QX9650 L739A644 + Blitz Extreme

[xgman]

George,
Congrats on the QX. You may find this kind of crazy, but I think you are having stability issues because of your FSB. My findings have always led me to stability issues whenever FSB was near 445, 450.... I cannot even boot Windows after 450 FSB, 455+ it will freeze at post. So, FSB has been a huge limiting factor for air/water on these QX's. Just a little FYI from my personal experience.

Try 10 X 400 with 1.3625 (bios selected) vcore and if you have ample air cooling capacity I am pretty sure you should be quad prime stable set to (12meg cache setting in multiprime).

I know, because you are raising the multi, bus speed and ram will be slower. But I see you are running CAS7 6-6 on those sticks. To combat this I have found the best setting on a 4.0 Ghz overclock for me was:

10 X 400 , 1.3625 Vcore bios selected (1.31 after vdroop)
Ram at 1600 mhz, Timings: 6-5-5-18 (2.05v , 1T)

As you can see, even though you are lowering FSB, you still have the ability to run CAS6 and tighter timings + I am pretty sure you will be able to achieve 4.0 GHz low volts stability.

Let me know of any results.

I heard that if you can set Static Read Value to "9", it helps to get higher fsb. I think with the Storm G5 he can set volts to 1.5-1.55 without harm. I have my phase running at 1.567v (reported). I find the 9650 needs higher volts to get in the high 4xxx's and also higher fsb stable.

George, I had a bit of a hard time getting this cpu up there at first, and it does seem to need more vcore than those ES samples, but under phase it now seems mostly stable at 4800 mhz approaching 1.6v set in bios, Odd thing is with sub zero temps, coretemp 95.4 reads the first two cores 8 deg higher than the 2nd 2. Is this normal. I know the readings for sub zero are not accurate obviously, but why so much difference between sets of cores. All that said, I think that this is one of those times when the ES's were really hand picked for the most part.

[xgman]

xgman, how's you cpu doing?

It just seems to need lots of vcore and bios tweaking galore. I'm on a p5k dlx. At first I tried it on a P5e3 dlx and 4GB ddr3 (2x2gb) and it was very hard to figure out what settings to use for best results and the 8gb ram didn't help. Then I fried that board, so I went back to the P5K dlx with 2gb ram. Now it's much easier to control, but no chance of hitting the high fsb I was used to on the qx6850. No matter, the cpu does clock higher now, and you just have to try different timings and settings in the bios to get control of this thing. Finally seems stable enough at 436/524ish x 11. It will go a bit higher but looses stability along the way. Definitely needs high 1.5's vcore to get there, but it seems like there is a point where it starts to not help anymore unless you have dual stage phase or better.

[eva2000]

coretemp can't read sub zero temps, so that's all over the place on sub zero heh

I'm just beside myself right now as to how poor this cpu is compared to all the QX9650 results out there retail or ES

[xgman]

coretemp can't read sub zero temps, so that's all over the place on sub zero heh

I'm just beside myself right now as to how poor this cpu is compared to all the QX9650 results out there retail or ES

I still don't understand why there would be such a huge difference between the 1st 2 and 2nd 2 sets of cores. I know the evap head is seated flat.

Anyway, go back to basics on your bios settings and work back up. It took allot of trial and error for me to figure out the differences is settings between this and the previous QX. You may still have a dud, but I thought so at first also.

[HDCHOPPER]

u will get it eva ! besides who needs sleep hahhahahahaa

[Dynasty]

coretemp can't read sub zero temps, so that's all over the place on sub zero heh

I'm just beside myself right now as to how poor this cpu is compared to all the QX9650 results out there retail or ES

swap the board out an try maximus formula or XE.....with appropiate bios.
you know the deal, I bet that qx9650 of yours would sing
under my P5K-Prem.(vmodded)......

my L739A663 under P5K. if you need more compares..

[/URL]

[eva2000]

I still don't understand why there would be such a huge difference between the 1st 2 and 2nd 2 sets of cores. I know the evap head is seated flat.

Anyway, go back to basics on your bios settings and work back up. It took allot of trial and error for me to figure out the differences is settings between this and the previous QX. You may still have a dud, but I thought so at first also.

it's normal some cpus just have different temperatures on each set of cores

u will get it eva ! besides who needs sleep hahhahahahaa

yeah i had my sleep now.. again to figure this cpu out

swap the board out an try maximus formula or XE.....with appropiate bios.
you know the deal, I bet that qx9650 of yours would sing
under my P5K-Prem.(vmodded)......

my L739A663 under P5K. if you need more compares..

http://img87.imageshack.us/img87/163...testpo0jx8.gif

Have you experienced anything similar on other boards with your QX9650 i.e. crap oc's on one board but excellent on P5K-Premium of yours ?

