The official HWBOT OC Challenge June 2010 thread.

[placid]

Does all versions Maxxmem allowed, or only the latest?

[aGeoM]

Does all versions Maxxmem allowed, or only the latest?

 

http://www.maxxpi.net/pages/downloads/maxxmemsup2---preview.php

 

* There are no changes at the main benchroutine since initial release.

 

So your reaced results will be comparable, from v1.05 to all further upcoming releases.

 

I had to use v1.74 bcz 1.82 doesn't run on my M2N32-Deluxe, checket the task manager, and CPU-Z was heating all resources, and got stuck.

[Massman]

1000p on DDR2 is nice. Could it be because of the NB speed?

 

In my DDR3 testing, I noticed that NB speed made quite the difference.

[aGeoM]

1000p on DDR2 is nice. Could it be because of the NB speed?

 

In my DDR3 testing, I noticed that NB speed made quite the difference.

 

yes indeed does, I´m on DDR3 now, and was a good jump from IMC ~2900->3200 plus thigh timings on memory.

 

With better cooling is possible do hit 1100 marks.

[Massman]

Yes - mostly because L1 and L2 caches run at core speed

 

If I not mistaken, the IMC in the S939 CPUS is running always 1:1 with the CPU frequency, which means that increasing the CPU frequency also increases the IMC frequency.

[zoro]

Not always. you can choose a different CPU:RAM divider if i remembered correctly

[Massman]

Ram != Imc

[aGeoM]

If I not mistaken, the IMC in the S939 CPUS is running always 1:1 with the CPU frequency, which means that increasing the CPU frequency also increases the IMC frequency.

 

Correct, but for S939, another factor must got our attention, HT Link frequency. Doesn't afect scores as IMC's does, but have influence too.

Squezze them all...

 

 

Another thing I notice, doesn't matter much the number of cores in use, results are identical.

[zoro]

What about L3 cache? i think it makes the maxmemm to run slower or not?

[aGeoM]

What about L3 cache? i think it makes the maxmemm to run slower or not?

 

Yes, it does. I can confirm, by comparing my Phenom II 955/965's with Athlon II/Sempron results.

[zoro]

Thanks aGeoM!

one other think we have to consider is the use of 32 or 64 XP,Vista or Win7. I am checking some myself.

[ageorge2500]

higher fsb is better than multi for black edition's and vista is better than win 7 .I have not try xp yet

[knopflerbruce]

OK... HOLIDAYS! Maybe I'll look for my 939 drive straight away to test some sticks.

[knopflerbruce]

OK, My PC5000 OCZ sticks are very good based on the memtest results, I could run 2.5-3-3-7-1T at 3.1v and... 326Mhz:D (tested one pass through test 5, that's the toughest one AFAIK) - I wonder if this particular test likes frequency over timings, if so maybe I should use these instead of my VX kit:)

 

I'll try in windows when I find my bench drive, it's somewhere in my mess

 

Edit: 2.5-4-3-7 335Mhz with 2 errors Too afraid to go any higher on the voltage for just quick testing, so I'll leave it at that I have more sticks to try. And... still onthe Expert, I bet they can do slightly more on the Venus:)

Edited June 9, 2010 by knopflerbruce

[chew*]

Ok guys don't waste your time on ln2, NB does not play a major role in l3 less cache chips on ddr II.

 

Pic in a bit when i shut this system down.......running 32m on athlon chip with ddr II O_o

[chew*]

check this out, the math is not adding up, wheres my boost from NB with l3 less cache cpu

 

http://chew.ln2cooling.com/athlon%20II%20255/LN2/maxmem2.jpg

 

Frosty shot for kicks.

 

http://chew.ln2cooling.com/athlon%20II%20255/LN2/IMG_2633.jpg

[aGeoM]

Ok guys don't waste your time on ln2, NB does not play a major role in l3 less cache chips on ddr II.

 

Pic in a bit when i shut this system down.......running 32m on athlon chip with ddr II O_o

 

I didn't say that, it's important but not the major, Maximum Latency read does.

[chew*]

I didn't say that, it's important but not the major, Maximum Latency read does.

 

very well aware of MRL, at best it got 30 points not 300.

[aGeoM]

very well aware of MRL, at best it got 30 points

 

yep how far did you go... 55..52 maybe.

 

Check my new score DDR3

[chew*]

yep how far did you go... 55..52 maybe.

 

Check my new score DDR3

 

40?

[aGeoM]

40?

 

Here is the full timings

 

 

[chew*]

yah thats not the problem , the problem is apparently this bench is useless and bugged apparently with l3 less cpu's.

 

It's rather obvious that there's no way in hell that that sempron will beat a phenom in anything mem intensive other than this bench much less intel in ddr II.

[aGeoM]

yah thats not the problem , the problem is apparently this bench is useless and bugged apparently with l3 less cpu's.

 

It's rather obvious that there's no way in hell that that sempron will beat a phenom in anything mem intensive other than this bench much less intel in ddr II.

 

I don't know about programming, and at some point you are right, does not make sense, but maybe... L3 is much slower than L1 and L2, and the program calculates the access time based in the 3 caches, making the score lower in L3 cases. Hopefully I also have here an 965 Phenom II, and we can test this "bug" theory. Until now what I notice is using lower MLR in conjunction with the other settings have showing an constant growing in scores.

 

BTW what is your default MRL at NB>4000?

 

Thx

 

EDIT

 

Run MaxxMEM w/Phenom II 965 @ same settings as before w/ Sempron, only Ganged was active didn't notice until I opened CPU-Z, but shouldn't be the problem:

 

 

EDIT 2

 

Played w/ NB, now @~3700 and lowered MLR to 43, tRW-TO and tRW-WB to 5:

 

 

Uploaded with ImageShack.us

 

What's the conclusion of it?

Edited June 10, 2010 by aGeoM

[chew*]

conclusion is bench is bugged and your MRL is alloiwng you to go very low for whatever reason. On ddr II board max lowest is 52 atm

[aGeoM]

conclusion is bench is bugged and your MRL is alloiwng you to go very low for whatever reason. On ddr II board max lowest is 52 atm

 

I'm on DDR3, and yes 52 is the lower I got booting at 9x on DDR2, and you can lower your MLR too, just configure lower NB in Bios and you will see. eg. 7x or 8x