Don't be afraid until you know the facts ;)Quote:
Originally Posted by Hornet331
Printable View
Don't be afraid until you know the facts ;)Quote:
Originally Posted by Hornet331
You might remember a previous post when i had problems getting above bclk 172.Well i have a new i7 920 which booted straight away @180!Anyway i have another strange problem.In the bios all 3 channels are showing for the memory but only 4096 is displayed at post:mad: ,when i booted into Vista x64the o/s is only showing 4gig.If i look at cpuz it is showing 6gig and triple channel.Anyone else had this problem?
If have done the following.
Reflashed and cleared cmos
Reloaded bios defaults.
All settings @ stock (when i first installed the cpu i used dram v1.66 and no more than 1.275 on qpi/vtt to test bclk 180)
Default volts:
cpu 1.2
dram 1.54
qpi/vtt 1.175
Re-seated the memory
swapped the sticks around
tested each stick in ddr3_1 slot -each test displayed 2gig
tested dual channel in dd3 3_1 and ddr3 3_3 -displayed 4gig
with all three channels only 4gig is displayed on post.
The memory appears to be ok.
Its looks like i have found the problem.One of the pins at the bottom of the socket 1366 was slightly bent:eek: .Not sure how i carefully put the cpu in and it only goes one way!After bending it back in postion (not easy used a small needle pinhead) its posts with 6gig.
Hey guys,
I'm just getting my head around overclocking this platform and found something slightly odd. It seems that in some cases more qpi/vtt voltage equates to lower stability.
At this point in time I'm testing @ 193x20, 1534RAM, 3454QPI (can't recall uclk right now, but it's in line with the 8:9 association) with turbo disabled (until I get a feel for the chokepoints of the rest of the platform) and it's stability wasn't 100%. I had the qpi/vtt volts at 1.3125 so I thought I'd try 'em at 1.325. To my surprise the system was significantly less stable. I lowered the volts than to 1.3v last night and have been stability testing successfully since. For reference, my vDram is 1.64v, my vCore is 1.2565 and my PLL is 1.9.
Has anyone else experienced this?
how are you stability testing?
1.325vcore, 1.86vpll, 1.3v qpi/vtt, 1.66v dram might be a good place to start. I think you're aiming at too little vcore and too much pll off the hop. unclk should be 16x - 18x max.Quote:
Originally Posted by Truckchase!
Quote:
Originally Posted by tonytiger
That's pretty interesting. I've had the same issue with my P6T Deluxe re missing 2 GB of memory, but having it show up in CPU-Z and the system running perfect in every other way. A number of others have reported the same thing in the ASUS forums.
Seems odd that a bent pin would show okay in CPU-Z, but not otherwise ... but great that it's working for you. I'm looking at an RMA, I think.
Two new results with relatively low Vcore and overclocked Dominator 1600MHz sticks (6GB)...
First image is 22x180MHz (3960MHz), 1804MHz DRAM 8-8-7-20-1T, 1.30000V CPU, 1.35000V QPI/DRAM, 1.65681V DRAM (all BIOS settings). The configuration is both memtest and stress test stable.
The second image is 21x190 (3990MHz), 1904MHz DRAM 8-9-7-20-1T, 1.28750V CPU, 1.43750V QPI/DRAM, 1.65681V DRAM (all BIOS settings). The configuration is both memtest and stress test stable. This config I am not 100% comfortable with as the QPI is much lower and significantly lower than Uncore, but the memory OC is just insane.
Thoughts? Which one seems better? The second has a better memory OC, lower CPU volts, but much higher QPI/DRAM volts than the first. Maybe I should skip all of these and settle for 21x180 (3780MHz), 1804MHz DRAM 8-8-7-20-1T, with 1.21250V CPU, 1.35000V QPI/DRAM, 1.65681V DRAM (BIOS settings)...
Edit: I guess I forgot to mention that I am now striving for the best balance between OC and voltages for 24/7, as opposed to raw clockspeed. After I managed to do 3.8 with only 1.2V I just cannot bear to force this chip to run at 1.4V for only 400MHz more.
what a difference the TRUE makes, i can do a moderate OC to 3.2GHz and temps won't pass 62C on prime95 w/ HT, maybe I can finally do 4GHz 24/7 on AIR!
