GA-EP45-UD3L + Q9550 Vcore

[RoscoP]

Good evening fellow enthusiasts. I recently purchased a Q9550 from my local MicroCenter :- ] (If you're going i7, go to Micro Center! They have the i7 920 for $199 boxed retail!). I dropped my Q9550 into my GA-EP45-UD3L, reset CMOS, and then fired up. It posted no problem, and I went right into BIOS. My main question I have is - what Vcore is normal for a Q9550? My GA-EP45-UD3L gives my Q9550 a "Normal" (same thing as AUTO?) Vcore of 1.28750v. I'm assuming if I left my Vcore set to AUTO (not Normal) it would put my Q9550's Vcore to 1.28750v?

My buddies Q9550 AUTO's to 1.22v, but he has a DFI X48 board. His CPU is completely stable at that vcore. However, I tried manually setting my vcore closest to 1.22v and I could not obtain stability (Doing Large FFT's on Prime95). Did I just happen to get a Q9550 that requires more vcore than some do? Is it true that every chip has a different Vcore given to it by the AUTO setting on the same motherboard? I'm not planning on overclocking this Q9550 until I can find out the lowest vcore I can run it at stock. I am plenty happy with the performance from this CPU at stock, but I do plan to at least up my FSB to 400 and run my RAM 1:1 eventually. I'm currently testing for stability on my Q9550 @ stock w/ a Vcore of 1.25000V.

If anyone has any suggestion or input I'd greatly appreciate it. Thank you.

[NimrodAUS]

This is a question that I have seen around the net. I was under the belief that the VID for 45nm Chips was 1.25V. Thats what I run both mine at. And that is what Both there AUTO/Normal Values are. However I have seen some people talking about High and Low VID Chips. So I Assume you have just got one that need slightly more volts to achieve the stock stable speed then others. I have seen some with as you say 1.22v I would be suprised if you cant get stock stable @ 1.25v. Also you may get a high overclock on the VID of 1.28750v Which is by no means a high voltage for that chip. If you are not overclocking leave it on Auto and when in windows use CPU-Z to see what volts it running at. If EIST and C1A are enabled in the BIOS it will throttle back the volts when not using much CPU power.

[RoscoP]

I have another question... If I were to put this same Q9550 in another motherboard, it would AUTO Vcore to 1.28750v also?

I went ahead and set Vcore to AUTO in BIOS and I am going to post a screenshot of CPU-Z below.



Does anyone think I should try and find out if my chip is stable with a lower Vcore @ stock? I think this chip would run at a lower Vcore and be stable. I'm afraid that my board is giving more Vcore to the CPU than is necessary. I know that 1.28750v is within reason.. but is it reasonable for stock speeds? CPU-Z reports a Vcore of 1.232v.

[Chike]

I don't see the reason to even care what the stock vcore is. What meaning does it have?
At my current overclock my vcore is still lower than your stock but the temps are 67c with full load IBT, and I see people with higher VID and overclock and lower temperatures. If I set my vcore as yours, temps will probably be well over 70c even at stock speed. I am not at all sure that lower VID is better than high VID.
It is probably factory calibrated with safety margins of allowed vdrop and other things, seems very reasonable to me to trust Intel to know what they're doing.

[RoscoP]

Thanks for the input Chike. I am still curious though, would this Q9550 have the same AUTO Vcore if I put it in another P45 motherboard? Nobody thinks I should try and find the lowest Vcore I can run at stock speeds? Sorry for such basic questions. I'm probably going to start pushing this Q9550 soon, but I was wanting to find this info out first.

Thanks for the help.

[Psycho101]

While VID isn't a sure fire way to tell if you have a good chip, I'd put money on a CPU with a VID of 1.2V clocking on lower volts than one with a VID of 1.3V.

Look at the proof you have in front of you Chike. You may have seen people with a higher VID chip run higher overclocks with lower temps, but you won't see many E7200's with a VID of 1.062V, or that are able to do 3.4GHz on a mere 1.12V. For instance, my CPU needs 1.3V to do 3.4GHz, and an incremental rise there after to get to 3.8GHz. It then needs a disproportional boost in Vcore for another 100MHz gain.