[eva2000]

Well so far it's even worse on Gigabyte X38-DQ6 F6B bios with this QX9650 L739A644, 4000mhz instant reboot even at 1.428v even kept FSB low at 333FSB x12 to rule out FSB.

More tests to come

[Dynasty]

Have you experienced anything similar on other boards with your QX9650 i.e. crap oc's on one board but excellent on P5K-Premium of yours ?

well only with my Asus Striker II/780i board of mine thats it....
its an pita trying to get this thing going right with 8800 ultras/qx9650...

Both my P5K-Prem. & Maximus. SE run the qx9650 just fine..
if you try other boards and still same poor OC you know the deal....

[eva2000]

Yeah it's looking like a below average QX9650 compared to ones around

Flashed to F7D bios now for more tests..

@4000Mhz - 12x333FSB at 1.5v bios (1.488v cpuz idle -> fluctuating between 1.440v to 1.456v cpuz load)

First 2 hrs of Prime95 showed 1.440v constantly



Then changed to 1.456v in CPUZ...



1.440v - 1.408v = 0.32v more vcore needed than on Asus Blitz Extreme water cooled.




Leaving the cpu oc'ing aside for the moment, decided to figure out the divider and timing performance of the Gigabyte X38-DQ6 and I must say end result is pretty good even if the dividers don't always work on all straps variations.

Here's Super Pi 32M @9x445FSB (2.5A divider = 4:5) 556Mhz 4-4-4-8 at 2.48v (don't know why it needs so much vdimm as Corsair 10000C5D did 570Mhz 4-4-4-x at 2.35v on Asus P35 chipset boards.

Straight boot from bios (NO memset changes in windows)







Only 6.172 seconds slower than QX9650 @9x445FSB 1:2 DDR3-1780Mhz 7-6-5-18 1T on Asus Blitz Extreme (P35 chipset)

[xgman]

Eva, it is looking like your 9650 may in fact be on the slow side of average. Also, I've never had any luck with gigabyte anything. More power to you. Maybe you should consider my PM if that is possible. I alos find it rather interesting that the few 9770 es's that are around don't really seem any better than the 9650 es's, and in fact some seem worse. I think in this case the 9650 es's were really the top pick of what Intel can dfo with this process at the moment whereas the qx6850 es's ended up being just about the same as the retails.

[MorGo7h]

Good clocks

[eva2000]

Eva, it is looking like your 9650 may in fact be on the slow side of average. Also, I've never had any luck with gigabyte anything. More power to you. Maybe you should consider my PM if that is possible. I alos find it rather interesting that the few 9770 es's that are around don't really seem any better than the 9650 es's, and in fact some seem worse. I think in this case the 9650 es's were really the top pick of what Intel can dfo with this process at the moment whereas the qx6850 es's ended up being just about the same as the retails.

Yeah thanks for the pm.. it's hard even think of forking out more after experiencing the disappointment of this cpu's oc'ing potential on air/water

Good clocks

thanks.. just wish they were better heh

Looks like Gigabyte X38-DQ6 vcore droop gets worse the higher you set vcore in bios. At 1.55v bios set vcore you get 1.54v idle but 1.488v load. Still for 4.1Ghz mark, X38-DQ6 is requiring around 0.05v more vcore than what I needed on Asus Blitz Formula/Extreme.

Still depressing to see this QX9650 needing this amount of vcore just for 4100Mhz

Straight boot from bios (NO memset changes)

[eva2000]

Swapped out Corsair 10000C5D for SuperTalent PC2-8000 memory. Seems SuperTalent likes tight subtimings fine but still like previous tests prefers CAS5 to CAS4.









Starting to get the hang of tweaking memory performance on Gigabyte X38-DQ6 F7D bios.. Supertalent PC2-8000 do run a bit easier than Corsair 10000C5D as they can handle the much tighter timings and performance level.

All results below are done straight from bios boot up (memset only used for display purposes of timings)



Very close to sub 11min 32M time now

[oscar3d]

Very, very good...