4GHz is going to be a lot hotter, and I mean A LOT.Quote:
Originally Posted by ap4lifetn
If you are using 6 sticks of Ram are you going to need a higher VDimm or qpi/vtt voltage?
heh, idle 36C, load 80-84C, my cores are splitQuote:
Originally Posted by dejanh
I'm pretty sure that with 12GB I would have to tone down the OC, but I am quite confident that 1800MHz would still be quite possible. Maybe 1900MHz would be a bit much. I do not think that there would be much change to QPI/DRAM voltage, or any for that matter. DRAM voltage, I have no idea as it seemingly has no effect on anything for me, at least in all of the tests so far.Quote:
Originally Posted by Lukee
All in all, not planning on getting 12GB any time soon as 6GB is already more than I need and because there are still a ton of project that I do that involve using 32-bit Windows so there more than 3GB is all useless gravy.
Yeah, the temps are split for everyone generally. Core 1-2 tend to be hotter than cores 3-4, but I guess reverse is possible too...Quote:
Originally Posted by ap4lifetn
Mines a bit different. I get say 75C/70C/70C/68C.Quote:
Originally Posted by dejanh
Are you measuring under Vista or XP? In Vista mine split the way you said, but in XP they are roughly 2 high/2 low.Quote:
Originally Posted by ReverendMaynard
On a different note...you know, after playing with those memory OCs, what the heck is the point in buying the ultra expensive 1866MHz and 2000MHz kits? I can get this memory up to 1950MHz before loosing a channel, and I can actually run it at 2000MHz in dual-channel configuration. But 1900MHz triple-channel is plenty plenty fast already and requires only 1.43750V QPI/DRAM. Heck, even 1600MHz was fast. It makes me think that these manufacturers just take the same chips, test them at higher speed, slap on a rating sticker and release it as "brand new high-speed" memory. Maybe if I was willing to crank up the QPI/DRAM to 1.65V as required by the Dominator GT sticks I would be able to run triple-channel at 2000MHz with the 1600MHz sticks. Not going to try that though, ever :p:
For anyone considering cooling, my core i7 overclocks much better, either higher or with less voltage with cooler temps. At 26C ambient, max OC for running chess stress test was 21x212, and even that was barely stable with 1.64Vcore, up to 1.68v no help. 21x213 no boot, no matter what.
Open the windows and cooled the room to 12C, and then it ran chess stress test easily 10x in a row at 21x215 (trying to get max score) and needed only 1.65 vcore. 21x215 in cold was very stable, 21x212 at 26C even using same Vcore and other settings would barely run, constantly crashing.
Just tried running linpack, and could lower vcore 2 notches and make a successful 25 run by dropping temps 14C, where same settings at 26C always failed.
Yup, I noticed this too having used three separate i7 940s. Even on pure air the same thing happens if you just drop your temps. Even just 3C - 5C difference in ambient can mean stability over complete failure. That's also why I was saying that this new i7 940 I got should be able to go as high as 4.6GHz if temps can be kept under control.Quote:
Originally Posted by rge
Not every chip obviously reacts exactly the same, but all of the ones that I tried did at least...
What about the IOH Temp?Quote:
Originally Posted by dejanh
Under load they can get quite high and if the increased bclck and the temp, it might be the source of the failure.
I am trying the best with the Corei7 920...
The results are almost the same that i7 965
http://i42.tinypic.com/2u74z0m.jpg
http://i40.tinypic.com/vzjxpw.jpg
http://i44.tinypic.com/zmk3kp.jpg
http://i43.tinypic.com/1z2qaew.jpg
http://i42.tinypic.com/2uzb9t4.jpg
http://i44.tinypic.com/34nnxqs.jpg
http://i39.tinypic.com/2hef4sl.jpg
i am doing my best, but i think it s the better on the air
IOH temp, either by software or IR, does not get much higher at 4.5 than 4.2 on mine, and at both are lower than all my early prime runs prior to putting a fan on IOH...which incidentally did nothing for stability...so not an issue on mine. But cpu temps go from 70's to near 100 from 4.2 to 4.5.
