The Sandra Memory Bandwidth Problem

[sww]

Did we ever figure this out? Did we just attribute it to poor BIOS optimizations? Or does no one care enough to research it further?

FYI: I am speaking of the problem where high FSB Athlon systems score very poorly on Sandra's Memory Bandwidth Bench. Abit boards seem to suffer from this the least, IMO, and Asus boards seem to suffer from it quite a bit. I'd really like to figure this one out.

------------------
Something like 37% of statistics are made up out of thin air.

[Ghettoweasel]

My 1520 tbird on a KT7A gets 635/697 with a 152FSB. At 153 it only goes to 655/706. I never did see the scores that were in the 700/800 range and I'm very skeptical that those are real.

------------------
1.2 tbird @ 1.52
Abit KT7A
256MB of Crucial PC133 @ 152 CAS2
Watercooling: Maze2, Cooling cube, Eheim1048
TNT2 Ultra
IBM 60GB ATA/100
Seagate 20GB ATA/66
Plexwriter 12/10/32A
Pioneer 16X DVD slot load
SCSI Zip
SB Live MP3+
Linksys 10/100

[Un4given]

Originally posted by Ghettoweasel:
My 1520 tbird on a KT7A gets 635/697 with a 152FSB. At 153 it only goes to 655/706. I never did see the scores that were in the 700/800 range and I'm very skeptical that those are real.

They are, but those are DDR boards. When you take a benchmark like Sandra that can stress a specific area of your system, and not have to worry about other system considerations, the scores tend to be much higher than the performance you see in real life.



------------------
Knight for the OC Crusaders

P3 700@945 (135x7.0 1.75V)
Asus CUV4X w/128MB PC150 (2-2-2)
Radeon 64MB VIVO 200/200
SBLive! Value
13.5GB Maxtor ATA100 (5400RPM)
Memorex 36X CD-ROM/TDK 8x4x32 CD-RW

[Ghettoweasel]

I beleive in the discussion that SWW was referring to was in regard to SDR scores in the 700/800 range and DDR scores more like 900/1000 range. I haven't seen the benchies so I'm still extremely skeptical.

------------------
1.2 tbird @ 1.52
Abit KT7A
256MB of Crucial PC133 @ 152 CAS2
Watercooling: Maze2, Cooling cube, Eheim1048
TNT2 Ultra
IBM 60GB ATA/100
Seagate 20GB ATA/66
Plexwriter 12/10/32A
Pioneer 16X DVD slot load
SCSI Zip
SB Live MP3+
Linksys 10/100

[DisCoDoC]

First post here. I have been a lurker for quite a while

I have scores with an Asus A7M of about 700/900 @ 148 FSB, but this is obviously with DDR RAM. This seem quite reasonable. Are you saying they should be higher?

DisCo

[sww]

I was partially referring to the posts that Ghettoweasel mentioned and partially to sapasion's and my threads about tbird systems with SDR that get benches in the 4-500 range with FSB 140+

------------------
Something like 37% of statistics are made up out of thin air.

[Un4given]

Originally posted by sww:
I was partially referring to the posts that Ghettoweasel mentioned and partially to sapasion's and my threads about tbird systems with SDR that get benches in the 4-500 range with FSB 140+

That would be a problem. At 945MHz, using a 135MHz FSB, I can get just over 400 on the Sandra FPU score, and that's with a P3.



------------------
Knight for the OC Crusaders

P3 700@945 (135x7.0 1.75V)
Asus CUV4X w/128MB PC150 (2-2-2)
Radeon 64MB VIVO 200/200
SBLive! Value
13.5GB Maxtor ATA100 (5400RPM)
Memorex 36X CD-ROM/TDK 8x4x32 CD-RW

[Siegfried00]

I seem to get a lot lower scores than I should, I'm running cas 2 at 148 and I still score below the comparison p3 1ghz cas 2.

------------------
Lian Li pc12 case
431 watt enermax ps
PIII 1ghz @1.1ghz
Swiftech MC-370 0A
ASUS CUSL2-C w/ipanel
256 megs cas 2 pc150
Leadtek gf2 64 meg 220/350
SBLive! 5.1
45 gig IBM 75gxp
12x Creative DVD-ROM
12x/10x/32x Plextor CDRW
3com 10/100 NIC
Sony CPD-G400
Windows 2000

[Molochi]

I assumed (yeah I know don't assume) that this was a spd or bios timing/setting issue. My a7v(1007) and a7v133(1003a) both scaled well in sandra2001se. At 156 on the a7v133 with bios set for 7ns memory I get ~650/720 (benches are at the bottom of sig2). I was impressed...

------------------
My system's specs dejour
My benchmarks, mods, and system pics

[sww]

You're probably right, but I think that most of the people here either know how to tweak out their boards or know where to look for info (they're here aren't they) and so I figured that it probably wasn't a BIOS problem. Mine DID turn out to be a BIOS problem, but I didn't know until I caught it in the act of changing settings back to defaults without permission. I upgraded to 1007 and haven't had any more problems.

Originally posted by Molochi:
I assumed (yeah I know don't assume) that this was a spd or bios timing/setting issue. My a7v(1007) and a7v133(1003a) both scaled well in sandra2001se. At 156 on the a7v133 with bios set for 7ns memory I get ~650/720 (benches are at the bottom of sig2). I was impressed...

------------------
Something like 37% of statistics are made up out of thin air.

[X-Tinman]

I have been working on that ram thingy myself, And have found that bios setting and GOOD Ram score fairly good on sandra. I just checked it at 150fsb,score 634-709 128 Mb Crucial Cas2. I thought that was pretty Good.

[sww]

Which mobo?

Originally posted by X-Tinman:
I have been working on that ram thingy myself, And have found that bios setting and GOOD Ram score fairly good on sandra. I just checked it at 150fsb,score 634-709 128 Mb Crucial Cas2. I thought that was pretty Good.

------------------
Something like 37% of statistics are made up out of thin air.