Team Xtreem 2GB (2 x 1GB) DDR2 SDRAM DDR2 1300 (PC2 10400)Dual Channel Kit + eVga680i

[Hutch]

You do realize you're testing DDR2, not DDR3, right?

[hecktic]

587.4Mhz CL4-3-3-20-2T@2.50v
http://img265.imageshack.us/img265/6...cl43320un1.jpg

[emoners]

we'll here what a friend of mine (Woody) did to his Team Group with the new heatspreaders:

[hecktic]

emoners,

Thats very clever. I might try that

[saaya]

hmmmm those heatspreaders look familiar

id say you should contact tg about this, the heatspreader you have is clearly inferior to the thermalright one, not to mention it wasnt even assembled properly and fell off

so whats the max speed you can get with 555 and 2.45v?
1300? not even 1320?

About the evga bios not going above 1400, i think thats a divider limit by the chipset, but not sure... you need a higher fsb to get higher memspeeds, in unlinked the max memspeed is limited to 1400 iirc.

[hecktic]

id say you should contact tg about this, the heatspreader you have is clearly inferior to the thermalright one, not to mention it wasnt even assembled properly and fell off

Yeah I tried looking for a phone number to a support line on their website and all I found was http://www.teamgroup.com.tw/teamgrou...iceContact.php

Thanks you for choosing Team’s products. If you have any question(s) to the product, please go to our FAQ section to trouble shoot your problem(s) , in case if you can not find an answer , please send an e-mail to the mailbox: rma@teamgroup.com.tw. Or dial the free-customer service line: 0800-821688(For Taiwan only), overseas please dial : +886-2-82265000 thanks!

I am not about to call an international number lol... Isnt there a sales office for TG in California? Doesn't one of their reps troll these forums.. or the TG part of XS forums?

But hey listen, on the other hand if I DO call them and they end up having me RMA this set back in, then arent they going to then reject them after seeing I have put more than the allowed voltage (2.35 - 2.45v) ??

Ive been on 2.50v for about 10 hours today with an insanely loud and fast desk fan blowing cool air into my PC. I read somewhere on the TG site that if they discover over-voltage usage that they will deny the RMA claim. But I think that only applies to dead RAM modules?

so whats the max speed you can get with 555 and 2.45v?
1300? not even 1320?

on CL5 my best run was this here:
651.3Mhz CL5-4-4-5-2T@2.45v
http://img526.imageshack.us/img526/2...0cl5445zt9.jpg (Bios has 1300Mhz locked in while MemSet read it at 651.3Mhz and CPU-z read it at 650Mhz)

on CL4 my best run was this here:
587.4Mhz CL4-4-4-20-2T@2.50v (Bios has 1175Mhz locked in while MemSet read it at 587.4Mhz and CPU-z read it at 586.7Mhz)
http://img518.imageshack.us/img518/1...cl44420gp3.jpg
(Currently testing 600Mhz CL4-3-3-12-2T@2.45v)

on CL3 I still have yet to test.

I was able to POST up to 1400Mhz with no problem at either 2.45 or 2.50 volts. The only problem stopping me from further testing those speeds at CL5 was the Windows XP I am running which is about almost 10 months old and has gone through a number of overclocking sessions. Now I have heard Windows XP and VISTA dont like the constant changing of hardware and if you do change your hardware constantly, the system files become corrupted and it prompts you to reinstall Windows or Repair it using Recovery Console on the Installation Discs... well this is what happened to me each time I tried and tested any speeds over 1300Mhz effective.

Im trying to overcome this by using an overclocking utility to increase the FSB from within windows. Sortof like MemSet 3.4 except it would offer FSB options as well as Latency options for memory overclocking. So far I cannot find anything that works with the evga 680i board I have.. got anything in mind?

About the evga bios not going above 1400, i think thats a divider limit by the chipset, but not sure... you need a higher fsb to get higher memspeeds, in unlinked the max memspeed is limited to 1400 iirc.

