clock speed for clock speed, you do take a huge performance hit running async. On synthetic benchmarks, even when the processor is 120 mhz higher, 200 * 11 (async) (no this is not your example, this came from when I just set up and tested my friends new rig) versus 166 * 12.5 (sync) the memory still had a nearly 500 point deficit in Sandra compared to the latter. You won't notice this much until you run data hungry apps like video or audio conversion, so for your average computer user maybe its not a big deal.

 

[Jeff7181]

If I had posted benchmark results the ego's would have flamed me saying synthetic benchmarks are meaningless, and real world performance will still take a hit.

 

[Cerb]

Originally posted by: Jeff7181

Originally posted by: Cerb
Give it up, Jeff7181.

What the hell are you talking about?

Stuff like

what program doesn't use 100 percent of your cpu?? don't they all ? if not, there would be no performace gain going from a 300mhz cpu to 500 mhz cpu (hypothetical) which wouldn't make sense

I swear... this is one of the best message boards I've found, but it's still full of some really strange people with HUGE egos. A lot of you need to take yourselves a little less seriously... I was pointing out that running a-sync does have it's advantages... like when you're stuck using slow RAM and can't adjust the multiplier any higher and you're left with just tweaking the FSB to overclock.

I need a new mobo...$57 staring me in the face at Newegg.

Then a bunch of you get your shorts in a knot, try and say I'm stating the obvious. Geeze... I started this post with the intent to inform others who are in the same situation I was in so they don't think they're stuck using a 166 Mhz FSB if their RAM won't go higher than 166. But damn, look how your egos and attitudes have turned it into an argument over where "220" came from.

Like

pulled out of his a$$. just used to prove his point. =|

Faster FSB is great--I'd just rather get RAM that can take it. The FSB is of course going to help, since...it's the FSB. Not "only goes to memoy bus," but, "goes to everything outside of the chip" bus.

 

[Whitedog]

Jeff7181, don't let these guys get to you. Best thing to do is pay no attention to them. A lot of the guys that post crap are just teen-aged wanna-be techs that know how to install Windows...etc. Then there's the group of 1 c.i. minds with the 1000 c.i. ego's that are just.... well, they think they're just sooo smart.

The majority of the people are genuine so don't be discouraged by the few bad apples.

[edit]I'm not insinuating that the people posting in this thread fit into the above descrition...[/edit]

You're posts are good info. And just so you know, I run my system at 66% memory to CPU ratio. I've done hours upon hours of self testing and found this to give the best performance overall.

The key is to run your memory at 6-2-2-2. It's much better to run your memory at faster timings in async mode than slow the memory down just so you can run it synch. I've read post where guys would set their memory to like 8-4-4-3 just to get it to run at 220/220 or something like that. LMAO!! I think all they are interested in is "hitting the high clock speed numbers" and not "real performance".

Anyway, in my system, I've got some crappy PC2700 that only runs at 7-3-3-2.5 at 166mhz, so I run it at 140MHz at 6-2-2-2. All benchmarks I run result in much higher numbers at these settings. I've got nForce2 and a t-bred "b" setup, so I can experiment with all the settings.

Current settings are 212 FSB - Mem @ 66% = 140MHz @ 6-2-2-2
CPU Feq is 2120 (x10 fsb).

I compared benchmarks with running my system synch @ 2158 (166x13 fsb and mem (XP2700+ settings)) and ALL the scores were higher with the 212/140 settings. More FPS is games and everything.

So rest easy knowing that running "synch" does Not always result in better performance.

People will surely argue with this, but that applies to those with the faster memory. Sure, if you can run your memory sync at 6-2-2-2 settings it is Absolutly faster!

Whatever. Have a nice day!

 

[pspada]

Ah, i see what you are getting at. But you are right, your situation only applies since you have crappy, er, lets instead say, slower RAM. My RAM will do 6-2-2-2 without a problem even when set to 494Mhz. But since my older KT400 chipset based mobo does not like a FSB above 166, I'm running it at 333 so it is insync with the FSB. When I instead run the RAM at or near 400Mhz, it induces wait states that cause the system to run perceptibly slower than when the RAM speed is set lower to match the FSB speed.

 

[DAPUNISHER]

Originally posted by: pspada
Ah, i see what you are getting at. But you are right, your situation only applies since you have crappy, er, lets instead say, slower RAM. My RAM will do 6-2-2-2 without a problem even when set to 494Mhz. But since my older KT400 chipset based mobo does not like a FSB above 166, I'm running it at 333 so it is insync with the FSB. When I instead run the RAM at or near 400Mhz, it induces wait states that cause the system to run perceptibly slower than when the RAM speed is set lower to match the FSB speed.

I agree with your conclusion on this, it applies to his particular setup and reflects the conclusions of This evaluation

From our results, the most important factor in memory bandwidth is the speed of the memory clock. This would suggest a certain desirability to sacrifice the other memory timings in hopes of pushing the memory speed higher. ***this has been my results as well*** However, as our results revealed, the speed of the memory combined with CAS latency has the most affect on the overclockability of a memory stick (our test memory would not run at the more aggressive speeds and CAS latencies). The memory timings on our particular setup that had the most impact on performance involved setting the Bank Interleave to 4 Way, decreasing the DRAM command rate to 1T, and decreasing tRCD to 1T. Just as other websites have suggested gains from certain memory timings (the Engineers at OCZ suggest that tRAS at 3T or 4T have a very significant increase on performance), I should remind you that these are my personal results from my test setup, and my particular combination of hardware created these "patterns."

What works best really depends on the RAM you have and your board's like or dislike of it. I use cheapo Apacer 3200DDR CL3 with TCC4 chips and in my P4PE/L it'll do 443DDR@2.5-3-3-6 in my NF2 boards it simply won't run tighter timings than it's specs of 3-3-3-7 and if I run asynch to test it tops out@425DDR with those timings even in SC-DDR mode.