What is Intel's recommended VTT? 1.35V?
Printable View
What is Intel's recommended VTT? 1.35V?
LOL :rofl: believe me, I tried my best to formulate it in a moderate way, because I knew you would say this. :)Quote:
Originally Posted by Movieman
I'm keeping the VVT=1.32v for now, :yepp: just trowing some ideas because all X56xx-owners seams to be in the thread.
1.4vQuote:
Originally Posted by =[PULSAR]=
run 1.4VTT and 1.4Vcore and be prepared to go and purchase a coffin for that cpu..one or the other but not both..
My X5680's are at 1.3675vcore and 1.3VTT
They will live to a ripe old age...like me!:p:
Actually, for 32nm it is 1,4V Vcore and 1,3125V VTT (980X specs).Quote:
Originally Posted by =[PULSAR]=
45nm Bloomfields were 1,55V/1,375V Vcore/VTT.
So yeah, 1,4 and above is definitely too much VTT for 24/7.
According to Intel's data sheets, the maximum safe limit for VTT is 1.35. This is for i7-900 series CPUs. I think I read that the VTT max is much lower for LGA1156 processors, something like 1.21.
http://download.intel.com/design/pro...hts/320834.pdf
The one I've seen at the first days of 980x-release, (and I saved it on my local drive), says 1.4v for max VVT. Has Intel changed this presently?
http://img.techpowerup.org/100415/Capture032961.jpg
Weird. Check this document, page 23:
http://download.intel.com/design/pro...hts/323252.pdf
Clearly says 1,4V max Vcc - and I am 100% sure on that. 1,55 will kill most 32nm chips on standard cooling.
This also states 1,4V for max VTT...
I searched around for data sheets on Gulftown and Westmere chips but couldn't find any. Sometimes Intel releases the data sheet after it launches the product.
This is the main page on Intel's site for the X5650.
http://ark.intel.com/Product.aspx?id...ec-codes=SLBV3
On the right it says "No Datasheet Available".
That document is dated March 2010. The one I posted was from February. So maybe Intel had a change of heart and decided to increase maximum allowable VTT.Quote:
Originally Posted by jcool
You are right, but these are almost identical to 980x. Probably with a better and more resilient Uncore that can take more torture and even higher VVT too?. Hopefully yes.Quote:
Originally Posted by sierra_bound
But I believe MM has a good point too, it is a good idea to be careful until we get to know tese chips better. But om the other hand, the VVT is the holy grail of high OC on these chips, and we should try yo figure out what they can really take in real life.
They sure seem to change numbers a lot. I distinctly remember seeing 1.3125V as max. VTT somewhere :shrug:Quote:
Originally Posted by sierra_bound
Worth noting is the fact that these are MAXIMUM voltages, Intel does NOT sanction the sustained use of said peak voltages. They are talking about voltage spikes with the CPU at stock clocks (and probably idle) - the high currents our overclocked and fully loaded Hexas are pulling make for a whole different environment.
Lets agree, right now Intel says max VVT=1.4v for 980x.
Intel has been conservative enough to reduce to vCore to 1.4v (from 1.55v),
does this VVT-increase means Intel has improved the Uncore dramatically?
If true, Intel said max VVT=1.35v for i7 45nm, and it has proven to take even 1.5v ,easily, for 24/7 use.
It is to early too say anything for sure about 24/7 VVT, but I believe there is a good hope that these chips can take a lot of VVT-torture. On the other hand, I agree with MM that we should be careful until we get to know these chips.
I guess those constant changes are due to the different CPU steppings and the advancements made in the manufacturing process.
I killed a Q3QR A0 hexa after ~10 minutes on SS (-30C), and I was only running Cinebench at ~1,65V Vcore. So it stands to reason that the same CPU would have died running WCG @ 1,55V Vcore on water.
As I said, I would not recommend going over 1,4/1,35V Vcore/VTT for long term crunching unless you are prepared to face the consequences.