i started using linxQuote:
Originally Posted by dejanh
after realizing prime was not enough to determine system stability
within a reasonable stress period. (can't afford priming for 10+ hours to see if its stable or not)
one question though
what "problem size" value is considered a rational setting?
i see in your screenshots using a size of 10000
the program on my system ιν 64bit mode can go up to 25000
which problem size do you find as reliable
thanx in advance
bios 1.3b6
222.5 x 19:D
HT > enabled
Turbo > disabled
Dram ratio 3
Vcpu > +0.31V
QPI > +0.18V
Cpu pll > 1.81V
DDR > 1.65V
IOH > 1.15V
http://i82.photobucket.com/albums/j2...22x19_linx.jpg
4.8Ghz with help of 4C outside temps with 1.68 vcore cpuz. (inside temps max is 4670)
http://valid.canardpc.com/show_oc.php?id=481012
You have your radiator outside?
Set my computer outside, wanted to see max in cold, but keyboard, mouse and monitor inside (wires though cracked door), no point in me freezing too:DQuote:
Originally Posted by Aberration
But after seeing the difference, I would not mind mounting the rad outside my window...except the cold season isnt long enough here.
great $105 Ram kit
http://www.newegg.com/Product/Produc...82E16820227384
1.7vdimm :up: 1.45 vqpi/vtt
^^ that is a great deal :up:
Nice overclock too!
That is a nice set of ram there Bobbylite.
My OCZ Platinum 1333Mhz would scale up nicely all the way up to 800Mhz @ 7-7-7-18 1T.
they go up about 900 Mhz @ 8-8-8-18 1T but unfortunately they will no scale up with speed with increased DRAM voltage :(
I gave them up to 2.1v and they did not seemed to respond, thats a real bummer as I kind of thought they have a great potential.
Bobbylite, do you mind see if your OCZ platinum increase with voltage, as that would be great :D
That's no joke. Air coolers simply aren't adequate for these chips once you breach the 4.0Ghz threshold. Decent H20 is the only way to go with these suckers. Or phase.....:p:Quote:
Originally Posted by dejanh
Was going to try up to 1.8vdimm :up: will report backQuote:
Originally Posted by The-Fox
Thanks, that is jolly good :D
BTW, you will probably find out that going above 1.76v~ Vdim will require some high QPI/VTT voltage.
I didn't get any POST with vDimm at about 1.76v when my QPi voltage was at about 1.5V
I suggest you try 1.6v for the QPI/vtt in case you don't get a POST, if it works for you without the additional QPI voltage than thats even better.
You can get 4ghz (HT on) with a good air cooler - but anything over that (with the exception of a few exceptional CPUs) overwhelms the ability of practically anything on air.Quote:
Originally Posted by Classic Satch
(Exception - put the machine outside in winter!).
I plan to put my 965 under water this coming week.
hey stevero, almost identical scores on that super pi run lol
http://img.techpowerup.org/090110/Capture029.jpg
So I have few questions that I seem not to be able to find answeres too:
1 - How do you determine VID?
2 - I have the EVGA board and I7 920. The first day with defaults (except for Raid turned on and Intel PWR saving turned off) and with stock cooling, the board automatically put my vcore at 1.25v. After installing my water cooling and not changing anything in the BIOS, the board put the vcore at 1.16v. Any ideas on that? Same BIOS by the way, literally, the only thing changed was the cooling.
3 - I have Corsair TR3X6G1333C9, 3x 2Gb 1333 Ram. If I set my Bus Speed to 170 do I set my divider at 2:6 or 2:8? On EVGA's forums Jacob states that the uncore frequency has to be exactly twice the memory multiplier. If I put it at 2:6, then the uncor freq would need to be 12, which is not an option. If I put it at 16, then they do not match. I am just trying to figure this part out as there is a lot more to do than there was with 775 chips. And I am just using 170 as an example.