VID's are completely individual for each CPU. They are tested and set individually. Intel set a VID for the purpose of auto to set a voltage that will definitely work at stock speed. It does, no doubt, contain a margin of safety. The margin of safety doesn't directly include Vdroop, as that would be an integral part of any CPU test. Intel would use equipment that droops under load because it's in the design specs. The resulting VID would then be set to the voltage the CPU needed - any droop and + a little extra to be sure and to counter higher amounts of droop on individual boards.

If we all trusted Intel to know what they were doing, we wouldn't be running CPU's out of spec. We'd all look at the rated speed and just "trust" them.

Your chip, if the VID is indeed 1.28750V, would be set to that in any compatible socket 775 board, P35, P45, X38, X48 or any of the older 9xx series boards too. To confirm your VID, run CoreTemp. It will show the CPU's hard coded VID value on the main screen.

There's not much benefit to trying for the least amount of volts at stock unless cooling is an issue, or you want to run the CPU in a HTPC, in which case you probably wouldn't choose a quad anyway. If you want to play, I'd be trying to find my max overclock using the standard VID (set manually in the BIOS, no settings on auto OR normal) and work from there.

[RoscoP]

Sorry to resurrect this post but I think my questions I have can be answered here.

I've started trying to see how high my fsb with this Q9550 can go on this P45 board. I never was able to get my old E5200 past 347 fsb. I didn't know how much vcore on the mch is necessary to get to 400 fsb. So far I am stable at 400 fsb with 1.24 v mch, but I needed to up my RAM voltage to 2.0v instead of 1.94v. Is 1.24v mch a reasonable voltage for 400 fsb? I have heard tweaking your GTLs and your VTT can possibly allow you to run lower mch voltage.

One crazy thing I noticed in switching between 333 fsb and 400 fsb. My RAM with 1.94v @ 800 4-4-4-12 would pass Prime95 Blend no problem for 10 hours. Whenever I moved up the to the 400 fsb and raised tRD to 8 (from tRD 5 @ 333), I had to increase voltage on the RAM to 2.00v to be able to pass Prime95 Blend for 10 hours. However, I did not need more mch voltage in going from 333 fsb to 400 fsb (1.24v mch)

I thought going to the 400 fsb would need more mch voltage, and not more RAM voltage. I can't clarify if I needed more mch voltage, because I was running 1.24v w/ 333 fsb.

I did notice that whenever I tried dropping my tRD any lower than 8 on 400 fsb resulted in a failure to POST, so I must have met the point to where I need more MCH voltage for the lower tRD. Does anyone think that a tRD of 8 is fairly optimal for 400 fsb 1:1 on the 333 strap?

I do plan on putting a northbridge fan on my P45. Everest reports my motherboard temp of 40C idle (is that the chipset temp?).

Any input or comments will be greatly appreciated.

[Conners]

I'd say your doing alright with your voltages. You might consider trying 2.66D (400MHz) and would give you 1064 ram freq. That is unless you just want to run a 1:1 strap.

Also how high will your mulit go and what id your overall goal? I think we can do a lot better than that if you like.

[Chike]

You had the RAM under it's voltaged spec at 333FSB, and you still have it under voltaged. OCZ page says 2.1v are needed, with EVP up to 2.15v, what's on the labels?
Have you tried higher voltage with lower tRD?

[RoscoP]

Chike, I have not tried a lower tRD and adding more voltage to the RAM. The OCZ RAM I have is rated 1066 @ 2.1v I don't need 2.1v to pass memtest @ 800 4-4-4-12.

Is my current tRD (8) optimal for the amount of MCH voltage (1.24) I'm using? Maybe I ought to try and lower tRD and increase voltage to my RAM?

Good news - I was able to do 400 x 7.5 = 3Ghz w/ stock vcore (1.28750v) - Prime95 Large FFT's for almost 15 hours w/ no errors.

I just got some zip ties to put my NB fan on - This is going to be a fun little project... I'll post some pics of the fan once I get it mounted. I'm looking forward to seeing some lower P45 chipset temps :- ]

Wish me luck :-P - Thanks for the replies guys - I will post back soon.

/edit

Connors - My max multi is 8.5 - I'm shooting for 3.2 Ghz (400x8 - preferably without a vcore increase)
About the RAM - I think I would see less performance with my RAM @ 1066 5-5-5-15 versus 800 @ 4-4-4-12. I could be wrong - the only way to know is to run Everest. I will probably do this.