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

[pspada]

For reference my RAM is OCZ Performance Series Rev. 2 PC3200, and it is rated for CL2 at 200/400.

 

[rogue1979]

I think every system is different and everybody should run benchmarks in the applications that they use most. Then use the data gathered to determine the fastest settings for your own system. I am running an Nforce2 at 200MHz synch for 400MHz DDR. I have to slow my timings down to 7-3-3-2.5 to hit 200MHz synched. It still gives me better framerates and 3Dmark scores than running 184MHz at 6-2-2-2.5. This is using the multiplier to run my cpu at 2300MHz at both fsb speeds. Either of these settings give me better performance than any asynched combos. But my 1700+ is unlocked, so cpu speed is not a factor. I put the emphasis on gaming, my Ti4600 is almost reaching 15K in 3DMark 2001. It is possible that tighter memory timings and a slower fsb could be faster in other apps, but like I said, gaming is my priority in the speed/performance arena.

 

[Whitedog]

My RAM will do 6-2-2-2 without a problem even when set to 494Mhz.

Eh? Come again? 494? That's 247Mhz - DDR494.... You have PC3200 that will 247Mhz @ 6-2-2-2??? Bull_f04k1n_$51t... To my knowledge, such memory doesn't exist.

LMAO! Really. That's a good one. :beer:

 

[Whitedog]

OK then... for arguments sake... Let's take the following scenario

Setting 1:
FSB = 150
Mem = 150 (100%) @ 6-2-2-2
MHz = 2100 (150x14)

Setting 2:
FSB = 200
Mem = 150 (75%) @ 6-2-2-2
MHz = 2100 (200x10.5)

Setting 3:
FSB = 200
Mem = 200 (100%) @ 8-3-3-3
MHz = 2100 (200x10.5)

My memory won?t do 200 period, so I can?t test this? I?d like to see someone test these settings and post the results.

My guess is #2 will result in most performance. Just a guess.

Cheers :beer:

 

[pspada]

Um, what's with the fsb/memory set to 150Mhz? Wouldn't that be 166? Besides, my CL2 memory is rated at 6-2-2-2 @ 400Mhz. And I don't change those settings when I change the FSB to 247, so it's booting with these settings at 494Mhz. Of course Winblows won't load or run at that speed, but I know that my AGP and PCI buses crap out when overclocked much, so I don't believe these is any problem, even at that speed, with the memory, or possibly even the CPU.

But I'm willing to try it on a NF2 mobo, if you'd like to lend me one.....

 

[Jeff7181]

Originally posted by: Whitedog
OK then... for arguments sake... Let's take the following scenario

Setting 1:
FSB = 150
Mem = 150 (100%) @ 6-2-2-2
MHz = 2100 (150x14)

Setting 2:
FSB = 200
Mem = 150 (75%) @ 6-2-2-2
MHz = 2100 (200x10.5)

Setting 3:
FSB = 200
Mem = 200 (100%) @ 8-3-3-3
MHz = 2100 (200x10.5)

My memory won?t do 200 period, so I can?t test this? I?d like to see someone test these settings and post the results.

My guess is #2 will result in most performance. Just a guess.

Cheers :beer:

I'd test it, but I don't have multipliers above 12.5 available.

 

[jjyiz28]

Originally posted by: pspada
Um, what's with the fsb/memory set to 150Mhz? Wouldn't that be 166? Besides, my CL2 memory is rated at 6-2-2-2 @ 400Mhz. And I don't change those settings when I change the FSB to 247, so it's booting with these settings at 494Mhz. Of course Winblows won't load or run at that speed, but I know that my AGP and PCI buses crap out when overclocked much, so I don't believe these is any problem, even at that speed, with the memory, or possibly even the CPU.

But I'm willing to try it on a NF2 mobo, if you'd like to lend me one.....

i think its your fsb/memory ratio that determines what speed you run you ddr400 ram. look in your bios.

whitedog, in my opinion i think #3 will result in better performance for most of the benchmarks. of course, this is just a guess. i too would like to see the results for #2 vs. #3.

 

[Whitedog]

Originally posted by: Jeff7181

Originally posted by: Whitedog
OK then... for arguments sake... Let's take the following scenario

Setting 1:
FSB = 150
Mem = 150 (100%) @ 6-2-2-2
MHz = 2100 (150x14)

Setting 2:
FSB = 200
Mem = 150 (75%) @ 6-2-2-2
MHz = 2100 (200x10.5)

Setting 3:
FSB = 200
Mem = 200 (100%) @ 8-3-3-3
MHz = 2100 (200x10.5)

My memory won?t do 200 period, so I can?t test this? I?d like to see someone test these settings and post the results.

My guess is #2 will result in most performance. Just a guess.

Cheers :beer:

I'd test it, but I don't have multipliers above 12.5 available.

Then crank down the MultX. Instead of 10.5x200 and 14x150, do 9x200 and 12x150. Both will give you 1800MHz. Do all the rest of the settings as stated. Should result the same.

Um, what's with the fsb/memory set to 150Mhz? Wouldn't that be 166?

You can do it however you want. I was just using 150/200 as a good spread. I don't know what kind of mobo you've got, but the A7N8X will allow you to pick whatever speed you want in 1MHz inc's from 133-250 or something like that.

For what this is worth, here are some benchie numbers from my testing. Remember, my RAM is not the screamer 200MHz kind like you have (and the way it was put "since you have crappy, er, lets instead say, slower RAM." isn't very polite. If we want to talk about "CRAPPY", we could talk about the long list of computers in someones SIG... )
Anyway, using the "CRAPPY" PC2700 I have, here is what got me the best results. (I'll just post 2 configs.. the rest don't contribute to anything)

Why 168 and not 166? Because I wanted to get the MHz as close to the same as I could. 2268 and 2270 are pretty darn close.

Comparing sync vs. async

Config 1:
168fsb x 13.5 = 2268MHz
Memory 168MHz (100%) @7-3-3-2.5
Sync
CPU test total - 11846
MultiMedia Test total - 25918
Memory Test total - 5020

Config 2:
227fsb x 10 = 2270MHz
Memory 151MHz (66%) @ 6-2-2-2
ASync
CPU test total - 11867
MultiMedia Test total - 25963
Memory Test total - 5220

Not a lot of difference, and not a very scientific test... but you can see the CPU and MM scores about the same, however the memory is faster at 151MHz using the faster settings.

Running some FPS test in games resulted in even more of a spread as I believe the Increased FSB made an impact. Nothing I tested worked better on the "sync" config. Nothing.

"When" I get some fast ram as has been stated by some, I'll be able to run "sync" at desired speeds... but until then, my system runs better async as stated... as does others, probably.

Cheers :beer:

 

[Jeff7181]

Ok... I ran 3DMark2001 and PCMark2002...

220x10=2200 Mhz RAM = 75% (166 Mhz) @ 5-2-2-2 (lowest stable)
3DMark2001 @ defaults
12,084
PCMark2002 @ defaults
CPU - 6811
RAM - 5541
HDD - 1084

200x11=2200 Mhz RAM = 100% (200 Mhz) @ 7-3-3-2.5 (lowest stable)
3DMark2001 @ defaults
12,862
PCMark2002 @ defaults
CPU - 6795
RAM - 6198
HDD - 1072

Then since I had to relax RAM timings to run it at 200 Mhz, I tested it again at 166 Mhz with the same RAM timing as I used at 200 Mhz

220x10=2200 Mhz RAM = 75% (166 Mhz) @ 7-3-3-2.5
3dMark2001 @ defaults
11,967
PCMark2002 @ defaults
CPU - 6746
RAM - 5524
HDD - 1088

 

[Whitedog]

Interesting...

Makes me Wonder just WHAT do those benchies test? How can the RAM scores almost match for both the 166MHz settings? One is timed 5-2-2-2 and the other 7-3-3-2.5? There's no way they should score the same... or even Close.

hmmmmmm......

 

[Jeff7181]

I think PCMark pretty much just cares about memory bandwidth... maybe Sandra would provide a more accurate depiction. However, my 3DMark score was higher too with the 200 Mhz RAM setting... and that's all I really care about, game performance.

I neglected to mention that I also tested the FPS in Soldier of Fortune 2 with 4XAA and 4XAF. I saved the game at a point with a couple light sources in view, and rain, and here's the results...

220x10=2200 Mhz - RAM = 75% (166 Mhz) @ 5-2-2-2 (lowest stable)
FPS = 35

200x11=2200 Mhz RAM = 100% (200 Mhz) @ 7-3-3-2.5 (lowest stable)
FPS = 42

220x10=2200 Mhz RAM = 75% (166 Mhz) @ 7-3-3-2.5
FPS = 34

 

[DAPUNISHER]

Good job Jeff :beer: That's the meat we like to see! The in game FPS are far more pertinent than 3Dmark and Sandra IMO but they do give an apple to apple comparaison we can all easily absorb the data from.

 

[Jeff7181]

I just upgraded to the 44.03 Detonators, and now I get 46 FPS in that same saved game in SoF2. Hehe... talk about squeezing every ounce of performance =)

 

[tbates757]

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

Dude, that was really really sad...

 

[jjyiz28]

Originally posted by: tbates757

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

Dude, that was really really sad...

what the hell are you talking about?? you contribute nothing to this thread 'cept this?? what's sad?? enlighten me

 

[Whitedog]

Originally posted by: jjyiz28

Originally posted by: tbates757

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

Dude, that was really really sad...

what the hell are you talking about?? you contribute nothing to this thread 'cept this?? what's sad?? enlighten me

You're like, an angry person or something... chill dude.

I think he was referring to your knowlege of programs and you making such a bold statement about them. NO, FEW programs utilize 100% of the CPU. Most programs sit idle 99.9999% of the time. Look at your task manager. Sure, when they are executing code the CPU is in work, but then not always the processor is being used 100%.

People shouldn't really comment about stuff they don't even know what they are talking about really... and this thread is FULL of such comments. But that's what makes AT forums so unique. Quite a comical place sometimes.

 

[Whitedog]

Originally posted by: Jeff7181
I think PCMark pretty much just cares about memory bandwidth... maybe Sandra would provide a more accurate depiction. However, my 3DMark score was higher too with the 200 Mhz RAM setting... and that's all I really care about, game performance.

I neglected to mention that I also tested the FPS in Soldier of Fortune 2 with 4XAA and 4XAF. I saved the game at a point with a couple light sources in view, and rain, and here's the results...

220x10=2200 Mhz - RAM = 75% (166 Mhz) @ 5-2-2-2 (lowest stable)
FPS = 35

200x11=2200 Mhz RAM = 100% (200 Mhz) @ 7-3-3-2.5 (lowest stable)
FPS = 42

220x10=2200 Mhz RAM = 75% (166 Mhz) @ 7-3-3-2.5
FPS = 34

This is Great!! The kind of stuff we need to see... It just proves to me how worthless synthetic bench marks can be. From now on, I'm going to just test performance where it REALLY counts... GAMES!

Thanks Jeff. I'll be setting my system to run sync this evening. I'm SURE there are instances where the faster FSB would be better, but who cares. Games is what we tweak for anyway, and if running sync makes my games run better, than so be it.

 

[pspada]

Originally posted by: jjyiz28

Originally posted by: pspada
Um, what's with the fsb/memory set to 150Mhz? Wouldn't that be 166? Besides, my CL2 memory is rated at 6-2-2-2 @ 400Mhz. And I don't change those settings when I change the FSB to 247, so it's booting with these settings at 494Mhz. Of course Winblows won't load or run at that speed, but I know that my AGP and PCI buses crap out when overclocked much, so I don't believe these is any problem, even at that speed, with the memory, or possibly even the CPU.

But I'm willing to try it on a NF2 mobo, if you'd like to lend me one.....

i think its your fsb/memory ratio that determines what speed you run you ddr400 ram. look in your bios.

whitedog, in my opinion i think #3 will result in better performance for most of the benchmarks. of course, this is just a guess. i too would like to see the results for #2 vs. #3.

My m/b allows you to adjust the memory speed, FSB, and FSB to memory ratio independently of each other.

 

[jjyiz28]

Originally posted by: Whitedog

Originally posted by: jjyiz28

Originally posted by: tbates757

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

Dude, that was really really sad...

what the hell are you talking about?? you contribute nothing to this thread 'cept this?? what's sad?? enlighten me

You're like, an angry person or something... chill dude.

I think he was referring to your knowlege of programs and you making such a bold statement about them. NO, FEW programs utilize 100% of the CPU. Most programs sit idle 99.9999% of the time. Look at your task manager. Sure, when they are executing code the CPU is in work, but then not always the processor is being used 100%.

People shouldn't really comment about stuff they don't even know what they are talking about really... and this thread is FULL of such comments. But that's what makes AT forums so unique. Quite a comical place sometimes.

i see, thanks a lot. i was thinking about games.

 

[Jeff7181]

Originally posted by: Whitedog

Originally posted by: Jeff7181
I think PCMark pretty much just cares about memory bandwidth... maybe Sandra would provide a more accurate depiction. However, my 3DMark score was higher too with the 200 Mhz RAM setting... and that's all I really care about, game performance.

I neglected to mention that I also tested the FPS in Soldier of Fortune 2 with 4XAA and 4XAF. I saved the game at a point with a couple light sources in view, and rain, and here's the results...

220x10=2200 Mhz - RAM = 75% (166 Mhz) @ 5-2-2-2 (lowest stable)
FPS = 35

200x11=2200 Mhz RAM = 100% (200 Mhz) @ 7-3-3-2.5 (lowest stable)
FPS = 42

220x10=2200 Mhz RAM = 75% (166 Mhz) @ 7-3-3-2.5
FPS = 34

This is Great!! The kind of stuff we need to see... It just proves to me how worthless synthetic bench marks can be. From now on, I'm going to just test performance where it REALLY counts... GAMES!

