FX-55 memory bandwidth memtest86 V1.51

[sauria]

I just got my first FX-55 ever and I'm running memtest86 V1.51 it and on my 2 x 512 GSkill LE TCCD.

I noticed that the bandwidth is about 200-300 MB/sec LOWER than what I got with the Newcastle 3800+ & (Clawhammer 3500+) CPU's in this MSI Neo4 Platinum/SLI board.

Anyone seen this patter before? Is it the 1M L2 cache? :confused:

[sauria]

Well, performance is good under Sandra memory bandwidth in Windows XP but I wonder why the memtest 1.51 is so much lower than any on my 512K L2 cache CPUS?

[EMC2]

Different timings or frequency? Maybe your MB defaulted to different settings during the change over on something you have set on "Auto".

You could always run both CPUs with cache disabled in memtest ;)

[Onepagebook]

THis is about the right number there.;)

[sauria]

Same settings, you think this ir normal OPB?

[Blindbat]

Same settings, you think this ir normal OPB?
At the same timing and setting, I have found my Athlon 4000+ to be underperforming in bandwith under memtest86 1.51, compare to my winchester 3500+.

So I am getting same results as you, and can only assume that clock for clock and fsb for fsb, the winchester controller is actually more efficient, but much more sensitive to voltage and harder to stabilize.

I am regaining the lost, by being able to reach FSB and CPu Multi that I could not reach with my winchester.

ClawHammer 4000+ ~ CAA2C 0502 BPMW
Winchester 3500+ ~ CBBID 0501 DPDW (not knowned for their ocing ability mind you)

[sauria]

Thanks a lot Blindbat -- I wonder if any 1M L2 cache chips perform better in memtest86? A pattern here? They beat my Winchester and Newcastles in Windows Sandra memory test though (may be because of 2.6 Ghz clockspeed)

[HARDCORECLOCKER]

:D I'm thru with memtest - shows only stability under DOS but WIndows is a different number......

Last time I used it I spent about one hour with finetuning till I got no errors anymore and Windows didn't even boot with the same settings............ :cord:

:toast:

[sauria]

:D I'm thru with memtest - shows only stability under DOS but WIndows is a different number......

Last time I used it I spent about one hour with finetuning till I got no errors anymore and Windows didn't even boot with the same settings............ :cord:

:toast:
I see, do you think that it does a good job of testing RAM in an isolated way?

[lsevald]

:D I'm thru with memtest - shows only stability under DOS but WIndows is a different number......

Last time I used it I spent about one hour with finetuning till I got no errors anymore and Windows didn't even boot with the same settings............ :cord:
:toast:
Agreed :) According to memtest86 1.51 my VX3200 kit is stable at 255 2-2-2-5 (3.4V). But since prime and SPI said different, I looked into GoldMem 6.64 (www.goldmemory.cz). Turns out my kit only does 215-220@2-2-2-5, and 245 at 2-3-2-5. I'm getting thousands of GoldMem errors at 240 2-2-2-5 in one pass. IMO even Memtest for windows and S&M memtest seems better than memtest86 at detecting errors.

[HARDCORECLOCKER]

I see, do you think that it does a good job of testing RAM in an isolated way?

:D Yep - sure. Good for finding out if You have a bad module.

:toast:

[sauria]

Well, using my BH-5 gave me a little beter result up by 100 MB/sec at tight timings but still doesn't match Newcastle or Winchester results.