Thanks
VID is shown on some earlier versions of Real temp, maybe core temp might show it, but im not sure how to tell. my EVGA board did that too!, THe voltages are always changing, however, you sure you might have not turned vdroop control on/off?Quote:
Originally Posted by Utnorris
Anyhow, im not fully understanding this ratios of the QPI/uncore, and all that, check out this link and see if it helps, im reading it here and there trying to understand these ratios, its def. a bit complicated
http://www.xtremesystems.org/forums/...&postcount=876
Thanks. Vdroop is off, meaning there is no Vdroop. I will take a look at that thread. It's just so damn confusing. :D
i feel you bro, my setup now, whatever I set the voltage to in the BIOS is about .5 less than what it is in windows.Quote:
Originally Posted by Utnorris
So now I have it set to 1.3625v I think and in windows I am running @ 1.423v
Check it out:
http://img.techpowerup.org/090110/Capture030651.jpg
Do you mind PM'ing me your BIOS settings, I cannot get anything to really run other than dummy OC. Any time I set up an overclock, my board start beeping upon restart and I have to shut it down and restart it. The settings are still there, but it doesn't make any sense to boot to windows if the board went berserk with the settings. It's a shame that dummy OC doesn't go higher than 3.3Ghz. I wouldn't have such a headache. Oh well, I guess I have to re-learn how to overclock. :D
I'm only going to address question 1. I asked a similar question about the CPU-Z configuration file option Sensor=0 to read Core VID. I got these two responses...Quote:
Originally Posted by Utnorris
1. Core i7 no longer reports VID since it manages the vcore supply via it's on-die PCU.
2. the core i7 reports no VID
So I guess it's a guessing game on which batches will be good or not.
Well not even phase can help you if you get one of the crappiest i7 ever made:Quote:
Originally Posted by Classic Satch
http://www.xtremesystems.org/forums/...1&d=1231637878
That's right, this is all I am going to get out of this dog chip. It does NOT work with the 20x or 21x mult (no not even 4,2 Ghz with 1,6V), only likes 19x.. which means you hit a brick wall above 210 BCK if you're shooting for real stability.
I need 1,46V VTT for 211 stable, 213 needs close to 1,6V VTT for linpack stability. Forget getting higher or forget real stability.
AND DO NOT BUY 3838A :rolleyes: :shakes:
Ouch bro! That really sucks.Quote:
Originally Posted by jcool
I know what you're talking about vis a vis those multipliers. My chip also hates 20x and 21x. I've been running it 211x19 @1.36v 24/7 with no issues at all. However, try to use 200x20 or x21 and it fails Prime within 30 minutes, regardless of voltage.
got a question guys. I got the EVGA X58 mobo, my NB temps are about 76ºc under a run on prime, and my Vregs are about 88ºc, is this too hot, or normal?
yikes that does sound bad but remember to isolate the memory stability side of things too as that can also cause instability and reboots under prime/all core stress tests.Quote:
Originally Posted by jcool
Thanks man, I gotta get a different CPU.Quote:
Ouch bro! That really sucks.
I know what you're talking about vis a vis those multipliers. My chip also hates 20x and 21x. I've been running it 211x19 @1.36v 24/7 with no issues at all. However, try to use 200x20 or x21 and it fails Prime within 30 minutes, regardless of voltage.
Thanks eva.. but I do know how to OC, ya know ;)Quote:
Originally Posted by eva2000
Might have needed a few hours to get used to the DFI, but I have built a few rigs with the GA-EX58 Extreme so far so I really know my way around its bios - different mobo, same :banana::banana::banana::banana:ty results :(
As for the memspeed, I tested with 6x and 8x dram mult, at high Vdimm and CAS9-9-9-24 with D9GTR mems :rofl:
As for mem stability, I can run my 6GB D9GTR at 200x19 at 10x ram mult and CAS 8-8-8-20, but it needs 2,07V to pass memtest. Which seems to be the DFI 10x issue because it does the same at around 1,9V with 12x ram mult and lower BCLK and also on the GB at 10x :shakes:
These are Diablos after all.
i7 CPUs have no predefined VID, it's determined by speed and temperature at bootup as MRC Code runs. To get absolute voltage, select manually and not Auto
yeah DFI 10x mem multi on bioses other than 12/18 are needing a bit more vdimm than before.. same issue with 01/06 bios but anyway still managed some nice mem clocks. Waiting on DFI to fix it in next bios :)Quote:
Originally Posted by jcool
Mixed 2x1GB Corsair DDR3-2133C9DF Dominator Samsung HCF0 modules + 1GB G.Skill PC3-12800HZ Samsung 816 HCF0 module together for first time Samsung HCF0 based triple channel configuration. The G.Skills are holding back the clocks as they need around ~0.1-0.15v more vdimm at same clocks as Corsair 2133 stuff, but still managed to be Super Pi 32M stable @DDR3-2191Mhz at 1.965v bios set vdimm.
