Simviation Forum - Core i7 overclocking raise mult, BClock, or Both?

Reply #30 - Jan 25^th, 2009 at 12:35pm

NickN Offline
Colonel
FSX runs fine... the problem
is you or your system

Posts: 6317

Jim right now the clockers are still working out the variables around these slugs and what they will and will not allow

The boards and BIOSs are still going through their growing pains too

you are playing with cutting edge hardware which is still in the dev phase AND in the phase for reverse engineering the ability out of it by the pros.

Do not expect me or anyone else who is a responsible engineer to assist you doing things only those who know what they are doing and are willing to lose their investment would do.

I am going to be quite frank.. and this IS a woodshed statement, … your post is whine that I would not assist you in blind suicide and putting it on me that I somehow told you to FO

I hope you see that now

When more information comes around and there have been confirmations on the use of settings which will pull more out of the chips without risk of loss, I will post and suggest it, not until.

Back to top

IP Logged

Reply #31 - Jan 25^th, 2009 at 1:03pm

raptorx Offline
Colonel
There's too much confusion...
San Diego, CA

Gender:
Posts: 434

NIck, I wasn't trying to put words into your mouth or twist what you said. I was just trying to express how I felt when I read what you said, I hope that makes sense.

I am very attuned to the chip's limits. I have no desire to push anything beyond cool, stable limits. I stopped at 1.4v on the core because I'm unsure how higher vcore will affect the useful life (for me 2 years tops). I see uberclockers pushing way beyond that but I'm not going there. Sure, 1.55v is the absolute max but how will it handle 1.45v provided temps stay under 80 deg C? Am I too conservative at 1.40v?

And I've noticed things change in Everest when I try different multiplier/bclock combinations. When the Bclock is higher, the latencies drop by as much as 15-20%. For instance, going from 30x133 to 29x140 and latency drops by 5ns! But when I start to raise the Bclock It's really hard to get stable. Am I just not pushing my voltages enough? I know I should not use Everest as much of a guide but it does show differences with different settings.

Do you run a default Bclock or do you drop your cpu multiplier one or two steps and raise the Bclock a little? How about the uncore clock, can I raise that a little (3200 up to 3333)?

As far as my BIOS, For 4.0GHz I use the XMP profile @ DDR2 1600 with 30x133, 1.375 vcore, 1.356 vQPI, 1.66 vDIMM, loadline enabled, SS disabled. Everything else on Auto.

And another thing: I have C1E disabled but C-state enabled because the Everest memory bandwidth jump so much when C-state tech is enabled. It goes up by 2500mb/s! Is that real? Or an artifact of Everest?

I'm NOT one of the clueless who's looking for irrational settings to squeeze every last MHz out of things. I know you didn't call me that but I want you to know I can be trusted to recognize the right way to test this stuff...that's why I'm asking all these questions. If I wanted to thrash I'd just start pushing voltage limits like the guys on Extreme Systems. But I know better. So I come here and ask you. I'm just looking to find out some of the things you've found out, even if it's not anything dramatic.

Sorry for being so sensitive.

-Jim

Rampage II Gene, i7 965 4GHz
Mushkin Redline DDR3 1600
XP x64 SP2
ASUS Matrix GTX285

Back to top

IP Logged

Reply #32 - Jan 25^th, 2009 at 1:18pm

Vodka Burner Ex Member

raptorx wrote on Jan 25^th, 2009 at 2:36am:

Well, I can't get my system stable at anything beyond 4.0GHz @ 1.4 vcore and 1.375 vQPI. I've tried a few different multiplier/bclock combinations and nothing works. Temps are no issue, nothing get's beyond 65 deg C so far. Memory runs between 1600 and 1700 no problems according to hours of memtest. So it's got to be the cpu.

I'm a little dissapointed in this chip now. Huh

I thought I would at least get to 4.2 GHz eventually w/o too much trouble. All I'm going to get is 4.0GHz out of a 965???

-Jim

Dude. You have a $1000 processor, running significantly faster than stock. It's a freaken supercomputer, and most of us would kill to even get an i7 920.

Enjoy the PC.

