DDR Suggestions?

[Greeny]

Hi all,

being as new as I am to the Intel platform I was wondering if any of you would have any suggestions for purchasing ram for an ASUS P4P-800SE. My requirements are not too lofty, but quite precise. I am currently running 205MHz 2-2-2-6-1T on 2 x 256MB of BH-5 which I use for testing.

I would prefer that any ram I purchase be able to run 1T as high as possible so that I can keep the performance accellerator on (IIRC this means 1T, feel free to correct me), I'm not too fussy about the other timings if I decide to go higher on the bus speed, but the tighter the better. I won't be going higher with the bus speed unless I decide to shell out on a 533FSB Dothan, which is still a remote possibility otherwise I'm at 205~ish MHz maximum as I'm limited by the Dothan as to my FSB speed and subsequent memory dividers.

Low voltage is another key requirement I don't want something I have to pump 4v through to get tight timings. I have 2.85~ish volts at my disposal.

At the moment I am very much considering:
Patriot PDC1G3200LLK
http://www.tekheads.co.uk/s/product?product=603313
http://www.gobeyondltd.com/datasheets/pdc512350032002700llk.pdf

OCZ4001024ELDCPER2-K
http://www.ebuyer.com/customer/products/index.html?rb=9004943846&action=c2hvd19wcm9kdWN0X292ZXJ2aWV3&product_uid=66693

I have read that the platinum is not ideal for the Intel platform, but I would be willing to take the risk at the reasonable price.

Your experience/opinions/suggestions would be welcome, I imagine a few of you have owned or still do own Intel Canterwood/Springdale boards. I apologise in advance if any of you mods don't like my links.

Cheers. :thumb:

[CandyKid]

Neither of the Patriot PC3200 modules are up to spec with that OCZ set you've listed, FYI.

They're probably not using TCCD IC's in them which is why the Patriot modules are listed at 2-3-2- instead of the OCZ listing of 2-2-2- which also probably means it's a Brainpower PCB... that's a good thing.

While this isn't a big difference, it does change your overclock headroom by quite a large bit. Many manufacturers are pairing a modified PCB with the TCCD IC's and reaching incredible overclocks at amazing latencies.

For instance, Corsair Memory has their TwinX1024-4400C25 (http://www.tekheads.co.uk/s/product?product=604624)'s which run DDR550 at CAS 2.5 and most systems can still manage a 1T if they're up to that kind of overclock. The website is incorrect in their listing, however, because this memory is not tested at CAS 3, it runs CAS 2.5!

Also, OCZ has their own PC4400 version, but it looks to run at CAS 3, unfortunately. Check it out: HERE ...I am not sure if the listed latencies on these modules are incorrect (as the Corsair listing is), but we can hope!

These PC4400 modules are NOT, however, rated to run CAS 2-2-2- at DDR400 while the TCCD + BP concoctions usually are. A plus side to them is that they overclock quite well, so if you're planning a small OC, go with the DDR400 @ 2-2-2- and if you want to wring it's gullet, get the DDR550 at CAS 2.5!

PLUR
CK

p.s. Some of the TCCD setups are power hungry, but you won't need to go over 2.85 volts.

p.s.s. Almost all high-end performance memory (OCZ, Corsair, Mushkin, Geil) can run at a 1T command rate. The 99.9% limiting factor, however, is not the memory, but the chipset and CPU. The command rate is influence by them much more than your memory.

[CandyKid]

I can't seem to edit my post, but the OCZ PC4400 link seems to be bad. Here is an updated version: HERE (http://www.ebuyer.com/customer/products/index.html?action=c2hvd19wcm9kdWN0X292ZXJ2aWV3&product_uid=80149).

PLUR
CK

[fivecheebs]

Dont worry CK. IIRC after 30 mins only Mods and admin can make edits.

[Starbuck3733T]

pardon my ignorance, but can you explain the whole "1T" thing.

any opinions on crucial ballistix PC4000?

[Greeny]

Starbuck - DRAM command rate:
http://www.lostcircuits.com/advice/bios2/7b.shtml

Thanks for the comments so far, I do agree that the patriot is not in the same league as the OCZ, I just liked the price for what is reasonably tight timings. The Dothan is a performer even with low memory frequencies (I run about 384MHz 2-2-2-5 on my desktop rig and it flies).

I've been speaking with my friend who works next door and he's got me half convinced some G-Skill might be a good idea. Sounds like I can get some stuff that would give me the option of 2.5-3-3-7-1T at about 250FSB which is an interesting proposition in terms of future proofing as I am very tempted to get a later revision 533FSB Dothan and take the setup further.

Further reading on the web though has hinted that high and slacker is not much different from low and tighter in any case on the P-4 which confirms my suspicians a little. Some are hinting that 1:1 ratio on slack may be slower than say a 5:4 on tight. Sandra benchmarks are one thing, but real world testing is another.