[eva2000]

Thanks seems on Gigabyte X38-DQ6 my limit is around 4176Mhz while on Asus Blitz series around 4234Mhz - well short of those gems doing 4.5-4.7ghz air/water

[Aivas47a]

Eva I'm seeing higher than expected vcore to run qx9650 stable on Maximus Extreme. With 1.55 vcore in bios (1.53 actual) I am OCCT stable at 9x450. Chipset seems to require higher voltages than I expected as well, for stability at above speeds I need 1.6v NB and 1.54 FSB Term. I'll post some screenies tomorrow. Haven't tried 10x400 to see how much that will let me drop voltages. Also have P5K3 Deluxe board on hand, would like to try same chip in that board which has been very good quadcore overclocker.

High fsb (450+) also seems more difficult to achieve on qx9650 than on G0 Q6600, at least on Max. Extreme board.

EDIT: OCCT stability at 10x400 only seems to require around 1.48 vcore, .05 less than 9x450. Hate to give up the full 1800Mhz spec speed of my ram however!

[eva2000]

Interesting guess it does depend on the board for how it handles vcore... FSB seems to be more related to cpu.. this QX9650 L739A644 easily does 470-475FSB on cpu pll volts = 1.5v on Asus board.

Swapped to 2x1GB Crucial Ballistix Tracer PC2-8500 memory and seems to be the best memory so far for X38-DQ6

Decided to do some single and quad Super Pi 32M tests at QX9770 speeds of 8x400FSB = 3200Mhz at stock vcore in bios

Straight boot from bios. Seems for 2.66C (3:4) divider, cpuz v1.42 and memset as well as coretemp 0.9.5.6 don't seem to recognise the divider and memory clocks properly. Only app that does is Everest Ultimate









Now quad Super Pi 32M

[Langley]

I have one like this

I have a difference of 6-7°c in core 0 form the others core with coretemp and i am working for a prime stability at 430 x 10 1,46Vcore on liquid, but after 2h of prime the core 0 have a error, i think there's a lot of problem with temp about 62° on full after 1h... good test eva u have a pvt!

[oscar3d]

eve2000, I just want to say thanks....

I was at the moment thinking about how my QX9650 could perform at 3200 (8x400). I see the volts in your CPU-Z, how much VCORE are you seting for this config?

[eva2000]

I have one like this
http://www.pctunerup.com/up/results/_200711/20071128125151_PIC_0010.JPG[/img]
I have a difference of 6-7°c in core 0 form the others core with coretemp and i am working for a prime stability at 430 x 10 1,46Vcore on liquid, but after 2h of prime the core 0 have a error, i think there's a lot of problem with temp about 62° on full after 1h... good test eva u have a pvt!

nice clocks still 4.3ghz at 1.46v

eve2000, I just want to say thanks....

I was at the moment thinking about how my QX9650 could perform at 3200 (8x400). I see the volts in your CPU-Z, how much VCORE are you seting for this config?

1.25v bios set

Some 3dmarks with new video card

512MB Inno3D 8800GT OC Edition 650/1900 cards






System:

• Intel Core 2 Extreme QX9650 L739A644 Retail OEM
• CPU Cooling: Thermaright Ultra 120 Extreme + 120x25mm Delta 75cfm fan + MX-2 TIM
• Gigabyte X38-DQ6 F6B & F7D bios supplied Gigabyte
• 512MB 8800GT Inno3D OC Edition stock 650/1512/1900 with 169.02 WHQL drivers
• 2x1GB Crucial Ballistix Tracer PC2-8500 dual channel kit
• Memory cooling: 120x25mm Thermaltake 81cfm fan
• 160GB Seagate 7200.7 NCQ SATA port 2
• 2x3GB Gigabyte I-RAM raid 0 purchased from Auspcmarket.com.au
• Pioneer DVD-RW
• 620W Corsair HX620 PSU supplied by Corsair
• WinXP Pro SP2 Nlite Fully Updated.

The 512MB 8800GT Inno3D OC Edition cards come factory clocked at 650/1500/1900 according to GPUZ. But Rivatuner 2.06 reports them as 648/1512/1900.

Idle and load temps range from 41-49C idle to 59-62C load in an open desktop bench system.



I clocked my QX9650 Yorkfield to QX9770 Yorkfield quad core speeds @3200Mhz - 8x400FSB with 3:4 memory divider for 533Mhz 4-4-4-8 2T memory timings.