Back to top

IP Logged

Reply #33 - Jan 25^th, 2009 at 1:25pm

raptorx Offline
Colonel
There's too much confusion...
San Diego, CA

Gender:
Posts: 434

Quote:

raptorx wrote on Jan 25^th, 2009 at 2:36am:

Well, I can't get my system stable at anything beyond 4.0GHz @ 1.4 vcore and 1.375 vQPI. I've tried a few different multiplier/bclock combinations and nothing works. Temps are no issue, nothing get's beyond 65 deg C so far. Memory runs between 1600 and 1700 no problems according to hours of memtest. So it's got to be the cpu.

I'm a little dissapointed in this chip now. Huh

I thought I would at least get to 4.2 GHz eventually w/o too much trouble. All I'm going to get is 4.0GHz out of a 965???

-Jim

Dude. You have a $1000 processor, running significantly faster than stock. It's a freaken supercomputer, and most of us would kill to even get an i7 920.

Enjoy the PC, not like it couldnt do anything.

I know, I don't want to be ungrateful here. I was able to get this platform because of a very generous Christmas bonus given to me by my company. I've got no complaints on how my system performs right now.

I'm letting my desire to find the working limits of this chip (moderate limits that is) overshadow its present capabilities.

-Jim

Rampage II Gene, i7 965 4GHz
Mushkin Redline DDR3 1600
XP x64 SP2
ASUS Matrix GTX285

Back to top

IP Logged

Reply #34 - Jan 25^th, 2009 at 1:30pm

NickN Offline
Colonel
FSX runs fine... the problem
is you or your system

Posts: 6317

raptorx wrote on Jan 25^th, 2009 at 1:03pm:

NIck, I wasn't trying to put words into your mouth or twist what you said. I was just trying to express how I felt when I read what you said, I hope that makes sense.

I am very attuned to the chip's limits. I have no desire to push anything beyond cool, stable limits. I stopped at 1.4v on the core because I'm unsure how higher vcore will affect the useful life (for me 2 years tops). I see uberclockers pushing way beyond that but I'm not going there. Sure, 1.55v is the absolute max but how will it handle 1.45v provided temps stay under 80 deg C? Am I too conservative at 1.40v?

And I've noticed things change in Everest when I try different multiplier/bclock combinations. When the Bclock is higher, the latencies drop by as much as 15-20%. For instance, going from 30x133 to 29x140 and latency drops by 5ns! But when I start to raise the Bclock It's really hard to get stable. Am I just not pushing my voltages enough? I know I should not use Everest as much of a guide but it does show differences with different settings.

Do you run a default Bclock or do you drop your cpu multiplier one or two steps and raise the Bclock a little? How about the uncore clock, can I raise that a little (3200 up to 3333)?

As far as my BIOS, For 4.0GHz I use the XMP profile @ DDR2 1600 with 30x133, 1.375 vcore, 1.356 vQPI, 1.66 vDIMM, loadline enabled, SS disabled. Everything else on Auto.

And another thing: I have C1E disabled but C-state enabled because the Everest memory bandwidth jump so much when C-state tech is enabled. It goes up by 2500mb/s! Is that real? Or an artifact of Everest?

I'm NOT one of the clueless who's looking for irrational settings to squeeze every last MHz out of things. I know you didn't call me that but I want you to know I can be trusted to recognize the right way to test this stuff...that's why I'm asking all these questions. If I wanted to thrash I'd just start pushing voltage limits like the guys on Extreme Systems. But I know better. So I come here and ask you. I'm just looking to find out some of the things you've found out, even if it's not anything dramatic.

Sorry for being so sensitive.

-Jim

no problem

Here is what I can relay right now and will come back to this later when I get more information

You should disable C state.. its only needed when TURBO mode is enabled and will cause the system to fluctuate in multiplier

Yes, you can tickle the latency by pushing BCLOCK and reducing multiplier.. you must remain in voltage ranges to do this. Its how I run DDR3 2000 product right now but there are quarks in it I am still trying to figure out.

The bottom line to all this tweaking and fiddling is not in the benchmark scores.. its in how does that translate to FSX? You can work things several different ways with the 965 but that does not mean it will translate better one way or another just because a benchmark value looks better.

I dont sit here all day long and play with i7 and my time is very limited with it right now since I am focused on GEX. Its way too soon for me to post DO THIS and DO THAT suggestions Jim. I need time to work out what is making a difference and what is simply making a benchmark program show better results.

Although I do have one i7 tower here pushing voltages it probably should not and running partial load to see how long it may go PooF! I already burned one 920 and a 940 engineering sample.. I am on my 3rd slug with that test tower.

