Teamgroup 2600 c10 2 x 4gb issues

MrPils

MrPils

MrPils

Associate
Joined
29 Jun 2009
Posts
549
Hi, hoping for some help with my difficulties with this kit.

System specs are below:

i7 3770k (custom water cooled)
Teamgroup 2600mhz c10 2 x 4gb
Asus M5F
MSI Lightning 6970
SoundBlaster Z
Seasonic 750w

I am unable to get the kit prime blend stable (7000mb allocated 24 hour test) at stock timings and volts. I have managed to get close, but not quite stable.

The defaults on this kit are 10-12-12-31 @ 2600mhz 1.65v. According to the Teamgroup and the OCUK site its supposed to be 1T command rate, however XMP is programmed at 2T. Changing this manually to 1T made no difference to stability though, a pass is still a pass and a fail is still a fail.

I've done loads of testing over the last 3 or 4 weeks as I've not had a memory kit capable of even 2400mhz before, so had to ensure my IMC can cope. I actually seem to have lucked out here as it goes right up to 2666mhz stable using default volts for vccsa and vccio. I have however tested the failing prime configs below using 1.1v vccsa and 1.2vccio and it made no difference. All testing below is at stock cpu speed.


Test results are below:

XMP / Auto : 2600mhz 10-12-12-31-2T 1.65v/1.2vccsa/1.25vccio Fails within 30 mins
2400mhz 9-11-11-28-1T 1.65v - Fails within 10-30 mins
2400mhz 10-12-12-31-1T 1.62v - Pass 24 hours
2600mhz 11-13-13-35-1T 1.65v - Pass 24 hours
2600mhz 10-12-12-31-1T 1.65v - Fails within 5-15 mins
2666mhz 11-13-13-35-1T 1.65v - Pass 24 hours
2800mhz 11-14-14-35-2T 1.65v - Boots to windows, not stable in the least.

I freely admit that i'm in uncharted ground here...I've had no experience of tweaking memory above 2200mhz before, so its possible there's a key timing or setting hidden somewhere in bios that I'm unaware of (Dram CLK period from 5 to 7 was the key for stability at 2600/2666 for example, it took me a good few days to figure that out!). To try and remove this possibility I have tested using spd, xmp, board auto settings and using the Samsung and TridentX presets (TridentX particularly sets very lax subtimings including LB14). I have found stable subs using the reduced primary timings listed as stable above though, so i'm as confident as I can be that i've done as much as I can to mitigate this possibility.


I've put in a webnote a few minutes ago to see what the OCUK techs can suggest, but thought I'd open this to the floor while I wait for a reply :).

Many Thanks!
 
Last edited:
Even though I can run 2600 and 2666 stable at default IMC volts? Primary timings are dependant on the memory modules aren't they rather than the IMC? I mean I can boot to windows at 2800mhz 11-14-14-35-2T (though its not stable) at near default IMC voltage (0.985vccsa/1.0875vccio).

I was fully prepared to face my cpu not going to 2600mhz...my true target was 2400mhz with decent timings with the possibility of more MHz given the kit. However I can't hit decent timings at 2400mhz either. That in combination with being able to get to windows at 2800mhz is what's making me question the memory rather than my IMC.


Got a reply to the webnote...an RMA number, no advice for further testing just send it back :(. Its valid for 28 days so ill probably do a little more testing over the next few days just to be certain, so I'm still open to any suggestions on anything to try...

Cheers!
 
No I haven't actually. Everything in that section is on auto, including the swizzle stuff as well.

Skews and Slews have always confused me tbh, never quite sure where to start! I was under the impression that messing with those was more for max validations than enhancing 24/7 stability? From what I understand these settings are dependant on individual systems too, there's no "good fit" setting for everyone's pc.

I haven't tried single stick testing either, but then that wouldn't necessarily tell me anything. I'm pretty close to stable in dual channel, so I wouldn't be too shocked if single channel everything is fine.

I guess it all boils down to whether primary timings are IMC dependant or not. If they are IMC dependant then I could have a weird IMC that cant run tight timings but can do big mhz. If as I suspect primaries are memory module dependant its almost certainly the kit. Its a shame there's so little solid info out there on exactly how to tell the difference between IMC and memory module instability. Its all witchcraft and miracles! :p

Teamgroup themselves suggested I test at 9-11-11-28-1T, and when this failed they suggested I contact ocuk for further testing/rma.
 
Yeah frustration is the key word at the moment...tantalisingly close at times but just not quite there!

I'll see if I can see any change in stability from playing with the skews. MCH duty sense is another I haven't touched either.

I'm glad to hear you would expect XMP to work as long as the IMC can achieve the memory ratio, that was my impression as well. Thx for that, its set my mind at ease a little :). Really don't want to get hit for double postage and a £12.50 fee because I missed something!
 
Is cross posting ok? I don't want to get my hand smacked :)

Edit: Posted in the other forum, thanks for the recommendation.
 