@DDR3-2191Mhz at 1.965v
Don't mind the slow PI time as just testing how far I can clock this mixed set of 3x1GB Samsung HCF0 modules :)
http://fileshosts.com/intel/DFI/DFI_...8m36s344ms.png
CPUZ Validation @DDR3-2211Mhz at 1.965v
Still at 1.965v vdimm set in bios so probably limited by max Bclk.
http://valid.canardpc.com/cache/banner/482092.png
http://fileshosts.com/intel/DFI/DFI_...cpuz_valid.png
Yeah nice :)
Forgive me though I don't care one bit about Spi etc, I care about max 24/7 clocks at absolute stability so this thing can crunch some serious WUs. So I need a good CPU FFS. I bet I could run yours at 4,5-4,6Ghz 24/7 :(
Give it to me hrhr :D
for 24/7 probably better off dropping mem clocks and tightening timings for more effiency in mem bandwidth and latency :)Quote:
Originally Posted by jcool
nooo not my baby i7 920s.. can't let them out of my sight :D
Yeah currently running CAS 7-6-6-15 at 210 BCLK 8x mem divider, works a treat :D
Amazing Ram speeds eva :up:
BTW, did you notice any difficulties running high ram speeds at high CPU speeds ?
When I had my i920 running on DryIce, I couldn't run high ram speeds above ~4520 Mhz.
I ran 221x21 with memory multi 6 with no problem but same cpu settings and memory multi 8 would fail in about 214-215 bclk.
I know the memory can do it as I ran 221bclk x19 and dram multi x8 with no problems.
I didn't try raising the vdimm, though I doubt it will help.
I probably got a weak IMC.:shrug:
does anybody know what the default QPI voltage is?
nice :)Quote:
Originally Posted by jcool
woah you ran mem at 221x6 = DDR3-2652 and 221x8 = DDR3-3536 :eek: or you mean 3x and 4x mem multipliers ?Quote:
Originally Posted by The-Fox
Haven't tried much on 8x mem multiplier but tried 6x mem with same 3x1GB HCF0 sticks and
max on 5x mem multiplier i got was 1096mhz 9-9-9-24 for DDR3-2191Mhz 32M pi stable at 1.965v
max on 6x mem multiplier i got was 1049mhz 9-9-9-24 for DDR3-2098Mhz at 1.965v
So could be higher mem multiplier harder to IMC/system ??
It depends on the QPI speed, but the range is 1.15v-1.375v according to Intel spec.Quote:
Originally Posted by liftedcj7on44s
If you leave the voltage on auto you will see that the voltage is always within the above range.
Yahh, I mean x3/x4=x6/x8, no way to run DDR3 @ 3xxx mhz :rofl:Quote:
Originally Posted by eva2000
It can be a bit confusing as some people (like me) refer to the DDR (effective) multiplier and some refer to the Speed (Mhz) multiplier (like you).
What I meant to say is that higher CPU speed (not memory multi) causes to achieve lower Ram speeds.
You would have hard time finding that as your ram speeds are way too high anyway :p:
For instance, try running CPU multi 18/19 and try to max out your RAM speed.
Then, try to achieve the same ram speed at higher multi (20/21) which will result in higher CPU speeds.
IMO, where the CPU is getting maxxed out, your IMC is having problems running high ram speed.
Yeah i experience something similar but i think it's possibly due to Turbo mode 21x on i7 920 as i reported at http://i4memory.com/102693-post9.html and usually happens once past >@210-215bclk , the fix is more uncore volts :)Quote:
Originally Posted by The-Fox
3x1GB Samsung HCF0 @DDR3-2076mhz 8-8-8-20 at 2.025v
http://fileshosts.com/intel/DFI/DFI_...9m00s453ms.png
I tried even 1.6v for the QPI/VTT voltage and it didn't helped, so I figured it was the limit of the IMC.
