Following on from a recent thread I started here, I've purchase an i7 920, a P6T SE mobo and 3 x 2 GB Corsair XMS3 1600Mhz CAS 8. Initially I just threw the ram in an accepted the default AI Tweaker settings and everything ran fine. Once I was confident that there were no major problems, I decided to get the RAM running at the designated speed, simply by upping the speed selection in the BIOS to 1333Mhz (followed by a quick test) and then to 1600MHz - everything else was left alone. At this point everything worked OK, or so it appeared, except that the timings were still at the 9-9-9-24 defaults. But during gaming (or very occasionally at other moments under load) I would get a BSOD. Errors sometimes pointed to a gfx driver issues, but having used the same driver and same card on the same OS, I was sceptical. I considered manually adjusting the timings, but given my occasional BSOD issues, I thought better of it. But then I discovered the XMP setting - I switched this on and immediately it appeared to read all the right information from the chips. Looking good... Bad gaming again and I was fortunate to have to ALT-TAB out of a game, so I notice that probe was indicating a high MB temp (in excess of 60C) and in a few minutes it tripped the PC Probe alarm as it continued to rise. At this point the fan speeds were all OK, and the CPU temp was registering 51C which is perfectly fine under load. Wondering if the previous change (to XMP) was at fault, I changed the settings back to manual and changed the RAM speed to something more modest. Obviously the tighter timings were kept, but nevertheless, this adopting this change seemed to make the difference and the MB Temp reduced to around the level of the CPU temp. I booted in Memtest86+ and ran overnight... 6 passes through the 6GB of RAM revealed 2200+ errors! So the RAM is dodgy right? Perhaps. Subsequently, I've reduced the RAM speed to 1066 and ran a long memtest again - 4 passes revealed no errors. Upped to 1333, 2 further passes revealed no errors. Perhaps there was a temporary problem while the board was bedding in? Unlikely, but I re-run memtest @ 1600Hz and a 4 passes revealed another 1000+ errors. So... is it simply faulty RAM, perhaps RAM that was placed in the wrong bin and isn't upto the task of higher speeds? Is it something to do with my BIOS configuration - at lower speeds the problem might be masked? Is is dodgy RAM that somehow manages to stumble through at lower speeds? Or is it the motherboard? 1600Mhz appears to be supported, but only if overclocked - I unserstood this to relate to the fact that officially DDR3 RAM on the 1366 socket was restricted to 1066MHz @ 1.65v max. If my motherboard doesn't support higher speeds of RAM, is it a valid warranty claim, or is the overclocking angle a get-out-of-jail-free card for ASUS? A lot of question, I know. I've logged an RMA for the RAM but I'm wanting to do all I can to diagnose the problem at my end before I send it off. I don't want to send the RAM off if I can determine that it is the mobo at fault. Any thoughts?
When I first built my system it defaulted the ram at 1066 which seems to be the norm for i7 systems . I set the timing's and volt manualy and ran memtest wich was fine . So i would say if you have set these right in the bios then the ram is bad and should be rma'd
As I understand it, 1066 is the max officially-supported (according to the specification) speed on i7... I have two concerns... the first is if it might be the mobo - clearly it's harder to test mobos than RAM. Secondly, if it is the RAM, is there any wiggle room from the supplier because running the RAM at the officially supported speed works fine. I got it from Scan and I've never had a problem with them regarding support and RMAs, so I'm not anticipating problems as such. I'm just gathering facts in advance... and I'm also keen to hear from people with similar kit and/or similar problems.
Did you set the voltage for the RAM at the manufacturers settings? If not, that could cause the problems.
It defaults to 1.66v, which I dropped to 1.64v - just in case intel's scare-mongering has some substance...
could you list all your voltages (in the BIOS)? If they all look fine then you might have a case for an RMA
I'm at work at the mo, so I can't be certain, but they are generally all set to Auto - I think XMP mode set specific voltages... I'll check tonight.
Sounds to me like dodgy ram, I had this a year or so ago when I had the asus x38 maximus extreme motherboard and ocz platimum 1600mhz ddr3 7 7 7 20. It would run fine until you gave the ram some load. But if you dropped it to 1333mhz if was fine. I would RMA and get another kit. Simon
I've ordered another kit from Scan - if the RMA'd RAM is confirmed as faulty, they have agreed to refund it - this way my PC isn't off-the-road for 2 weeks in the meantime.
XMP probably sets 1.66v (my Corsair 1600Mhz CL8 does, at any rate). 1.66 is actually fine; as detailed in a Bit-Tech article of yore it's actually not that voltage that shouldn't exceed a set value but that the differential between memvolt and uncore shouldn't exceed 0.5v (if that highly confusing explanation didn't help, try 'it's always worked for me'). Have you tried MemTesting it at XMP?
yes, post all settings here, such as RAM timing, voltages, clock speeds and RAM multipliers. im not sure whether it defaults to 1066 means you have to increase Bclk to get 1600Mhz on your RAM, is there unofficial multipliers available to select on your motherboad? in Asus's terms on PureSilver's "uncore voltage", it is the difference between IMC controller voltage and DRAM voltage.