Yea your probably right, although nTune by nvidia allows for crazy options for memory overclocking but I am not so sure nTune is stable as a program as I have had my own fair share of problems with it in the past as many other evga users have had as well.

I also figured the FSB on the cpu needed to be upped a little bit more to get a good ratio with the memory as you may have noticed on my previous results I have had crazy cpu:ram ratios like 11:17 etc.

Thanks for the tips.

You do realize you're testing DDR2, not DDR3, right?

Yes ofcourse. I dont care much about DDR3 right now at least until the crazy overclockers here at XS report a more wider variety of DDR3 memory brands that can run CL4 on the high end DDR3 modules rated at like PC3-15000+

[hecktic]

some more testing with CL4 timings...

584.4Mhz CL4-3-3-14-2T@2.50v
http://img524.imageshack.us/img524/9...cl43314cv2.jpg

[hecktic]

new record at CL5

601.0Mhz CL5-4-4-5-2T@2.50v
http://img520.imageshack.us/img520/5...0cl5445hd8.jpg

Edit: Even better CL5 @ 650mhz with slightly lower voltage

651.3Mhz CL5-4-4-5-2T@2.45v
http://img526.imageshack.us/img526/2...0cl5445zt9.jpg

[saaya]

ntune sucks, forget about it
for ocing within windows use clockgen, send me a pm with your email address and ill send you the clockgen beta. it was removed from the author cause its buggy, but its better than nothing.

took me 5 hours googling to find that beta finally on some ocers ftp ^^

how are you testing stability?
did you increase chipset voltage to make sure the chipset isnt holding you back?
check the manual or teamgroups site for what settings they recommend, ddr2 1300 is really pushing the limits...

your cas4 results are very nice!
im pretty sure its faster than 1300 555... go and try it out
and maybe your mem is good with cas3 as well so maybe cas3 with cmd 1t would be even faster

and its a LOT less problematic to run sub 1ghz memspeeds on 680!

yeah there is a teamgroup rep here on xs but they are usually very busy.
i just hit him up on msn and told him about this and he said hes going to look for the thread. maybe hes only checking the tg support section so maybe post a new thread there and link to this thread here.

about losing the warranty... seeing as the memory didnt come with the heatspreaders they were advertised with and that they just fell off i dont think they will refuse your rma. the memory was not in perfect shape before you got it... having said that, dont bench any further with 2.5v for now.

a thing i learned the hard way: always try your new hardare at stock, ALWAYS! and then mod it or void the warranty by ocing and overvolting and voltmodding it.

[saaya]

the best way to oc on memory 680, in my opinion, is to set the right divider in bios with unlinked. boot with a low fsb and memspeed and then push it all up in windows with clockgen.

you need to know the target fsb and target memspeed, then calculate the memory divider, then set the right divider with a low fsb and mem speed in bios, boot and increase the fsb. the memory mode is called unlinked but the ratio is actually locked. so when you increase the fsb the memspeed will track with it.

this is a great way to avoid straps and other timings problems and especially those nasty bios holes in 680 bioses.
i didnt calculate all dividers, only the ones interesting for around 900-1000 memclock with cas3, but you can extend the table easily if you know how to use excel. in case you do, please post it on xs again to share the knowledge with others

here i hope this helps:

[Nanok]

http://www.arcticsilver.com/arctic_alumina.htm

[mad_skills]

we'll here what a friend of mine (Woody) did to his Team Group with the new heatspreaders:

I like this very much, nice work.

[Kunaak]

someone should try this ram on a P35 or X38.

1300 on a 680i isnt exactly hard...

[saaya]

someone should try this ram on a P35 or X38.

1300 on a 680i isnt exactly hard...

good point! i cant remember seeing any high ddr2 clocks on x38 boards... it seems all those crazy 1400 1500 1600 ocs were on gigabyte or foxconn p35 boards?