Last edited:
A-ha, a fellow sufferer! Misery loves company and so on ;). Id be interested to hear if you bought your kit from the same batch, mine were ordered on the 8th July.


Thanks for the assistance 8 Pack, its much appreciated. I was trying to get stock timings stable to be honest...that would do me just fine. Hell id settle for 2400mhz 9-11-11-28 but I cant get that either. I've tried up to 1.7vdimm near enough at both stock and the Xtreem 2400LV spd settings (same chips) and prime dumps within a few hours. I get odd hangs and pauses in use and the odd display driver recovery too, so its not just prime stability im lacking.

If I drop the timings to 11-13-13-35-1T I'm 100% stable at 2666mhz. I've dropped vdimm down to 1.63v now (1.62v errors, so that's the real minimum for the sticks at those timings). 1.61v is the minimum needed for 100% stable 11-13-13-35-1T 2600mhz. XMP timings at 2400mhz takes 1.58v (1.6v actual), this is kind of what I was aiming for at 2600mhz to be honest (1.60-1.65v listed as stock).

vccsa and vccio have been at stock the whole time, though I have tried up to 1.25v on both on occasion to try and force stability with XMP settings, but no joy.

The Raw Mhz profile (LB14) isn't stable for 10 mins at 2600mhz if you correct primaries to 10-12-12-31 with 1.65v. Testing at 1.675v now and its been going about 20 mins, I expect it to last around an hour if it falls in place with previous testing.

I'll get on to some single stick testing tomorrow night, and just on the off chance ill try the black dimm slots (though I know they're not as good).

I'd just like to do as much as I possibly can to be sure its the memory and not the IMC before I send it back. No sense in wasting ocuk's time and my money :)
 
Heres XMP timings on the Maximus 5 Formula:

2600mhz
1.65vdimm
1.20vccsa
1.25vccio

CL 10 - TRP 12 - TRCD 12 - TRAS 31 - CMD 2T

TRRD 8 - TRFC 152 - TREFI 10360 - TWRT 10 - TRPT 10 - TFAW 24 - TWRD 10 - CKE 8 - TWCL 8

TWRDR 5 - TRWDR 6 - TRWSR 6 - TRRDD 5 - TRRDR 5 - TRRSR 7 - TWWDD 7 - TWWDR 7 - TWWSR 5


Lasts around 10 - 15mins on an 8k - 720k 7000mb blend.

Here's a stable 2600mhz:

1.63vdimm
0.925vccsa
1.05vccio

11-13-13-28-1T
4-128-10360-5-4-4-16-5-7
1-4-4-1-1-5-1-1-4

Reduce any primaries and you get a prime thread or two drop, even if you use XMP second and thirds listed above.
 
I have pages and pages of notes...the last half written in blood lol. Just cannot get those 3 primary timings down to stock.

As an aside should I be raising vccsa or vccio if im pushing past 1.65v? Don't want to damage the cpu.
 
Quick note...my CPU is at stock for all those tests :). Left it at the offset +0.030 that I need for 4.6 (bad core) and dropped multi to 39x, so its actually overvolted.
 
If your sticks act like mine then 11-12-12 at those timings will drop just as fast, 11-13-12 will last an hour or two and 11-13-13 would be stable...
 
Yeah the sticks themselves should be much better than we are seeing, I know theres no compatibility issue with my board as many over at XS are using exactly the same combo as me (that's why I picked this kit specifically). I went for a Samsung kit over Hynix as I was more concerned with latency than raw MHz given I didn't know my IMC limits.

I think its more likely you have a bad batch (or at least a couple of bad kits that myself and James have been unlucky enough to get). Regardless, I'll hold off on sending my kit back until after the weekend so I can say I've covered everything I can test before sending them back.

Thanks for looking into it, really appreciate an experienced opinion :).
 
Last edited:
I seem to have had some success...

First I tested the sticks individually in red slot 1 - as expected they seemed to run fine. I then did the same in red slot 2...and one failed prime within 5 mins.

I have now swapped the sticks round and am running dual channel again - stick 2 is now in slot 1 and vice versa. So far its done 2 hours blend at xmp/auto, LB 14, CLK 14 whereas before it would fail within minutes. That's with cpu @ 39x and +0.030v, 1.65vddr, 1.1vccsa, 1.2vccio, 1.2 2nd vccio and all other voltages auto.

I'll leave it running and see if it fails by tomorrow night. A ray of hope at last though! :)

Probably worth trying this yourself James...
 
Just for this test (3 hours now and still going strong, I'm trying not to get my hopes up too much!) its all at auto...if its stable then I'll re-run with my previous 2nd and 3rd timings from the stable 11/13/13/35 run and see what happens. Trrd was at 4 then, trrd 3 would pass 32m but not hyperpi or prime.
 
Last edited:
Mine dropped a thread after 3 hours 24 mins. The other 3 threads were still running this morning (12 hours).