I only oc'd the GPU core leaving memory currently @1900Mhz

3Dmark06

@ stock 650/1512/1900 = 13,875
@702/1620/1900 = 14,410
@725/1674/1900 = 14,668
@738/1728/1900 = 14,780

3Dmark06 = 14,780
http://service.futuremark.com/compare?3dm06=3964423



Haven't tried GPU core clock higher than 738Mhz for now. I wanted to run some Pcmark05 to see how single 512MB 8800GT pairs with 6GB Gigabyte I-RAM Raid 0

Pcmark05 = 14,975
http://service.futuremark.com/compare?pcm05=1239691

Pcmark05

@ stock 650/1512/1900 = 14,699
@ 738/1728/1900 = 14,975



Then just bumped multiplier from 8x to 10x

3dmark06 @738/1728/1900 = 15,501
http://service.futuremark.com/compare?3dm06=3964806



PCmark05 = 17,021
http://service.futuremark.com/compare?pcm05=1239759

[oscar3d]

Because of the pictures you've just posted I'm hating you

I'm having spasms of envy.... those superclocked GT's look gorgeous!!

So 1.25V freaking AWESOME!! Thanks for the numbers, gonna try them later, and see if I can push for lower volts...

I see you just turned your QX9650 into a QX9770 right on the spot

Are you planning this as yourfinal 24/7?

[eva2000]

Yeah these 8800GT are very nice

24/7 i prefer 3600-3800Mhz heh - 8x450FSB or 8x475FSB if possible

Putting it altogether GPU/Shader/Memory clocks for 512MB Inno3D 8800GT OC Edition I managed to easily bench at 742/1836/2052 with exception for 3dmark2003 I ran with lower shader clock of 1783mhz.

Managed to easily break 16k barrier for single 512MB Inno3D 8800GT OC in 3dmark06 and pulls off very decent runs for 3dmark05/2003 and just under 75k mark for 3dmark2001

• 3dmark06 = 16,088
• 3dmark05 = 24,427
• 3dmark2003 = 45,860
• 3dmark2001 = 74,961 (LOD = 4), 72,188 (no LOD)
• AM3 = 245,219

[eva2000]

System:

• Intel Core 2 Extreme QX9650 L739A644 Retail OEM
• CPU cooling = Dtek Fuzion Rad Tower Box
• NB chipset cooling = Radiical Chipset Block/Corsair Cool single 120mm fan Rad/MCP350
• Asus Blitz Extreme 0904 bios
• 128MB Gainward FX5200 PCI
• 2x1GB OCZ PC3-14400 Platinum - DDR3-1800 rated at 8-8-8-27 at 1.9v
• 500GB Seagate 7200.10 SATA on SATA1 Port
• LiteOn 165H DVD Burner
• Sony FDD
• 1KW OCZ ProXStream
• WinXP Pro SP2

Moved QX9650 back to Blitz Extreme with Dtek Fuzion and now core 0/1 temps no longer move even under load LOL

idle



load



similar core 0/1 vs core 2/3 temp differences on X38-DQ6 http://i4memory.com/showpost.php?p=80956&postcount=27 but with lower temps on Dtek Fuzion vs Thermalright Ultra 120 Extreme air

Immediately after ending Prime95.. temps



With Dtek Fuzion and Asus Blitz Extreme I can definitely clock my QX9650 L739A644 higher than before. Max on Gigabyte X38-DQ6 on Thermalright Ultra 120 Extreme was 4176Mhz at 1.60v idle/1.568v load for Super Pi 32M. On same air cooling with Asus Blitz Formula was around 4234Mhz at 1.560v.

On Asus Blitz Extreme so far at 4200Mhz at 1.552v idle / 1.544v load

Comparing Super Pi 32M and Everest numbers

Asus Blitz Extreme - broke 11min barrier easily





Gigabyte X38-DQ6

[eva2000]

Finally some decent numbers cpu clock wise... still about 250-400mhz behind other QX9650 cpus I've seen in terms of clock speed though

Wprime 32M @4400Mhz



Super Pi 1M & 32M @4330Mhz





512MB Inno3D OC Edition stock @742/1836/2000
3dmark06 = 16,033



Interesting that my run on Blitz Extreme scored lower in SM2/SM3 scores than my run on X38-DQ6 with cpu at nearly 200mhz lower cpu clock. Could be the higher FSB on X38-DQ6 @9x456FSB and/or the higher memory clock on video card at 2052mhz vs 2000mhz