XMS2 v2.1 queries

[KTE]

I've tried two pairs of your XMS2 PC2-6400 C4 v2.1 2x1GB kit with ProMos IC's and one pair of Crucial Ballistix Tracer PC2-8500 C5 2x1GB kit with Micron D9 IC's inside the last 4 weeks.

The ProMos IC's you use on XMS2 C4 do not oc well, far worse than the Micron ones in v1.x and they do not respond to voltage much above stock 2.1V. They die with 2.35V a few times. Fair enough, since that was never guranteed or touted, I have no complaints. But I'd definitely prefer you use Micron. :)

But can you tell me why these chips performed 0.1-0.3s faster in 1M than the Crucial at the same exact hardware/timing/OS/bench? :confused:

I mean I tested on XP Pro SP2 32-bit with Gigabyte P35 DS4 and Abit IP35-E on the same day.
Same OS, set everything up, restart. MHz/timings changed through BIOS only, reboot with the Crucial in, check Task Manager memory/system cache until its similar value and run 6x 1M.
Reboot and repeat but now install the Corsair in.

The Corsair was consistently faster on every attempt! :confused:
No kidding they were badly limited but performing very quick.

I'll give you a quick comparison. ;)

Crucial: Fully tweaked OS, E6750 8x450.7 3605MHz 676MHz 5-5-5-5-5 PL6 tRFC 45 tWTP 13 tWTR 13 tRRD 3 tRTP 3, 1M = 13.953s
http://img243.imageshack.us/img243/7785/1m675mm7.png

Crucial: Fully tweaked OS, E6750 8x450.7 3605MHz 563MHz 4-4-4-4-4 PL6 tRFC 45 tWTP 10 tWTR 10 tRRD 3 tRTP 3, 1M = 14.453s
http://img149.imageshack.us/img149/5139/cdt1mhw4.png

Corsair: Non-tweaked OS, E6750 8x450.7 3605MHz 563MHz 5-5-5-5-8 PL7 tRFC 42 tWTP 14 tWTR 11 tRRD 3 tRTP 3 (best RAM allowed), 1M = 13.906s
http://img45.imageshack.us/img45/4681/6thbesttimeat3600gh0.png

Corsair: Non-tweaked OS, E6750 8x450.7 3605MHz 540MHz 5-5-5-5-15 PL7 tRFC 42 tWTP 14 tWTR 14 tRRD 5 tRTP 5, 1M = 13.968s
http://img529.imageshack.us/img529/5629/360554085551513968gh6.png

There's plenty more at 3600/450/540 which gave 14.1-3s for Crucial and 13.98/14s for Corsair.


The main qs is: are there any DRAM timings/register performance values which are set and we cannot control?

[Yellowbeard]

I am not a super PI user very often but, a difference that small could simply be the PCB differences between what we use vs the Crucial. Also, the Promos ICs have voltage regulation on the ICs so, going much past 2.1v is only going to produce heat.

[KTE]

Neither am I. :)
But that difference (is v.big in 1M terms) on one system after you have everything tweaked as much as possible (OS) is completely hardware related and usually requires only more CPU MHz to achieve. It's quite impossible otherwise unless a specific hardware be quicker due to a certain timing. The confusing thing is, I repeated rebooting and swapping sticks and I obtained the exact same pattern. The ProMos modules were quicker than the D9 by a good margin even in 32M and bandwidth.

Yeah 2.2V well cooled looks like max feasible on the ProMos.

Thanks for the reply.

[Yellowbeard]

I'm certain Bachus will read this and he may have some input. He's much smarter than I am. Corsair only hired me for my good looks as you can easily see :D

[KTE]

Hehehe :D

I don't mind you beard if its really that yellow.... => :D

Cheers.

[Yellowbeard]

It itches though.

Actually, if you did not recognize him, it's Graham Chapman.

[KTE]

He looks too different to recognize with every look. ;)