There are things we simply do not know yet Jim. For me to tell you its OK to jack up the Vcore to 1.55 would be utterly irresponsible until I can confirm the rest of the settings which may compensate for that increase keeping the chip safe, and, if its really worth it

1.45v is out of range from Intel who sets that max at 1.375v

http://processorfinder.intel.com/details.aspx?sSpec=SLBCJ

and lets get something very clear.. it is a 45nm CPU built on Penny tech.. that means it probably has the same voltage limits as a Q9650 in which 1.40 was the max I would ever suggest online in forums someone use with a Q9650.

Are you asking me if it’s safe to go to 1.45 and leave it there?

I will let you know when I confirm the slug being pounded here in a test tower is not dead in some weeks.. or I get positive reports from others with the same findings..

Until then… its all on you.

Back to top

IP Logged

Reply #35 - Jan 25^th, 2009 at 2:06pm

NickN Offline
Colonel
FSX runs fine... the problem
is you or your system

Posts: 6317

One other thing too...

Look at the link above and check the THERMAL SPECIFICATION

its blank

Intel, although has hinted.. never posted the safe thermal limit for the i7 9x series so be quite aware that what has been posted on the net and in the clocking fourms is in fact a GUESS @ 100c and a top end of 80c for proper operation

There is no science in this at this point... we are all functioning on reverse engineering testing and trial

http://download.intel.com/design/processor/datashts/320834.pdf

look at page 79

@ 130w that places the max CASE temp at 67.9c

The case is NOT the internal diode temp and that is typically 10c higher (average)

which places the ESTIMATED or GUESSED TJMAX at between 95-105c

we took 100 to hit both

The safe temps that have been posted are purely an educated GUESS and we may ALL be wrong

cuz a QX9770 has a 136w spec and its max is 70c!

Back to top

IP Logged

Reply #36 - Jan 25^th, 2009 at 2:22pm

raptorx Offline
Colonel
There's too much confusion...
San Diego, CA

Gender:
Posts: 434

I really appreciate it Nick.

I put a box around my tweaking to try to keep things safe. I say I get frustrated when I can't breach certain values and still stay within that box. Like that 1.4v vcore. But that's a good thing because it will keep me from allowing myself to "go for it". I get discouraged instead, but that's a result of having to accept keeping it safe. The system will appreaciate that more than I ever will.

Thanks for being patient and understanding and keeping me informed. When I came into overclocking the Core2 was well known and there was time-tested data on the operational limits. It was easy to know better what to push and how high to push it.

Now it's new for EVERYBODY and it's hard to reign in my enthusiasm.

-Jim

Rampage II Gene, i7 965 4GHz
Mushkin Redline DDR3 1600
XP x64 SP2
ASUS Matrix GTX285

Back to top

IP Logged

Reply #37 - Jan 25^th, 2009 at 3:14pm

NickN Offline
Colonel
FSX runs fine... the problem
is you or your system

Posts: 6317

you gotta give the platform some time.. by the time the Q9650 was out in full swing on the market we already knew almost 5 months before how that was going to unfold in tweaking and limits based on the established tech

there is no established tech for any of this in i7

Look at page 23 in that link above

Processor I/O supply voltage for DDR3: 1.575 and you are using 1.65.. we know that is a safe limit because Intel spilled the beans to the motherboad companies.. what we DONT know is all the safe limits in that spec sheet but are working them out.

Back to top

IP Logged

Reply #38 - Jan 29^th, 2009 at 12:13pm

raptorx Offline
Colonel
There's too much confusion...
San Diego, CA

Gender:
Posts: 434

Well I tested my system with Linpack today. I ran 5 passes max stress and everything passed. Temps never got over 63 deg C (ambient at the time was around 70 deg F). I'm at 1.375 vcore, 1.356 vQPI in the BIOS, all else on auto.

So I've got nothing to complain about AT ALL! Embarrassed

« Last Edit: Jan 29^th, 2009 at 4:21pm by raptorx »

Rampage II Gene, i7 965 4GHz
Mushkin Redline DDR3 1600
XP x64 SP2
ASUS Matrix GTX285

Back to top

IP Logged

Reply #39 - Jan 29^th, 2009 at 3:57pm

NickN Offline
Colonel
FSX runs fine... the problem
is you or your system

Posts: 6317

nope

them's some dang nice numbers!

Most would kill to get that...

You do get what you pay for

Back to top

IP Logged

	Search the archive: