P5WD2-P low bandwith?

MikeyLikesItSI · Oct 4, 2005

OK, i expected higher sandra memory bandwith benches then this... 6080/6067

63% efficiency

280FSB, 3920mhz (P4 630), DDR2 @ 700 (4:5), timings 4-4-4-10

Does that seem right? While i'm on sandra benches, file system benchmark is 92 mb/s. Thats with a 2 seagate baracuda 80gig 7200.7 drives on RAID 0. sound about right?

d94 · Oct 4, 2005

MikeyLikesItSI said:
OK, i expected higher sandra memory bandwith benches then this... 6080/6067

63% efficiency

280FSB, 3920mhz (P4 630), DDR2 @ 700 (4:5), timings 4-4-4-10

Does that seem right? While i'm on sandra benches, file system benchmark is 92 mb/s. Thats with a 2 seagate baracuda 80gig 7200.7 drives on RAID 0. sound about right?

dude that is low
when i run 300fsb 4-4-4-12 i get bout 8k
im not sure what to say...

MikeyLikesItSI · Oct 4, 2005

d94 said:
dude that is low
when i run 300fsb 4-4-4-12 i get bout 8k
im not sure what to say...

What speed are you running your memory at?

RangerXLT8 · Oct 4, 2005

MikeyLikesItSI said:
OK, i expected higher sandra memory bandwith benches then this... 6080/6067

63% efficiency

280FSB, 3920mhz (P4 630), DDR2 @ 700 (4:5), timings 4-4-4-10

Does that seem right? While i'm on sandra benches, file system benchmark is 92 mb/s. Thats with a 2 seagate baracuda 80gig 7200.7 drives on RAID 0. sound about right?

That looks about right to me. When I run DDR2 660 3-2-2-6, 265FSB, 3.71ghz I get roughly 6500, 80 % effeciency.

Nexus Realized · Oct 4, 2005

I'm not familiar with Intels systems in any way. Is the effeciecy level always that low on them? On my A64 when running the bandwidth benchmark it gives me a 93% effieciency level. :shrug:

MikeyLikesItSI · Oct 4, 2005

RangerXLT8 said:
That looks about right to me. When I run DDR2 660 3-2-2-6, 265FSB, 3.71ghz I get roughly 6500, 80 % effeciency.

exactly, your running 200 mhz slower, and 15 fsb slower, and running your memory slower, but scoring 500 points higher, and getting 17% higher efficiency then me....

Nexus Realized said:
I'm not familiar with Intels systems in any way. Is the effeciecy level always that low on them? On my A64 when running the bandwidth benchmark it gives me a 93% effieciency level.

Show off!

Yeah, A64's integrated memory controller make them much more efficient.

MikeyLikesItSI · Oct 4, 2005

OK. changed the multiplier back to 15, lowered the FSB to 267, left the timings and DDR2 speed the same. Now the efficiency is up to 75% and a bench of 6398/6396.... why would backing down the FSB (resulting in about 85 mhz more clock speed(14x280 vs 15x267) yield such a jump in score and effiecency? i've locked hyperpath enabled, and performance mode to turbo.

d94 · Oct 4, 2005

im running my ram @ 750mhz

BigStan · Oct 5, 2005

Seem like my foggy brain remembers having an issue like this once a while back. I think I increased the voltage. But I'm not sure exactly where. As counter inutitive as it seems, I think it was the cpu voltage. But I could be wrong. it might have been the mem voltage.

Also, run Everst Mem benchmarks at both multis and see if you are really losing speed.

MikeyLikesItSI · Oct 5, 2005

I did discover why the efficiency numbers fluctuate so badly. At 280 x 14, sandra reports a fsb of 300, and the memory follows suit.

It still doesn't explain why it goes up 300 points for simply lowering the FSB and changing to 15x.

BigStan · Oct 5, 2005

Have you disabled EC1? Also try disabling Hyper Path3.

When running the 14 multi by using the Minimum option of SpeedStep, Sandra and a couple of other proggies my not work right. AI Booster and CPUz should give you a good readings on your fsb and multi.

You are the first person to repot high Sandra scores at lower fsb. There just has to be a setting somewhere that needs to be adjusted. Did you try a little more voltage?

orionlion82 · Oct 5, 2005

my rig suffered when not running 1:1.
yeah, theres been alot said about how things are supposed to happen, before the point of deminishing returns, but ive never been boxed in by such notions.
i tuned my rig for memory bandwith over a period of days, and the absolute best i ever got was 1:1 with an FSB bump. i was folding these things that are called QMD's - they are all about bandwith. nothing ever beat 1:1... memtest, sandra, whatever...
if youcrave bandwith, then 4:5 looks like the wrong place to me

d94 · Oct 5, 2005

actually, if i run 1:1 my computer gets funki..4:5 or 3:4 runs NICE and i get high bandwidth

BigStan · Oct 10, 2005

MikeyLikesItSI - I'm seeing some other posts of ppl taking a performance hit when using the 14 multi with 6xx procs. Did you ever find a solution?

MikeyLikesItSI · Oct 10, 2005

Bigstan: I haven't done much further testing. In fact i have another problem to overcome first.

I loaded the 0519 bios the other day and restarted without a problem. I started playing some NFSU 2 and it reset. no big deal i had adjusted vcore lower so i put it back to where it was. I booted back up, installed another game that i wanted to try out, the installer requested to restart the computer after the game was done installing, and it hung on the bios screen, when its trying to start the USB controller.... i powered on and off a few times and it didn't change. Unfortunatly i had to run out the door and haven't been home since. I'm gonna pull the reset jumper and see if that solves anything.... anybody seen this problem before?

After that is resolved, i want to do some more testing with 14x multi. I know you can't compare the effiecency percentage due to sandra reporting the wrong FSB, so it will have to be a output/numbers comparison only.

MikeyLikesItSI · Oct 11, 2005

Ok, More results. I tried 295x14, gave me 5900's!! i can't believe how sandra is falling off with higher FSB. I went back to 267x15 (highest stable clock), and i'm getting 6110. Memory right now is at 2:3 and it puts out the same numbers as 4:5 does.... Maybe sandra just isn't compatible with this chipset? only reason i say this, is because when using the 14x multi, sandra reports higher then actual FSB.

Sentential · Oct 11, 2005

What version are you using? The old one wont read the newer chipsets properly

Pinky · Oct 11, 2005

Have you verified the results with 3mark05 at 290 versus 267? That may solve the "is sisoft" to blame question. If 3dMark05 chokes at 290 as it appears Sandra does, then that would solve the mystery.

BigStan · Oct 11, 2005

They are not really the same are they - 2:3 ratio of 267=400 and 4:5 of 295 is 368 right? Or did I do my math wrong again? Will your mem do 2:3 at 295? It's pretty good mem.

P5WD2-P low bandwith?

MikeyLikesItSI

Member

d94

$30 a phone

MikeyLikesItSI

Member

RangerXLT8

Member

Nexus Realized

Member

MikeyLikesItSI

Member

MikeyLikesItSI

Member

d94

$30 a phone

BigStan

Member

MikeyLikesItSI

Member

BigStan

Member

orionlion82

Member

d94

$30 a phone

BigStan

Member

MikeyLikesItSI

Member

MikeyLikesItSI

Member

Sentential

Contributing Member

Pinky

Member

BigStan

Member

Similar threads