[hecktic]

ntune sucks, forget about it
for ocing within windows use clockgen, send me a pm with your email address and ill send you the clockgen beta. it was removed from the author cause its buggy, but its better than nothing.

Alright Ill scrap nTune... in fact I need to give my PC a good reformat and clean install of XP SP2.. I can pick between Pro and Home... which one should I use?

I want to make sure everything is perfect in terms of the OS.... I hate it when windows craps out BSODs due to excessive overclocking lol... the constant changing of hardware configurations like CPU and Memory clocks for example... it just spits out some errors and/or BSODs.

But all the testing I have done thus far are stable 100%.

took me 5 hours googling to find that beta finally on some ocers ftp ^^

oh I see

how are you testing stability?

A mix of benching tools like Super PI mod 1.5, Orthos, 3dmark06, Memtest86+, and some other misc. programs like game applications (call of duty 4, unreal tournament 3 beta, world of warcraft, crysis, etc.)

did you increase chipset voltage to make sure the chipset isnt holding you back?

Actually I did not do that yet.... I figured since my board allowed my CPU to POST and remain stable on AIR up to 2000QDR (e6300 @ 3.50Ghz w/ 500FSB) with just under 45/46 C temp. wise and only 1.4v FSB I figured the chipset voltage did not need to be touched.

However now I am finding it more and more convincing that when I alter the QDR on any mode, be it link-sync, unlink, user defined-auto, etc. when attempting to increase the QDR the POST screen will take me back, way back to 1067QDR or default speed of the e6300 chip with default memory timings per JEDEC I believe at 5-5-5-18-2T @ 800Mhz

I must have spent a good hour going through dozens of alterations in voltages on RAM and CPU FSB and CPU Core with combinations of user defined QDR values above 2000 AND even alterations on the Memory Speed with no success and instead this constant same result as described above.

So this could be the chipset needing more voltage as you say?

The most I ever got and it was a one time only thing, 2015 QDR which was about 5 or 8 Mhz more on the CPU clock and when using Link-Sync mode it provided about 10 or so more Mhz to the Memory.

This was all with a multiplier of 7x by the way... when I tried using 6x I got the same road blocks.

So chipset voltage then? MCP or SPP ? and how much more voltage?

check the manual or teamgroups site for what settings they recommend, ddr2 1300 is really pushing the limits...

TeamGroup tested this RAM under Asus-P5k which I beleive is the same board CapFTP is using with his TG RAM that was able to go well beyond 1300Mhz as seen here:
http://www.xtremesystems.org/forums/...d.php?t=169305

I know I can at least achieve 1400Mhz since it is an option in Link-Sync mode for a provided QDR value... and so far on the life of this 680i board, every time I use Link-Sync. it has given me good results.

your cas4 results are very nice!
im pretty sure its faster than 1300 555... go and try it out
and maybe your mem is good with cas3 as well so maybe cas3 with cmd 1t would be even faster

and its a LOT less problematic to run sub 1ghz memspeeds on 680!

Yes the CL4 results are simply amazing and Im hoping with a better heatsink cooler with TIM that actually covers all the ICs and not half of one like the stock TIM on both RAM modules, I think I should be able to pull off CL4 at 1200Mhz even.... so far Ive come extremely close by about 25Mhz off at 1175Mhz as seen in my previous results.

It can run CL3 with no problem with the right voltage. Right now as we speak I am running a lower voltage with lower speed 533Mhz at CL3 w/ only 2.10v and my FAN is on super low quite speed hahaha

I hope you saw my results at 651.3Mhz CL5-4-4-5-2T@2.45v

Do you know what kind of TIM can be used on RAM ICs ?

yeah there is a teamgroup rep here on xs but they are usually very busy.
i just hit him up on msn and told him about this and he said hes going to look for the thread. maybe hes only checking the tg support section so maybe post a new thread there and link to this thread here.

about losing the warranty... seeing as the memory didnt come with the heatspreaders they were advertised with and that they just fell off i dont think they will refuse your rma. the memory was not in perfect shape before you got it... having said that, dont bench any further with 2.5v for now.