I am looking for some good OCable RAM (re: 1600 MHZ and up for 24/7). And around $200-$300. I already have a 940 with a asus p6T mobo.
Any suggestions.
Did anybody try to test his system with XP64bit? I did only see Vista64 test screens but how are the Benchmarks with XP64bit?
http://i259.photobucket.com/albums/h...8/CIMG0661.jpg
Still waiting for cooling.
http://i259.photobucket.com/albums/h...ekT2008/a1.jpg Looks like a prime candidate for water.
I just picked up this. Incredible timings and speed for the price - let's hope it holds up - I'll know next week.Quote:
Originally Posted by avrionis
http://www.newegg.com/Product/Produc...82E16820227381
Man that looks like a killer chip man. Wish I would've gotten a "B". My "A" is sitting, unopened, waiting for it's mem & mobo to get here...Quote:
Originally Posted by DerekT
not a bad choice :up:Quote:
Originally Posted by Brahmzy
and nice Subbie!
Guys, something interesting: My dog chip goes way higher when I disable HT (even though I am NOT temeprature limited) - I think that's pretty interesting, especially for those of you that are getting issues beyond 4Ghz with HT enabled thinking their temps are too high (mine are subzero so its a nonissue).
I could easily bench 3D at >4,4Ghz and even run prime for about 10 minutes at 1,55V load Vcore with this setting:
http://database.he-computer.de/Bilder/temp/3d06i7.jpg
ok so if i wanted to run 2.0vdimm where should my QPI voltage be. And the reason i ask what the default voltage is because my MSI Eclipse only lets me change voltages by + or -. i cant select a certain voltage. So if default where 1.15 then i could select +0.010, 0.020 so on and so on. So i have no idea what to set qpi voltage at with 2.0vdimm on the ram.Quote:
Originally Posted by The-Fox
Keep the QPI within .5v of VDIMM. (to avoid insta-death, no long term consensus yet)Quote:
Originally Posted by liftedcj7on44s
errr. how am i supposed to keep the QPI voltage 0.5 from the vdimm if i dont know what the QPI voltage is? Is there a program that will tell me what the QPI voltage i have set is??????
You're overcomplicating the matter. If default is 1.15v simply set +.35 or .40 in bios and juice the ram.Quote:
Originally Posted by liftedcj7on44s
I DO NOT KNOW WHAT DEFAULT QPI VOLTAGE IS. this is what i have been trying to say!!!!!!!!
if my QPI voltage is the one in this screenshot, I don't see how I can have it .5 from my Ram voltage. Impossible.
Default is 1.1v, I don't even think the max voltage allowed in the BIOS is as high as my RAM voltage :confused:
http://img.techpowerup.org/090111/Capture033.jpg
no youre safe. it has to be within .5 to be safe and you are there. But i need something like that to tell me what QPI voltage i am running!!!!!!
Quick question: isn't using the i7 turbo feature to reach your 'max' overclock sort of cheating? Doesn't the increased multiplier only apply to one core? When you get the turbo multiplier to show up in CPU-z it's really only increasing the multi on that one core right? So lower voltage should be required for turbo enabled vs disabled for a given clock speed. Meaning you lower your clock speed as you enable turbo to maintain your same (well, close) non-turbo max clock. I.E. 20x200 turbo disabled ~= 20x190 turbo enabled, on CPU-Z.
No, not really.Quote:
Originally Posted by JWMc
20x with turbo enabled = 21x permanent on all cores, 22x if only 1 core is loaded - unless you're on a mobo that doesn't disable the current limit. Then the CPU will drop back to 20x if it detects 130W are exceeded.
Got my i7 running at 3.4 Ghz at the moment.
