Asus p5ad2-e and corsair xms2 issue ?

[skipio]

I have just bought a p5ad2-e with a Corsair kit of 2 xms2 5400C4 Pro 512Mb memory sticks. The memory revision is XMS5402v1.1.
I can not get the motherboard to work with the 3:2 memory divider/setting (i.e 800fsb, 600 memory). The 400 (1:1) and 533 (4:3) settings work fine. The 600 (3:2) setting defaults to 1:1 (i.e 800fsb 400 memory)! Has anyone else experienced this issue? Has anyone tried corsair or other memory at 600 (3:2)?

[Xassius]

I have just bought a p5ad2-e with a Corsair kit of 2 xms2 5400C4 Pro 512Mb memory sticks. The memory revision is XMS5402v1.1.
I can not get the motherboard to work with the 3:2 memory divider/setting (i.e 800fsb, 600 memory). The 400 (1:1) and 533 (4:3) settings work fine. The 600 (3:2) setting defaults to 1:1 (i.e 800fsb 400 memory)! Has anyone else experienced this issue? Has anyone tried corsair or other memory at 600 (3:2)?

Just to note, there is no point running asynchronous dividers with DDR2.

Better to aim for low latencies with the 1:1 divider.

[skipio]

Xassius, you may be right on that although i am not sure how low one can push the latencies even at 400. The lowest i've seen is 3-3-3-8-3 at 400 and the corsair should do 4-4-4-12-4 at 600 and 667. Sure the latencies are lower at 400, but they memory speed is also faster. Will have to read around do some tests on that.

However apart from the performane issue there is the validation issue, i.e. Asus sais p5ad2-e does 600 (3:2) with corsair xms2 5400c4, corsair sais the memory is good enough for 667 so why is my system not capable of doing this? Might there be something wrong with the memory (although i tested it at 533 (3:4) with memtest86) or maybe there is something wrong with the board? or?

Wierd, there is noone around with orsair or any other memory that can simply switch his p5ad2-e memory divider to 600 (3:2), confirm with cpuz the memory setting and let me know if it works?!

[OneyedK]

Did you set vDimm to 1.9Volts? (I give them 2.1V 24/7, they don't run hot...
If not, do that and try again...
No problem whatsoever with the non-pro version of your sticks...
@DDR533 latency settings: 3-3-2-8-4
(on P5AD2 Premium)
@DDR600 latency settings: 4-4-3-11-4

[skipio]

I have set the memory to 1.9V, i also tried 1.95. I am using 4-4-4-12-4 which is the memory timings the 5400C PRO modules should work fine at 667 and have also tried 'by SPD'. But whenver the 600 (3:2) option is used cpuz reports 400 (1:1) and in the SPD section it always reports 4-5-5-15. The wierd think is that according ot the asus QVL for memory on the p5ad2-e, the 5400C Pro is validated at 600 while the non Pro is not mentioned. OneyedK could you please let me know what does cpuz report for the non pro sticks in the SPD section and what is the memory revision?

[Xassius]

huh?

My ram can run DDR2-720 at 3-2-2-4 on my AA8 duramax 3:4
and Run DDR2-675 3-2-2-2 on my P5AD2-E 1:1

[Xassius]

skipio...my review...here

www.ocprices.com - should be first article. Async + Sync basic testing done

[skipio]

Xassius, gone to your page.
But the differences in the "DDR2 Dividers - Sync or Async" in the p5ad2-e review, isn't really big. The 280 fsb/ 3:4 ratio gives 1% lower pcmark, 3% lower everest bandwidth and same picalc as 280fsb/ 1:1 ratio. These %ages are pretty similar results to the ones Viperlair gives for badwidth and latency between 1:1 (3-3-3-10-4), 3:4(4-4-4-12) and 2:3(4-4-4-12). And from what i have seen , Anand suggests that tRAS between 2 and 12 gives about 3% difference (all other timings and fsb beeing the same).
It seems to me it is a close call. Moreover, since you managed to reach 434 memory speed at 3:4 with 325fsb, you should have been also able to reach 420 memory speed at 2:3 with 280fsb. Then i wonder if maybe the further increase in memory speed may have turned around the above low % differences.

I've been o'cking since the days of the Pentium 133 & 150, but usually i preffer to stay within 20-30% from the prescribed 'default' values. The reason i want to use the 2:3 memory setting is that with 225fsb cpu speed it gives 337 memory speed i.e. 667 DDR2 which should be able to run with 4-4-4-12-4 or possibly 3-3-3-10-4 and then nothing is o'cked more than 30%.