Thanks. Ill make a post linking it to this one in the TG section of the XS forums. Thanks again for speaking to him about this.

I am no longer benching at 2.50v .... I dont wanna fry the RAM on the stock heatspreaders I have on them now.

a thing i learned the hard way: always try your new hardare at stock, ALWAYS! and then mod it or void the warranty by ocing and overvolting and voltmodding it.

Thanks Ill keep that in mind.

the best way to oc on memory 680, in my opinion, is to set the right divider in bios with unlinked. boot with a low fsb and memspeed and then push it all up in windows with clockgen.

you need to know the target fsb and target memspeed, then calculate the memory divider, then set the right divider with a low fsb and mem speed in bios, boot and increase the fsb. the memory mode is called unlinked but the ratio is actually locked. so when you increase the fsb the memspeed will track with it.

this is a great way to avoid straps and other timings problems and especially those nasty bios holes in 680 bioses.
i didnt calculate all dividers, only the ones interesting for around 900-1000 memclock with cas3, but you can extend the table easily if you know how to use excel. in case you do, please post it on xs again to share the knowledge with others

here i hope this helps:

THANKS A LOT!

I will let you know if I find anything.

Also I am using P31 bios for Evga 680i if that matters.

someone should try this ram on a P35 or X38.

1300 on a 680i isnt exactly hard...

good point! i cant remember seeing any high ddr2 clocks on x38 boards... it seems all those crazy 1400 1500 1600 ocs were on gigabyte or foxconn p35 boards?

Yah I may do that in the near future.

[kojiro91]

i want to thanks to @emoners, to give us such good info from the picture, because i still figuring what that for.. heheh thanks so much....

for TS.. i think you ram can run faster than that bro.. show us more number plz hehee

[Aerou]

with that voltage ... why not try CL4? it is my personal favourite

[hecktic]

with that voltage ... why not try CL4? it is my personal favourite

with what voltage ??

[hecktic]

1T works with 2.1v @ DDR2-533Mhz , & DDR2-667Mhz both at CL4
1T works with 2.25v @ DDR2-800 at CL5 and runs extremely well under SuperPi:
http://img45.imageshack.us/img45/121...8at225vdf4.jpg

[saaya]

id say install xp pro but home wont make a difference for ocing/performance afaik

and some other misc. programs like (call of duty 4, unreal tournament 3 beta, world of warcraft, crysis, etc.)

ah, those are the most reliable stability tools i know and they are so much fun too

Actually I did not do that yet.... I figured since my board allowed my CPU to POST and remain stable on AIR up to 2000QDR (e6300 @ 3.50Ghz w/ 500FSB) with just under 45/46 C temp. wise and only 1.4v FSB I figured the chipset voltage did not need to be touched.

However now I am finding it more and more convincing that when I alter the QDR on any mode, be it link-sync, unlink, user defined-auto, etc. when attempting to increase the QDR the POST screen will take me back, way back to 1067QDR or default speed of the e6300 chip with default memory timings per JEDEC I believe at 5-5-5-18-2T @ 800Mhz

I must have spent a good hour going through dozens of alterations in voltages on RAM and CPU FSB and CPU Core with combinations of user defined QDR values above 2000 AND even alterations on the Memory Speed with no success and instead this constant same result as described above.

So this could be the chipset needing more voltage as you say?

The most I ever got and it was a one time only thing, 2015 QDR which was about 5 or 8 Mhz more on the CPU clock and when using Link-Sync mode it provided about 10 or so more Mhz to the Memory.

This was all with a multiplier of 7x by the way... when I tried using 6x I got the same road blocks.

So chipset voltage then? MCP or SPP ? and how much more voltage?

ok, so i guess your on an evga 680 board right?
well dont complain that it resets mem to 800 555, the striker extreme board would just not boot and youd have to reset the bios and risk killing the board

well you should be able to get above 1300 even with a much lower fsb...
about chipset voltages... id say nb to 1.6 and sb to 1.5? not sure though i havent played with 680 for some time.

