Kingston HyperX 8500

[Spirit7]

Hi all,

Slight issue with my brother's computer. He's recently purchased 2x2GB of Kingston HyperX 8500 1066MHz DDR2 RAM, for use with his Gigabyte EP45-UD3LR motherboard.

With both modules in, he gets IRQ_NOT_LESS_OR_EQUAL stop errors, but with 1x2GB module in it's fine. He has two identical sets of RAM (so 8GB in total), and both sets cause the same problem.

Any idea what I should check? I've changed DRAM Performance mode in the BIOS from ''Turbo'' to ''Standard'' and will let you know if that changes anything, but at the moment I'm at a loss...

Cheers!

 

[Plec]

What's the full spec - CPU in particular?

If it's an 800 FSB CPU and you're trying to run the memory at 1066MHz it won’t be happy in that setup. If it is, drop the memory frequency to ~800MHz and see if you can get 4gig stable (check you're using correct voltages and timings - and also set the CPU to default settings if it's clocked for now).

If that works try the full 8 gig - if it fails to post/crashes try upping the NB voltage (MCH) as is sometimes requires more juice when all the motherboard DIMM slots are filled.

If it's none of the above run memtest on each individual stick. If, however, it was and you now have 8 gig running - still run memtest and check that it's stable.

 

[Spirit7]

CPU is a Q8400, 650w Corsair PSU, Velociraptor, 500GB F1.

 

[Plec]

That rules out the 800 FSB theory then...

Have you tried manually setting the timings and voltages for the memory? I have the same sticks and they require a lot of juice to run at 1066MHz - 2.2v. (i would use the 800 timings while you're trying to get things stable too.)

I would also try upping the NB voltage (MCH) up one/two voltage increment/s - even when trying to get 2 sticks stable.

If you do manage to get 2 sticks stable you will probable still need to up the NB again to get 4 sticks stable.

If none of this helps i would memtest all sticks individually.

EDIT: I'm assuming that he had memory working absolutley fine in the MB before the Kingston?

 

[Big.Wayne]

Any idea what I should check?

So your brother was running with 2GB(2x1Gb) of system memory then decided to upgrade to 4GB(2x2GB) Kingston HyperX 8500 and all hell broke loose!

It should run perfectly on a stock system hmm

I would load BIOS Defaults, then boot into windows and check CPU-z, the Kingston would have probably defaulted to DDR2-800 so reboot into BIOS and select DDR2-1066 and then run some memory diagnostics (Memtest or Prime95 Blend)

Theres a chance the BIOS reset may sort something, failing that you may have either overheating sticks or overheating Northbridge, if its none of the above then you may have been unlucky and received a faulty stick (**** happens!).

If it's an 800 FSB CPU and you're trying to run the memory at 1066MHz it won’t be happy in that setup.

As far as I know its not even possible to run the memory higher than DDR2-800 when using a 200MHz-FSB/800MHz-System Bus chip? . . . . unless that was something an nVidia LGA775 board could do which I know nothing about heh!

Have you tried manually setting the timings and voltages for the memory? I have the same sticks and they require a lot of juice to run at 1066MHz - 2.2v.

That's strange as I just mentioned here I got some of this Kingston PC2-8500 last week and it runs flawlessly at 1.80vDimm (tested using 1.82v and 1.84v as well just in case my P5Q-E over-volts beyond what I set in BIOS).

I wonder if maybe Kingston are using different grade chips in their PC2-8500 HyperX

How confident are you of your testing results Plec? like did you reset your system back to bone stock apart from selecting the memory frequency and the vDimm or was it an on-the-fly test on an overclocked machine?

Reason I ask is because I am impressed with the Kingston stuff and may be buying some more for a few builds, thanks in advance for any feedback!

 

[Plec]

As far as I know its not even possible to run the memory higher than DDR2-800 when using a 200MHz-FSB/800MHz-System Bus chip? . . . . unless that was something an nVidia LGA775 board could do which I know nothing about heh!