Thats as close as I've ever come to stable at stock settings, however the timings and bandwidth were absolutely horrible. On a whim I swapped the sticks around again and it failed within 5 mins.

Im not willing to go over 1.65v to get them stable at stock settings, if they dont do stock they can go back. For testing purposes though ill slowly up the vdimm and see if they do actually become stable at any point. Itll take a while at 3hrs+ between failures though :eek:.


For James2k - my stable 11-13-13-35 3rd timings set in the order they appear in bios (M5F compared to your M6...) :

TWRDR 1
TRWDR 4
TRWSR 4
TRRDD 1
TRRDR 1
TRRSR 5
TWWDD 1
TWWDR 1
TWWSR 4


Edit - On consideration I think ill try re-running with vccsa and vccio at stock first and see if it makes any difference to the failure time. If it fails quicker then I know I need to look at these two voltages, if not then ill add a bit of vdimm and see where we get to.
 
Last edited:
8 Pack said:
Guys what about mem test??? Prime is not a memory test exclusively!!! mem test please!!!
Click to expand...

Im testing at stock everything...3.9ghz with an offset overvolt and 100mhz pcie. Im also stable at 4.6ghz/2600mhz using 11-13-13-35-1T using the same offset and voltages, so my hardware is fundamentally sound despite the frustration im going through at the moment.

Im really not keen on memtest (bootable) as a stability test as from experience I have found it reports stability when windows is not. I always have to add a little voltage to get from memtest stable to prime stable (though memtest stable will run super pi32m, 3dmark etc). Its good for dialling in your benching settings, but I would never use it to test 24/7 stability. I really only tend to use memtest when im unsure if I can boot to windows and dont want to risk corruption. HCI memtest for windows gives me much the same result...it doesnt find anything that Prime wouldnt find quicker, and sometimes finds nothing where prime fails. A good example of this is my Crucial Elite which needs 1.56v for 2000mhz 9-9-9-21-1T for memtest stablity but will still bsod in games and drop prime threads. 1.58v is required for actual stability.

My Prime95 settings are custom 8k - 720k and 7000mb allocated. I find this picks up errors quickest on SB/IB, for some reason that 720k test really pushes this architecture hard.

Cheers for the help so far...Im particularly looking forward to seeing your results from a current stock kit (id be happy to take that kit as my rma if it passes stable, just incase there's an issue with opening stock).
 
Will do when I get home tonight - it'll be late though :)

What settings do you want? The 10-12-12 from last night that lasted 3 hours or so or the 11-13-13-35 with tweaked subs?
 
Got HT disabled...nothing I do takes advantage of it so no point having it on with all the extra heat. This chip is about the hottest running 3770k I've ever seen...crying out for a delid but I haven't got the guts to do it lol :).
 
Heres my AIDA64 at XMP / AUTO.

Vccsa and Vccio voltages dropped for an overnight test tonight...same as last nights fail after 3 1/2 hours but with everything default except 1.65vddr. Curious as to whether it drops a thread any quicker.

 
Quick update:
Default vccsa and vccio failed in 7 minutes...considerably faster.
Default vccsa and 1.2 vccio failed in just under 2 hours.

As a reminder 1.1vccsa and 1.15 vccio failed in 3 1/2 hours.

It looks like im finally seeing some reaction to voltage change now ive swapped the dimms round in the slots, hopefully I can tune in to stability.

Next step will be a run at 1.15vccio to see if it fails faster than 1.2vccio, and therefore if its worth trying further increases. Following that ill be doing some independant vccsa testing, then hopefully putting them both together. The point in that being once I have found where increasing voltage no longer increases stability I can set both voltages to their respective values.

Still finding it somewhat annoying that XMP is not stable. XMP should always be 100% stable and offer you a base to tweak from. I am still getting the feeling I need a slight nudge of vdimm (1.67v - 1.68v), but we shall see.
 
This thread is kinda split as theres a portion of it over at ask 8pack...I didnt want anyone thinking it had gone unanswered or abandoned so theres a couple of replies there. This is notable enough to duplicate here :

8 Pack said:
I have this stable now. Will post settings tomorrow morning!! On Z77 seconds can be fairly tight with TRAS at 35.
Click to expand...

And my reply :

MrPils said:
Interesting...so TRAS needs to be relaxed from the XMP setting of 31? Thats disappointing...but also explains why my 11-14-14-35-1T timings were always easily stable. I'll do some TRAS testing this weekend.
Click to expand...


Damn you TRAS! Its quite feasible its benchable and mostly stable set low (under 30), but needs raising for full prime stability (or perhaps it just needs big volts). That would explain the reputation for low tras in samsung based kits anyway.

Still, XMP should set 35 if it needs 35 for stability, it shouldnt take this much effort to get them stable. This has been quite a poor experience with Teamgroup for me so far!
 
You must log in or register to reply here.
Back
Top Bottom