I know I can at least achieve 1400Mhz since it is an option in Link-Sync mode for a provided QDR value... and so far on the life of this 680i board, every time I use Link-Sync. it has given me good results.

well the setting is there but that doesnt mean 1400 will work
yes, linked is ok, "unlinked memory mode" should be renamed to "unstable memory mode", nvidia really did a crap job with that...

tg tests 1300 on a p5k? thats weird... you should get much better results on 680 than 1300 then...

Yes the CL4 results are simply amazing and Im hoping with a better heatsink cooler with TIM that actually covers all the ICs and not half of one like the stock TIM on both RAM modules, I think I should be able to pull off CL4 at 1200Mhz even.... so far Ive come extremely close by about 25Mhz off at 1175Mhz as seen in my previous results.

It can run CL3 with no problem with the right voltage. Right now as we speak I am running a lower voltage with lower speed 533Mhz at CL3 w/ only 2.10v and my FAN is on super low quite speed hahaha

I hope you saw my results at 651.3Mhz CL5-4-4-5-2T@2.45v

Do you know what kind of TIM can be used on RAM ICs ?

well not amazing id say but they are very good.
our cellshock 1150 555 can run up to 1200 with cas4 with 2.45v
thats the best ive seen so far, not all kits will do that of course...

you can use any termal grease you want, but id recommend silicon based stuff so you dont risk shorting the sticks.

Thanks. Ill make a post linking it to this one in the TG section of the XS forums. Thanks again for speaking to him about this.

I am no longer benching at 2.50v .... I dont wanna fry the RAM on the stock heatspreaders I have on them now.

np man
if he doesnt reply within a few days lmk and ill hot him up on msn and nudge him

THANKS A LOT!

I will let you know if I find anything.

Also I am using P31 bios for Evga 680i if that matters.

hmmmm i think thats the best one so far right? im not sure...
ask evga what bios they would recommend you

1T works with 2.1v @ DDR2-533Mhz , & DDR2-667Mhz both at CL4
1T works with 2.25v @ DDR2-800 at CL5 and runs extremely well under SuperPi:

1t only works up to 1000, thats the highest ive ever seen.... usually 1t only works up to around 900 i think, so it only makes sense with 333 really...

[WoutZoR]

Just wanted to share the innards of my Team Xtreem PC6400 C4 (not the one's mentioned in this topic).

Under the heatspreader


Micron D9GMH


Except for the Chips that are used, the innards look the same as the 1300's?

[saaya]

hmmm i dont understand why team uses thermal pads and not thermal paste like we do...
The writing on those gmh chips looks really weird... are those real micron chips?
or maybe micron ics made by a third party mfg?

[WoutZoR]

Maybe the chips look a bit odd/overbright because I used the flash when making the photo.
I think because these (PC6400 C4) are the cheapest Xtreem versions (NOT Dark) they just applied leftovers from several batches on the pcb.

The thermal paste/pad that Team uses looks like the same Pink stuff that AMD applies to their boxed heatsinks.

Instead Cellshock uses 'real' thermal paste as can be seen in this review:
http://www.techpowerup.com/reviews/C...C2_8500/1.html



But maybe this is getting a bit offtopic...

[saaya]

well what i mean is look at the writing on the chips...
the dot on the corner is not filled out completely, and its size and position is slightly off on some chips

plus the writing on some of the chips looks way bolder than on others...
if i wouldnt know it any better id say those are fakes :P
but since micron produces those chips in more than one fab i guess this is just from some other fab where the chip labeling is beeing done... very sloppy

[hecktic]

sorry for not updating this thread guys.

I have just gone through some terrible family related emergencies and am still in the process of getting back to norm.