[CandyKid]

The Command Rate setting defines how many clock cycles it takes for the CS (Chip Select) command to latch a signal onto the DRAM.

In English, that means that a 1T command rate requires one clock cycle for the chipset/MCH (memory controller hub) to select which bank/chip the memory access commands are going to. If you have only one single-rank memory module, this process is simple. If you have four double-rank memory modules, then this process is complex and will probably require two clock cycles, 2T.

The Command Rate is more hardware dependant and rarely memory dependant because it isn't an internal memory access latency (I.E. CAS) and doesn't require the memory to actually do anything, it's the MCH that's working at this point.

As for the PC4000 memory, I'd say ditch it and look for CAS 2.5 rated PC4400 modules as they'll be TCCD IC's (Higher quality, better latencies, etc).

PLUR
CK

[Starbuck3733T]

The Command Rate setting defines how many clock cycles it takes for the CS (Chip Select) command to latch a signal onto the DRAM.

In English (snip)

That was english :D

<-- electronic engineer, I just always wondered what 1T meant for ram, but that was all the explanation I needed. You're a real wealth of knowledge... you'll do well here.

as for my memory, I had 2x1GB corsair PC3200 XMS (non-LL) in my box which I thought was holding me back, so I sold it and got the 4x512 ballistix. I've heard things about TCCD being good stuff, but everybody seems to still be stuck on BH5 (which I have 4x256 of, but 1GB isn't enough memory for the work I do)

hrm.. so I should be looking for TCCD and a brainpower PCB if I decide to upgrade. I've got myself a new chip (swapped out the 3.06 mobile for a 2.8C northwood on my IC7) and I'm hopping to break 3.67 GHz which shouldn't be hard considering the cooling I've got and the fact that if I loosen the timings on the ballistix to cas 2.5 I should have plenty of headroom needed to get to the FSB to break 3.67..)

that and a few board mods are coming, VDD to 1.7V, additional capacitor for northbridge, vcore droop, etc.

[CandyKid]

ROFL, good luck with that, if you grab some PC4400, you shouldn't have any problems.

"You're a real wealth of knowledge..."

Thanks, I appreciate it!

PLUR
CK

[Bambini]

hah
<---junior year of schooling in El. Engineering ^_^

Sadly, I have no idea about any overclocking stuff. I want to learn more about it all, but I'm still worried about frying my stuff, since I dont have alot of disgressionary money.

All I do know is that I have 2x512mb corsair xms, and the timings that cpu-z give me.

[CandyKid]

This is getting off topic, but keep an eye on my DDR-II guide as I'll throw in a lot of DDR-I and other 'older' system OC information as well because most of it still applies. With all the cooling expertise around here, you ought to be able to run a very nice overclock without risking your hardware.

PLUR
CK

[jaguarking11]

ill throw in my two cent as well.

Tcc seems to be the way to go these days as it dosent need too mutch voltage (like bh5 does) to hit tight timings and high speeds. Highest voltage on tcc I have seen was 3.6v but I belive tcc dosent need anything over 2.85v to hit high fsb with low timings. But on the down side it seems tcc dosent like intel chipsets as most are tweaked to run on amd platforms.

bh5 is very good ram if you can get a good set but 3.3v on ram would be a good thing, but be warned bh5 is aging and allot of ram sticks sold on forums have been under very high voltages 4v+ and thus very prone to frying. But bh5 in some cases bh6 is one of the best to get for the oler intel chipsets

Now a third option. Very underated and decent Hynix d5 ram. Usualy dosent need more than 2.65 to hit 270fsb @ 2.5-3-3-5 acording to reasearch. Unfortunetly its very rare for this memory to breack the 300mhz barrier stably. Highest I have seen so far was 290 @ 2-4-4-5 witch is slower than 2.5-3-3-5 @ 2.85v. I personaly run this ram, I have gone as high as 270mhz on it but anything over that my cpu craps out being that my asus mobo is severly drooping on vcore.

[Greeny]

I bought the OCZ Platinum Rev 2's. Probably not the fastest around, but thats a pretty decent price for TCCD's, I can't find much else in the same price range. I'll take the gamble as to whether its a good overclocker rather than paying a ridiculous £200+ for the G.Skill...

[Greeny]

Hmm, nice bandwidth score but it's random rebooting my ASUS board at the minute. I'll have to try the sticks in my DFI and see if I can swap them out with the cheapo OCZ platinum sticks I've got in it. Think they're CH-6 so they should run better.

DIMM's are burning hot too the touch in comparison with BH-5 tester sticks at pretty much any voltage setting (have tried 2.55->2.75). Mildly worrying.

DIMMS also wiggle from side to side in slots a bit too much for my liking...