At the moment i got no idea why the 2:3 divider on my asus p5ad2-e defaults to 3:4. Asus claims the p5ad2-e should have no issues with the 2:3 divider and the corsair memory is rated for ddr2 667. So i am not sure if there is an issue with the corsair memory (although it has passed 3 memtest86 passes at 225fsb and 3:4 memory i.e. ddr2 600 ) or with the motherboard.

[OneyedK]

Sorry for the late reply...

spd timings of my XMS5402 rev.1.1

And this is where I suppose you want to get...
--> 2.1V vDimm

I've tested the 2:3 setting too, but the memory bandwith is much lower then... So I woudn't recommend even trying that...

All these are achieved with my "old" P5AD2 Premium...
Hope to be able to re-test everything with the E-version next week...
W8ing for my new nexXxos XP cpu cooler to arrive...

[skipio]

OneyedK thanks. The spd timings are the same here so at least i know there is no issue with the p5ad2-e reading wrongly the corsair spd.
Regarding your link/picture. Yes, thats what i would like to get, but prefferably with the 2:3 ratio. I've been looking around and the 1:1 and 3:4 rations don't seem to show that much performance difference. The less the cpu is o'ced the less it'll throttle.

[Xassius]

quite frankly, there's not much point of a 3:4 divider, let alone a 2:3 divider - don't even go there

Also note: I am not running 434mhz on that ram. If you read the article, it is the P5AD2-E that is reporting wrongly - wrong dividers.

[OneyedK]

Ablosutely right, no need to use a divider...

Unless you have a mobo like mine (non-E-version) with no really working pci-e-lock

[skipio]

Xassius, thanks!!! I missed that point in your review, yes, zpuz missreports the 3:2 memory divider or the p5ad2-e doesn't pass it properly. So when i set it, zpuz reports 1:1 and Everest DDR2-0!! So the divider works fine.
OneyedK thanks!!! With 4-4-4-12-4 settings at 3:2 i.e. DDR 675 there seemed to be an instabiltiy at VDDR=1.9V or 1.95 so following your post above i've set it to 2.1V and it looks fine.

[skipio]

As for benchmarks. With 225fsb for the cpu, VDDR=2.1V i get the following:
DDR 450 (1:1), 3-3-3-10-4, everest gives 5900(read), 2243(write), 79.6 ns
DDR 600 (4:3), 3-3-3-10-4, everest gives 6487(read), 2575(write), 74.9 ns
DDR 675 (3:2), 4-4-4-12-4, everest gives 6595(read), 2656(write), 74.4 ns
DDR 675 (3:2), 5-5-5-12-5, everest gives 6496(read), 2629(write), 78.1 ns

So any divider gives better results than 1:1. And DDR675 (3:2) with very high memory timings gives a better latency than DDR450 at low memory timings. Even with 5-5-5-12-5 the 3:2 divider gives a lower latency than 1:1 at 3-3-3-10-4. So i do not understand why i should avoid the 4:3, or the 3:2 divider?

[OneyedK]

Well, you wouldn't need a divider if you were capable of running high fsb, 1:1... If you cpu doesn't allow you to, you should go with a devider...
Can you switch your cpu to x14??? If you can, try again and see the results
What cpu have you got under the hood?

Did anyone have the courage yet to take of the HSP's off the Corsair?

I'm assuming that they could do with more juice, but I don't dare to voltmod the board before I'm sure that the chippies can handle the voltage...

(killed two sticks of tccd on my P4C800 e deluxe because I didn't know they were not capable of handling 3.3V... stupid me...)

[OneyedK]

Datasheet of Micron chips
Used in XMS5402v1.1

max. vDimm: 2.3V
max. op. Temp: 85°C

In search of voltmod for P5AD2 E Premium

[skipio]

I got a 3.4GHz 200fsb cpu, and yes i can switch it to x14, but i haven't tried anything above 225fsb yet. But dropping the divider from 17 to 14 i doubt i'll gain much. I can't see it going above 250-260fsb with 14. To get the DDR2 to 667 with 1:1 and cpu divider 14 i'd need 333fsb for the cpu but that would give 4.6GHz which is impossible. With memory at 4:3 i'd need 250fsb but that would only give 3.5GHz.
But i am worried about the hard disk data despite having forced pci to 33 and pci-x to 100. Plus at 225fsb the cpu is at 50 Celsius on idle, mobo is at 28 Celsius on idle with a Gigabyte Rocket (not pro) cooler and throtles quite a lot. Before i try anything with the 14 divider i need to get a better cooler. Something which will drop to around 45 idle, any suggestions?