Anyways, some even more bad news is that my pair of TG PC2-10,400 (1300mhz) ram has failed on me as of last thursday (1/24) and the very odd part is that the RAM Heatspreaders were never making solid contact with the RAM ICs throughout this entire time from when I started this thread to now, of which in that time frame, I was running this RAM at various underclocked and overclocked speeds and timings for prolonged periods of time and then finally choosing 266.5Mhz (533Mhz Effective). And to add on to that, the RAM had failed when I tried to lower the voltage from 2.35v @ CL3-3-3-9-1T by 0.10v adjustments with successful POSTs on every downgrade of voltage rating defined into the motherboard BIOS except at 1.95v upon when the POST spat out this message right after displaying information about BIOS version:

WARNING: The CPU has been changed or the BIOS has been flashed. This requires a hard reboot. Please turn off your system in order to restart.

I was running Link-Sync. mode with my CPU at 1.86Ghz.

So I did a hard reboot and then I kept getting a long beep followed by no POST.

I then figured the memory controller had something to do with it after looking up the POST code in the Evga manual for my motherboard (680i A1)

So I decided to reset the CMOS and clear it by 3 methods; once by the jumper switch w/ disconnection of all incoming power followed by a minute of waiting until finalizing the process by turning on the main power from the back of the PC, AND then once by the removal of the battery from the motherboard w/ disconnection of all incoming power followed by a minute of waiting until finalizing the process by turning on the main power from the back of the PC, and the third method was the combination of the two above PLUS the switching of the RAM from channel 1 or bank 1 to channel 2 or bank 2 (black to blue slots if I remember).

All 3 methods had failed with no POST and instead the long beep; same post code but blank screen. Not even a warning blah blah error message like the first time. I attempt now to try 1 Stick of TG ram at a time but its still not working and giving the same results of no POST with the long beeps and blank screen.

So now I reset the CMOS again and used some other ram; Corsair 800mhz 2gb dual channel XMS kit. Upon the switch of the RAM to this pair, I recieved a POST, BUT with that weird Warning blah blah error message from before.

So I reset the CMOS again and FINALLY I get a sucessful and normal POST.

So I try and stick the TG ram back in and hope maybe the CMOS clearing had actually worked this time and not the last times which would explain the Corsair memory being able to POST and not the TG. But, it doesnt work, the TG ram spits out with no POST and instead the same annoying long beep followed by a blank screen.

So I think to myself, well maybe, just maybe its the cooling of the TG ram that is the problem. So now I replace the thermal pads on the TG ram out of desperation and try and use them again to get a sucessful POST, but nope its a no go even with the new thermal pads on. The aluminum heatspreaders on the TG ram just simply suck and are very annoying to remount if need be, like in my case.

So I have another pair of TG ram also at the same specs of this pair that failed on me, but with only 1GB (512mb x 2) instead of my current 2GB (1GB x 2)

So loaded up the 1GB kit and low and behold sucessful posts left and right at the same underclocked and overclocked combinations with various speeds and timings like the 2GB kit.

At this point I call up EVGA to confirm the problem is not related to the motherboard and all I got from them was that quote "you should disconnect the power and disconnect all peripherals for about a half hour and then reconnect everything and you should no longer be experiencing those errors and the unsuccessful attempts to POST w/ the previously used combinations of RAM timings, voltage, etc."

So I did what he said but still its the same problem.

I have concluded that the 2GB (1GB x 2) PC2-10,400 (1,300Mhz) kit rated to run at CL6 6-6-6-18-2T w/ 2.35v ~ 2.45v is officially dead and was defective to begin with when receiving the RAM modules with improper mounting of the heatspreaders which later resulted in the RAM falling out of the heatspreaders and loosing all contact between any RAM IC and the heatspreader AND the actual heatspreader being faulty itself due to the RAM falling off AFTER a manual re-mounting of the RAM and making sure contact was made between all RAM-ICs and the aluminum heatspreaders.

This is so depressing having spent over $200 (after tax+shipping) for this high performance RAM.