BIOS settings:
vCore - 1.265
QPI/DRAM - AUTO
vDimm - 1.65 (1360 Mhz)
QPI multi - 18x (3060 Mhz)
I'm having some trouble with my QPI voltage. I've tried setting it withing 1.25-1.27v, and I can't get it stable, and Prime crashes almost instantly at those voltages. What am I doing wrong? I can't see why it should be higher
I have ran mine at 4.2 GHz with HT enabled literally since I bought this chip in exception of the last week or so as I was trying different settings. Well anyways, I decided to try to disable HT and see if I can go higher. First off I wanted to see if I was stable, so here is where im at now, the only problem is my temps, I will try to go as high as this votlage allows, I will not increase it any further until my new cooling parts come in.
Another thing is if you look in the HW monitor window, my vregs and NB are getting really hot, I don't think this is normal, here is a shot of my voltages, any suggestions?
http://img.techpowerup.org/090112/.jpg
http://img.techpowerup.org/090112/Capture035.jpg
Ah thanks for the info. My memory of how turbo boost worked wasn't good. So the single core multi thing I was thinking of with turbo was really the same multi for all 4 cores, except 3 cores are off when the CPU gets +2 multi. How can you get the +2 multi to kick in realistically though when something in windows is going to activate one of the other cores?Quote:
Originally Posted by jcool
I am now trying 20 x 191 with turbo boost enabled for (21x) 4011mhz load, (12x) 2292 idle. I wonder if I can drop my vCore and/or vQPI due to the decrease in memory speed by going from 200 to 190 BCLK even though my clock speed went up 11mhz. I reduced the memory voltage to 1.62v and it's stable so far.
My new 24/7 I think for i7 940. Going for the best OC/voltage ratio.
Currently 3.8GHz CPU, 1603MHz 7-8-7-20-1T DRAM, with 1.20000V CPU, 1.23125V QPI/DRAM, 1.55081V DRAM (all BIOS settings)...C-states are enabled hence the post-run clocks are lower.
Screenshot below...not bad for a low-voltage OC I think :up:
My i7 920 runs at the same speed of yours but with a very low voltage i.e below 1.1VQuote:
Originally Posted by nemesisrobot
http://img.photobucket.com/albums/v4...4Ghz_htoff.jpg
6hrs prime without errors.
http://img.photobucket.com/albums/v4...hz_superpi.jpg
Now it's running at 1.024V idle and 1.03V on load.
You can try to increase your Vcore to make it more stable.
Man, I've found it to be far better to run two instances of Intel Burntest for stability purposes. It stresses your chip waaaaaay more and you'll discover an error far faster.Quote:
Originally Posted by tot31
http://www.xtremesystems.org/forums/...d.php?t=197835
What kinda of cooling are you on, Tot?
I'm on air using Deepcool Ice Blade Pro with 2 x 120mm fans (push/pull) setup. My room temperature is normally between 30C - 32C.Quote:
Originally Posted by YukonTrooper
http://www.modernopc.com/store/image...eblade_LRG.gif
http://www.modernopc.com/store/images/iceblade_04.gif
I'm at 3.8ghz using 1.31v CPU. Why are you guys going for such low voltages as 1.1v? Are you using a micro case or something where heat is a big deal?
I'm living in a very hot climate country where all around the year is always summer. Like I posted before that my room temp is around 30C - 32C, sometimes can go up to 34C, so heat is really a big problem to me...:(Quote:
Originally Posted by Hot'n'Chilly
Made a few updates to my i7 OC Optimizer tool :) :
- more multipliers for ram and QPI. (from P6T bios)
- ability to enter the QPI speed in terms of MHz or GT/s.
- Turbo mode options.
http://overclocking.gogar.com/i7form.png
See sig for link, pm me if you find any errors.
What are your load temperatures, Tot? I bet at 4.1GHz they are :explode2:
I've now got my QPI at 1.35v, and that allowed me to lower vCore to about 1.22-1.23 in the BIOS. Any lower seems to BSOD on Prime within an hour. How can you guys be running at sub 1.2v at higher clocks?! :confused:
1 got max temp 72C on 100% load running at 3.5Ghz with 1.06V. I think if I run at 4.0Ghz the temp will be 90C or above.....Quote:
Originally Posted by YukonTrooper
http://img.photobucket.com/albums/v4...i7_35Ghz-1.jpg
based on my own experiences, and from reading this thread, seems like there is a high correlation between week number and OC ability on the i7 920. The earlier the week, the better OC. For example, I've tested three different batch 3841xxxx 920s (week 41) and they were all duds. The chips that are getting phenominal OC's in the 45-4600Mhz range are all earlier chips, like 3834xxxx - 3836xxxx batch. Has anyone gotten a really good OC out of, say, a week 39 or later?