[skipio]

It looks like the Corsair XMS2 works fine at 4-4-4-12-4 at 675 and 2.1V and if i find any problems i can still take it to 2.2V on the p5ad2-e, so i'm not going to worry with a volt mod.

[OneyedK]

Ah, I didn't know the E-version went up to 2.2V

About aircooling... Can't tell you much about it...
A few friends of mine achieve good results with the cnps7700cu
but they all say it's very heavy...
I stick with water...

[skipio]

Water is expensive though, plus there is no guarantee it'll work with future setups, mobos etc

[OneyedK]

We're going offtopic now... But anyway...

Water is not as expensive as buying a different aircooler with every new setup you create...

I've build the cooling system externally from the case, so there are only two tube going into the case. That's not only very easy, but it gives better performance as well.

Some tips to cut expenses on water:
1) don't use fancy watercooling tubing, simply go to a DIY-store and buy some PVC tubes. I use simple 10/14mm tubes. They comme really cheap.
2) the reservoir is coming from a supermarket, it actually a curver bow to keep salad fresh, just drill a few holes in it and it's a perfect reservoir.
3) the radiator, I bought a Nextcool X2, but a copper heater-core from a car works even better, and comes much cheaper.
4) the pump, simply buy it in an aquarium store, pumps are the same and always cheaper then in a PC-shop.
5) water, since demineralised water needs a lot of additives to keep it clean (nobody wants the water in his tubes getting green after a while, yuk!), I use coolant from the gas station... I only have to switch it every two years...

The things you cannot save money on:
1) fans on the radiator: I go for silence and performance, so I use Aluminium titans (the aluminium als extends the mass of the radiator, so I win some performance against high cfm-fans (Enermax) at low speed).
2) the cpu and optionally gpu blocks: they really make a difference, so read lots of reviews befor you buy!

Don't worry about new motherboards and new sockets.
As long as the cpu's don't get bigger all of a sudden, the block will fit.
The zalman ZM-WB2 was not meant for LGA775, but the guys at Zalman created this nice adapter ZM-CS1 to fit a CNPS7700 on the lga ;-)
And guess what, it fits the waterblock perfectly, you only need to buy two 15mm M3 screws to hold the whole thing down.
It's not because it's not documented, that it won't fit

Most manufacturers don't change their blocks, they simply offer different holddowns for different sockets...

So what are you waiting for? Get the water pumping and the wind blowing!

One last thing, stay away from "watercooling kits", they always save some money on some parts, and that will cost you maximum performance...

[Xassius]

As for benchmarks. With 225fsb for the cpu, VDDR=2.1V i get the following:
DDR 450 (1:1), 3-3-3-10-4, everest gives 5900(read), 2243(write), 79.6 ns
DDR 600 (4:3), 3-3-3-10-4, everest gives 6487(read), 2575(write), 74.9 ns
DDR 675 (3:2), 4-4-4-12-4, everest gives 6595(read), 2656(write), 74.4 ns
DDR 675 (3:2), 5-5-5-12-5, everest gives 6496(read), 2629(write), 78.1 ns

So any divider gives better results than 1:1. And DDR675 (3:2) with very high memory timings gives a better latency than DDR450 at low memory timings. Even with 5-5-5-12-5 the 3:2 divider gives a lower latency than 1:1 at 3-3-3-10-4. So i do not understand why i should avoid the 4:3, or the 3:2 divider?

Try PC Mark...sandra may spell out different results. You results look weird - especially when 5-5-5 barely had any performance hit over 4-4-4

[skipio]

Xassius Sandra gives similar results to Everest. Maybe one needs to repeat the benchmarks a number of times and use the average value. I have done only 2-3 repeats and used the max value per case. I'll give it another go and see if the results are consistent. The bandwidth rates do not so a significant difference, however the latency shows that DDR2 600 with low timings seems to have a much lower latency than DDR2 675 at high timings. Which is what you have been talking about earlier on. I think there is a limit that low speeds at low low timings may be better than high speeds and high timings. But nemer mind that small differnces between DDR2 600 and DDR2 675, there is a significant difference between 1:1 at low timings and any other ratio even at higher timings. I think that unless i drop to cpu multiplier 14 and increase the fsb, 1:1 doesn't seem such a good choice.

[skipio]

OneyedK i decided to stay away from water at least for the time beeing and i got a Thermaltake XP120 heatsink. Keeps the cpu on average at -5 to -7 degrees from the old heatsink. I think water would have done -8 to -12 but i think i need to do some reading before i go with it.

[OneyedK]

Thermalright XP-120 is a very good choice for aircoolers...
(last time I looked, it wasn't ready for LGA775, but I see Thermalright solved that problem recently)
Hope it'll keep your baby cool