Excuse the ambiguity, i was referring to if the system was clocked – i.e. that it’s more difficult to get the memory at its dedicated 1066MHz if you’re using an 800FSB CPU. Although, like you, i’ve found it’s more than possible if you’re prepared to tinker (thanks to another previous thread encounter.). As such, I’m beginning to revise my thinking on this – but hesitant to voice it in some threads as it’s easier to get a ‘newbie’ clocked keeping the memory as close to 800MHz, initially, when faced with a 800FSB CPU and 1066MHz capable memory.

That's strange as I just mentioned here I got some of this Kingston PC2-8500 last week and it runs flawlessly at 1.80vDimm (tested using 1.82v and 1.84v as well just in case my P5Q-E over-volts beyond what I set in BIOS).

I’m running my present memory at 1.9V (just checked) but then i’m running slightly under 1066MHz. However, i’ve installed another set into a friend’s P45 system and it required the full 2.2v @ 1066MHz and 5 5 5 15 timings (although i've yet to tinker so this may get revised). Because of the latter i’m inclined to suggest the manufactures voltage settings when an OP is faced with a problem with memory - or are clocking their system for the first time. I tend to work down from stock voltages with memory...

How confident are you of your testing results Plec? like did you reset your system back to bone stock apart from selecting the memory frequency and the vDimm or was it an on-the-fly test on an overclocked machine?
Reason I ask is because I am impressed with the Kingston stuff and may be buying some more for a few builds, thanks in advance for any feedback!

I installed it into a clocked machine (current spec), it replaced a corrupt set of Corsair. As mentioned above i started off at stock voltages (2.2v) as i was very close to 1066MHz – and worked my way down, incrementally, to its present setting, and a very stable clock (although not the 3.8GHz that i originally had. This i put down to the weather but have since found it’s due to a failing pushpin – hence a trip to the bay and a bolt-thru-kit waiting to be installed.)

Hope the above hss helped clarify things - and, if after reading the above, you would still like me to run a stock voltage test on my system, for a few days, just shout as it wouldn't be a problem.

 

[Big.Wayne]

Excuse the ambiguity, i was referring to if the system was clocked – i.e. that it’s more difficult to get the memory at its dedicated 1066MHz if you’re using an 800FSB CPU.

That't true to some extent but I *think* that's due to the motherboard defaulting to the 200-strap which is tight and expecting a 200MHz-FSB/800Mhz-System Bus, I found that on boards that allow an adjustable strap setting this to 266 or 333 (according to clock) removes the problem you were describing!

i.e an E5200 running at 266x12.5/3325MHz will be easy to set up using DDR2-1066 on a board that supports adjustable straps and support for fast memory! no different from a native 266Mhz-FSB/1066MHz System Bus chip!

I’m running my present memory at 1.9V (just checked) but then i’m running slightly under 1066MHz. However, i’ve installed another set into a friend’s P45 system and it required the full 2.2v @ 1066MHz and 5 5 5 15 timings

That's interesting, does sound like maybe there are different grades of chip used in the HyperX?

I tend to work down from stock voltages with memory...

I do almost the opposite, well I do start from stock [Auto] vDimm, if that passes all tests I go straight to 1.80vDimm and get Memtest loaded up (10-50 loops #5, 10-50 loops #8, 10-20 loops All), if that passes its straight into Windows and get Prime Blend 64-Bit, IBT, LinX, and Windows Memtest on the case! That may sound OTT but consider I test each stick separately as well as in pairs! takes a couple of days!

One random bit of info is that in some cases the memory will pass (bootable) Memtestx86 for 24 hours without showing one single error but will then go on to fail Prime Blend within an hour. The fix in these examples has always been to nudge up the vDimm one or two clicks and its sorted?

From that I am gathering that Prime Blend 64-Bit is a slightly better error checker than (bootable) Memtestx86 or IBT/LinX? I know these errors can be caused by other things like a flakey northbridge etc but as already said this fix to these Prime Blend errors has been fixed with a vDimm boost, dunno!