If vCore has killed your A0, does it means that VVT didn't kill it?Quote:
Originally Posted by jcool
I'm not recommending to use too much VVT, just suggesting that Intel's recommendation on VVT-increase for 980x, plus these chips probably has a better and more resilient Uncore than 980x, plus i7 45nm ability to take 1.5v VVT for 24/7, all togteher suggests that these chips may be able to take a good deal of VVT for 24/7. But nobody knows how much torture they can take yet, and it is a good idea to be careful.
Not sure what killed it, but I didn't run high VTT so I am guessing it must have been Vcore.
Don't forget the 5650's are still B0, and thus potentially more vulnerable to high voltage. If you read FUGGER's OC report on Gulftowns, he said the same things really... A0 die pretty easily, B0 are a lot tougher but still not perfect, and B1 he found "to be unkillable under controlled conditions" (-100C and colder) so that gives at least some security.
I used 1,52V vcore and 1,43V Bios VTT for short-term benching on the 5680, but I turned down the chiller beforehand so the CPU never exceeded 30 or 35C.
Plus I only ran short tests like 3DM Vantage and Cinebench, crunching is a whole other story.
I see what you mean. My X5650 is a B1 :cool:.
I will thank you all for your inputs. It has been a great help to get a better idea about these chips. :)
Sorry guys if this got OT for others, I couldn't find this X56xx-gang all together in another thread. So when i found them, couldn't resist to push (and rub) it. ;)
Well I've got this up and running and will let it contribute for the next month on my i7
http://img191.imageshack.us/img191/8...unitygridy.jpg
Isn't your X5650 a B0 not B1?Quote:
Originally Posted by Sam_oslo
Thanks for your help..:up:Quote:
Originally Posted by Adhmuz
It is a B1, according to EVEREST. A Model C, Stepping:2, according to CPU-z too :) Both means it is B1 ES :clap:Quote:
Originally Posted by Movieman
EDIT:
http://img.techpowerup.org/100416/Capture034.jpg
Well, just finally joined (I used SATI years ago), with the laptop in my sig :D I know it's not much, have to run it at 70% or the fans are 100% all the time, and it'll take ~8 hours per task, but I'm glad to help.
I also can't assure 24h, but I'll be there for the May week, let's break some records!
I started crunching on an iBook G4 so don't feel bad. It's the thought that counts. :up:Quote:
Originally Posted by mursaat
One question:
I can think to use this to test the stability of my new OC, it seams to be as good as Prime at 100%, but then it may BSOD frequently.
Would a BSOD harm running tasks dramatically? or just those last minutes?
A BSOD may damage the WU and cause it to "Compute Error" and upload for no credit. You can also get "Compute Errors" if your OC is not quite stable enough to process properly, yet not bad enough to BSOD.Quote:
Originally Posted by Sam_oslo
If you want to use WCG to test stability then I would suggest you follow the steps below carefully because if you piss the WCG server off it will shut you off from getting new WUs to crunch temporarily (kind of like a quick 1 day ban here on XS except it is automated so you can' t try and talk your way out of it:rofl:).
From BOINC Manager ...
1. Suspend Network communication
2. Suspend BOINC
3. Turn BOINC off completely. When exiting say "Yes to unload when closing", if you do not get this question then go into Options, Advanced and recheck the exit question box.
4. Copy the BOINC data directory
5. Apply your OC
6. Open BOINC Manager and turn activity to Run Always (leave network communication off)
If you get lots of errors then you can stop BOINC completely (like in step 3) and replace the data dir with the copy you made in step 4 (make sure to create another copy).
If you are good to go (no errors) then turn Network communications back on.
Rinse and repeat for each round of testing.
The reason for this method is that you are limited by the number of WUs per core you can download in a day and if you are OC testing you can throw compute errors very quickly (even without a BSOD) and you will then be shut off until 12:00 AM UTC at which time WCG will give you another chance ... BUT ... that chance makes you earn the number of WUs it will download based on your ability to complete and upload valid WUs.
That vs. I have never had a WU error if my system can Prime Blend with custom settings that use all free RAM for 2 hours.
Your choice ... let us know how you do :up:
^^ Oops, I didn't know all this. Thx for good explanation, good to know. :)
I'm glad I asked before trying it. ;)
Hi everyone,
Add me to the list as well.
My system should add a few results as well
Core I7 at 4400 mhz :)