Thanks Jeff. I'll be setting my system to run sync this evening. I'm SURE there are instances where the faster FSB would be better, but who cares. Games is what we tweak for anyway, and if running sync makes my games run better, than so be it.

Yes... in my origional post, I was saying that sometimes the speed of a person's RAM prevents them from overclocking the CPU more, in which case it is better to run the RAM a-sync in order to attain a higher CPU overclock which makes up for the RAM running out of sync. But if your RAM can keep up, by all means, run it in sync.

 

[pspada]

Ah so! Jeff, you left out an interesting test question: what if you run the processor at 166, and run the memory at 200? What (crummy synthetic) benchmark and (good real world) game performance would you get in comparison with these current test scores?

And this then poses another question - is game performance enhanced by the extra cache of the Barton chips over the actual default clockspeed of (for example) a 2600+? How about when both chips are running at the same clockspeed?

I wish you were in San Diego and we could try this with my ram.... :sun:

 

[Whitedog]

Originally posted by: Jeff7181

Originally posted by: Whitedog

Originally posted by: Jeff7181
I think PCMark pretty much just cares about memory bandwidth... maybe Sandra would provide a more accurate depiction. However, my 3DMark score was higher too with the 200 Mhz RAM setting... and that's all I really care about, game performance.

I neglected to mention that I also tested the FPS in Soldier of Fortune 2 with 4XAA and 4XAF. I saved the game at a point with a couple light sources in view, and rain, and here's the results...

220x10=2200 Mhz - RAM = 75% (166 Mhz) @ 5-2-2-2 (lowest stable)
FPS = 35

200x11=2200 Mhz RAM = 100% (200 Mhz) @ 7-3-3-2.5 (lowest stable)
FPS = 42

220x10=2200 Mhz RAM = 75% (166 Mhz) @ 7-3-3-2.5
FPS = 34

This is Great!! The kind of stuff we need to see... It just proves to me how worthless synthetic bench marks can be. From now on, I'm going to just test performance where it REALLY counts... GAMES!

Thanks Jeff. I'll be setting my system to run sync this evening. I'm SURE there are instances where the faster FSB would be better, but who cares. Games is what we tweak for anyway, and if running sync makes my games run better, than so be it.

Yes... in my origional post, I was saying that sometimes the speed of a person's RAM prevents them from overclocking the CPU more, in which case it is better to run the RAM a-sync in order to attain a higher CPU overclock which makes up for the RAM running out of sync. But if your RAM can keep up, by all means, run it in sync.

PLUS, you should have added "When your multiplier is limited to 12.5x" as is your case.

In my case, I'm not limited to MultX (2100+), so I can just keep cranking up my MultX.

Don't you have a 2500+ Barton? And you're limited to 12.5x with the A7N8X board? Weird.

 

[Jeff7181]

Originally posted by: pspada
Ah so! Jeff, you left out an interesting test question: what if you run the processor at 166, and run the memory at 200? What (crummy synthetic) benchmark and (good real world) game performance would you get in comparison with these current test scores?

And this then poses another question - is game performance enhanced by the extra cache of the Barton chips over the actual default clockspeed of (for example) a 2600+? How about when both chips are running at the same clockspeed?

I wish you were in San Diego and we could try this with my ram.... :sun:

I did what you suggested...

166x12.5=2083 Mhz RAM = 166 @ 7-3-3-2.5
3DMark2001
11483

PCMark2002
CPU - 5786
RAM - 5318
HDD - 1092

FPS in SoF2 at my saved game point
32

166x12.5=2083 Mhz RAM = 200 @ 7-3-3-2.5
3DMark2001
11480

PCMark2002
CPU - 5724
RAM - 6105
HDD - 1046

FPS in SoF2 at my saved game point
33-34 (fluctuated very fast between the two)

 

[Whitedog]

I can see running the RAM @ 166 and the FSB @ 200 ('cause your RAM may not be fast enough to run @ 200), but I certainly don't see the point in running the FSB @ 166 with the RAM @ 200? I don't know why he even suggested doing that? There's no point to that.

 

[Jeff7181]

Originally posted by: Whitedog
I can see running the RAM @ 166 and the FSB @ 200 ('cause your RAM may not be fast enough to run @ 200), but I certainly don't see the point in running the FSB @ 166 with the RAM @ 200? I don't know why he even suggested doing that? There's no point to that.

As you can see, 3DMark score went down, but FPS increased by 1-2. So maybe it's not so bad afterall. I think it would make more of a difference if Athlon XP's were as bandwidth hungry as P4's, but they're not, so latency is more of an issue than maximum bandwidth.