• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Memtest/IntelBurn/Small FFT fine but blend test fails?

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Flaw

New Member
Joined
Nov 26, 2009
Hello there!
I'm trying to troubleshoot the TDR issues I'm having with my GPU and came accross this rather peculiar behaviour with the blend test of Prime95.

My system is a Q9550 OCd to 3.4GHz (FSB 400, 1.24Vcore, load temps under IntelBurn are around 65) on an Asus Commando with 4GB of Kingston HyperX 800MHz CL4 ram.
The PSU is a Zalman ZM600-HP, and my GPU is an Club3D Radeon HD4890 Superclocked Edition.

Overclocked:
I can run InterlBurn/Small FFT tests over night with 0 errors.
I've also just ran Memtest86+ over 20h with 0 errors.
Now, when I try the prime95 Blend test, it fails within the first minute.
Usually within the first 10 seconds, on all cores.

Stock settings: (by manually setting the FSB to the default of 333MHz, I don't trust the automatic values.)
Just as above except that Blend test now runs for max 5minutes on one of the cores, usually atleast one fails within the first minute.

Any ideas what could cause something like this?

Also, I'd like to point out that HW Monitor reports my PSUs +12V actually giving out around 13V, 12.97V-13.05V to be precise.
My multimeter is out of batteries atm so I can't confirm this value, but is it were to be correct, is something like that serious?
All other voltages reported by HW Monitor seem to be relatively close to what I know they should be.

Thanks for your time in advance.
 

gwar2288

Member
Joined
Oct 6, 2009
It means there is something wrong with the CPU-Ram Link.

The Ram is fine, as it passed memtest.
The CPU is fine, as it passed the Small FFTs

Thus, the link is the problem more than likely.
 

redduc900

Inactive Moderator
Joined
Dec 17, 2000
Location
Portland, OR
Take FSB Termination Voltage off AUTO (1.200V at default), and bump it up to 1.25 - 1.3V. And if you have 4 x 1GB sticks of RAM installed, try bumping the NB Vcore up to 1.35 - 1.40V (I believe default is 1.31 - 1.33V). Also manually set the Memory Voltage to 1.95V at 4-4-4-12-2T, w/ DRAM Static Read Control disabled. Try each of these settings changes one at a time, so you can compare the length of time before Blend test failures, then pinpoint the setting(s) that need adjustment.
 
OP
F

Flaw

New Member
Joined
Nov 26, 2009
Thanks for the replies.

So, at the moment I'm sitting at 1.3V FSB Termination Voltage and 1.4V NB VCore.
(I had already set my memory timings to the correct values, except for command rate, which I can't find in the bios. CPU-Z reports it as being 2T so I thought that's ok.)

While I went through the voltage increments one at a time, things seemed to improve each time. First at 1.25V FSB Termination and 1.35V NB VCore, all cores except one failed within the first 1minutes while the last core kept on going without a hitch for 15minutes (which is a new record by miles).

After bumping up to 1.3V FSB Termination and 1.4V NB VCore, now only 2 cores fail within 2minutes. The other two usually go on for about 5-6minutes before either one fails, and the last core standing will continue calculating happily for around 15minutes before failing.

Could this be a case of insufficient/bad power?
 

redduc900

Inactive Moderator
Joined
Dec 17, 2000
Location
Portland, OR
Post the VID of your chip, using Real Temp--> 'Settings' page--> Max Core VID. Next compare the VID w/ the Core Voltage shown by CPU-Z while running Prime. If the voltage shown in CPU-Z is lower than the VID, bump up the Vcore until the loaded voltage is slightly higher than the VID.
 
OP
F

Flaw

New Member
Joined
Nov 26, 2009
RealTemp says that the Max Core VID is 1.3000V.
I bumped my vcore up to 1.35V in bios, which translates to 1.28V under stress.
The blend test is still a no-go.

Given the huge vdrop of this board that means I'd have to go almost to 1.4V Vcore to get above 1.3V in stress. Is that safe? I remember reading somewhere that the max vcore for this chip is somewhere around 1.365V.
 

redduc900

Inactive Moderator
Joined
Dec 17, 2000
Location
Portland, OR
'Absolute' Max. Vcore (Vcc) for your chip is 1.45V, w/ 1.3625 being the max. recommended voltage. Loaded Vcore needs to be higher than the VID of the chip. While running Prime make sure to monitor the core temp's w/ Real Temp or Core Temp, and if Distance to Tj Max drops below 25C then lower the Vcore and / or improve your cooling.
 

bing

Low Profile Senior
Joined
Jul 13, 2006
Location
Indonesia
Since this is an old P965 chipset and was made around core2 65nm generation, the VRM might not suitable for oc-ing latest 45nm quad.

I guess the VR circuit is really similiar or identical to P5B Deluxe model, any thought of doing the simple pencil mod to improve the vdroop issue like this HERE ?

Again, I'm not sure if its the same VRM circuit as P5B Deluxe model, make sure you double check and suggesting to search for pencil mod in the net as well.
 
OP
F

Flaw

New Member
Joined
Nov 26, 2009
Ok then.
I'm now sitting at 1.4V Vcore, which translates to 1.328V under load according to cpu-z.
And the blend test is still a no-go. First core (as in first to fail, not core #1) failed after around 2minutes, the next two pretty much at the same time at the 5minute mark, and the last one is still going, 19minutes now.

EDIT: Ah, missed your post there before pressing reply bing.
I've looked into that, I've actually got a nice picture showing where I should do the pencil mod on the Commando.
But I'm also on the verge of buying myself an Asus P5Q Deluxe if I can't get this thing sorted out with the Commando (referring to the TDR problems I'm having with my GPU), so I haven't resorted to that as of yet.
I know very little of the pros and cons of doing the pencil mod, actually all I know is how to do it and that it is supposed to reduce the vdroop on this board.
 
Last edited:

bing

Low Profile Senior
Joined
Jul 13, 2006
Location
Indonesia
But I'm also on the verge of buying myself an Asus P5Q Deluxe if I can't get this thing sorted out with the Commando (referring to the TDR problems I'm having with my GPU), so I haven't resorted to that as of yet.

If you're looking for P45 chipset mobo, suggesting to go for EP45-UD3P or UD3R series.

I know very little of the pros and cons of doing the pencil mod, actually all I know is how to do it and that it is supposed to reduce the vdroop on this board.

Well, since your Commando is out of warranty period, can't see any significant on the 'cons' part.
Especially in your particular situation on grabbing new P45 mobo, the "cons" part is beaten by the "pros" part by big margin. :D

Just my 2 cents. ;)
 
OP
F

Flaw

New Member
Joined
Nov 26, 2009
If you're looking for P45 chipset mobo, suggesting to go for EP45-UD3P or UD3R series.

I would if one could find one of those anywhere these days.
Literally not a single store over here has that thing. I'd have to order it from god knows where and it'd arrive by spring next year, most likely.

Also, P45 is not set in stone. I just figured it'd be a good choice for my system while I'm planning on staying with just 1 gpu. Suggestions are welcome regarding that matter too.
 
Last edited:
OP
F

Flaw

New Member
Joined
Nov 26, 2009
And I'm back.
This is solved now, turns out the other 2GB pair I have is actually mislabeled PC5300.
I always thought SPD was just reporting wrong values, now I know I'd do well to trust SPD over the label.

Now that I'm only running with the other 2GB kit in the machine I can get to 450FSB np.
Anyways, thanks to everyone who contributed.