[CandyKid]

Be careful if the slots are loose, could indicate bad connections... also, your memory should get quite hot while burning it in. Most IC's will work much warmer than you'd imagine.

However, from my own experience, TCCD's require just over 2.8 volts to run 2-2-2- at DDR400 and sometimes they need it at any speed. Just a nuance of the IC.

PLUR
CK

[Greeny]

I have an incling the'yre just incompatible, the 2 x 256 of BH-5 I've had in it are the only sticks which do not reboot the machine so far. I've tried the cheapo platinum sticks and they don't work either. Don't think the board is reading the SPD's right on the two sets of Platinum EL ram either, they both come in at 2-4-4-8 if you set them to bySPD.

I've got a set of OCZ Gold sticks left to try, they're based on oldschool Samsung chips and have been in quite a few different boards so they might well work. Not anywhere as good timing wise as either of the newer sticks though they have a celing of 240MHz~ish.

The TCCD's are in my DFI at 220Mhz asynch now (10 hour memtest86 passed) so theres nothing wrong with them. Had to slacken timings to achieve the 220 so they're nothing special. No voltage options whatsoever on the board though so that's not too bad for stock.

Only really two things it could be atm and thats a PSU failure or incompatiblity (since the larger DIMMS would draw extra juice). I can still try using 2 different slots for the sticks as well and reversing the order of the DIMMS (weird how that works sometimes). I'll also try 2.85V, I just wasn't sure how high to go with them the manufacturer says 2.75V but many people say low voltages are best for the TCCD I was playing it safe.

The looseness I refer too is just the ability to rock the DIMMS side to side, they fit tighter in the DFI. In the ASUS I can put my finger on top and wiggle the DIMM from side to side far more than any other DIMM I've tried. I don't think its a problem just slightly different PCB dimension or something it just struck me as strange. Certainly the slots are not loose on the board.

[CandyKid]

220MHz w/out increasing the voltage will definately mean loose timings. Go ahead and try the 2.85 volts just to see if it needs the juice.

Did you happen to have the latest BIOS, btw?

PLUR
CK

[Greeny]

I have the latest BIOS in the ASUS, its a beta BIOS though to accept the CT-479 socket adapter for the Pentium-M. A new rev. may be forthcoming at some point, I'll keep my eye on it. Can't say it's given me any problems yet barring the DIMM compatibilities.

I'll pull the DIMMS and retest them in the ASUS at 2.85V at some point soon. The DFI has no adjustment of any voltages.

[CandyKid]

That adaptor may be the source of your problems (as well as a beta BIOS)... I wish you had a desktop CPU to try!

PLUR
CK

[Greeny]

I can get a P-4 if I need to from work, but I don't think I'll need it.

It's had a full test with the BH-5, nothing wrong with the setup it's working 100% stable.

[CandyKid]

I'm just saying that different IC's/different PCB's run slightly differently and while they're mostly compatible with normal setups, you have something rather unique.

I'd try it, just from my own curiosity!

As for the DFI, I'm really suprised it doesn't have voltage options!

PLUR
CK

[Greeny]

I've got it sussed, the DIMMs just don't work in the blue slots, am currently testing in the black slots and haven't rebooted yet. The gold stuff random rebooted as well in the blue slots and the voltages look fine so my PSU theory was wrong as well which only left the old "shuffle them in the slots" routine.

Aye the DFI doesn't have any options as it's not a lanparty it's a DFI855GME.

I'll stick some benchies up later.

Cheers CK.

[CandyKid]

ROFL... stupid blue slots! That's really wierd, I always hated when a good board decided to act up like that!

Good luck, let me know how it works out!

PLUR
CK

[Greeny]

In the end it was more than just the slots, the black slots just bought me more time in between a reboot or a lockup. I've come round to your idea about the CPU being the problem. Dropping the FSB was necessary to get the thing stabilised, I have a sneaking suspician that the CPU was simply saturated with too much ram bandwidth. Not once did I see an error in Memtest, which would seem to indicate that the RAM was fine operating at the higher frequency as you would usually see errors before you get the lockup. Interesting. I've dropped about 50MB/s of bandwidth down to just under 1.8GB/s and now its completely stable (and very bastard fast). The BH-5 by comparison was only generating about 1.6GB/s and hence showed none of the instability.

[CandyKid]

Heheh. Not to toot my own horn, but I have a fair amount of information when it comes to memory.

BTW, are you sure you're quoting the numbers correctly? 1.8 and 1.6 GB/s? Should be a bit more than that, eh?

People like to hold onto the myth that BH-5's are the best IC's ever made, but it's because they were one of the first 'best' IC's and thus have a lot of stigma surrounding them. IMHO, the TCCD's are a better choice paired with a modded BP PCB and you're set for upwards past PC4400, no problem.

PLUR
CK