I have a week 37 and its a dud with HT on, HT off im 4.3ghz stable 1.4375 vcore
Something I think people need to quit confusing in their overclocks is that a 920 running at 20x200 = 4Ghz is NOT the same thing as 21x190 = 4Ghz with Turbo mode enabled. The higher multiplier from Turbo mode does not function in the same manner as actually selecting a higher multiplier. Turbo mode only engages when there are idle cores. When 1 or more cores are running idle it increases the speed of the cores that are under load. If however you have all 4 cores running under full load such as when you are running Prime95 then Turbo mode is not increasing the speed of the cores to 4ghz in the example above. You are at 3.8Ghz. And that is why if you disable Turbo mode and try running 20x200 instead your system most likely isn't Prime stable unless you increase the vCore more.
Are you generalizing all X58's with that statement? If the person is using a Gigabyte board that 21x multi does not drop to 20. It's permanent :up:Quote:
Originally Posted by burningrave101
Asus sticks @ 21 with all cores loaded as well provided your temps are in the clear.Quote:
Originally Posted by ReverendMaynard
gotchya, so it has a TDP limit. The Gig's don't which is why I bought it :)Quote:
Originally Posted by Truckchase!
Well I'm just generalizing it based on how the technology works according to the info I've read about Intel Turbo mode.Quote:
Originally Posted by ReverendMaynard
http://www.behardware.com/articles/7...hitecture.htmlQuote:
Turbo mode and overlocking
The Turbo mode of the Core i7 is not an architectural characteristic but a functionality that Intel has already implemented on certain versions of the Core 2 Mobile under the name IDA (Intel Dynamic Acceleration). This mechanism consists of accelerating in a dynamic and temporary way the clock speed of one or several of the cores when others are not called for. The concept is based on the fact that many applications consist of one or two threads and therefore do not use all of the multi-thread processing potential of a multi-core processor. When the case arises, the Turbo mode comes into play (under control of the operating system) and increases the multiplying coefficient of the one or several cores in question.
Thus, Nehalem is an evolutionary step in the control of internal voltage and clock frequency. Up until now this control was handled by the operating system and the processeur offered the possibility of external control of the multiplying coefficient (FID: Frequency Identifier) and voltage (VID: Voltage Identifier) which forms the foundation for EIST (Enhanced Intel SpeedStep Technology). With the Core i7, these parameters are no longer external and it alone can change them in order to have control over its thermal dissipation. Indeed, the processor is capable of estimating the power it consumes at each moment (voltage x intensity of the current consumed) and of course to control it with the help of frequency and voltage parameters. Therefore, a Core i7 model will no longer be characterized by its maximum FID and VID but by its maximum TDP. Software (the BIOS and operating system) no longer controls the FID/VID combo as it did with previous generations but rather « Power-States » (or P-State). These power levels are defined based on the overall TDP of the processor at the maximum frequency outside of Turbo mode.
As an example, the Core i7 has an announced TDP of 130 Watts at 2.93 GHz, or 22 x 133. For each intermediate multiplying coefficient, the estimated TDP equals:
TDP[coeff] = (coeff / max_coeff)3 x TDP_core + TDP_uncore
The “uncore” part of the processor (IMC and L3 cache) is not subject to P-states and thus 20 of the 130 Watts is constant. We therefore obtain for example at 14 x 133 = 1.86 GHz:
TDP[14] = (14/22)3 x 110 + 20 = 48 Watts.
The multiplying coefficient varies between 12x and 22x which gives us ten P-states between 37 and 130 Watts.
The Turbo mode thus operates within the framework of this internal control of the processor’s overall TDP. The absence of activity of one or several cores results in a lowering of the overall TDP and thereby offers the Turbo mode the opportunity to accelerate cores that are called for.
With this new protection mechanism via control of the TDP, there is the obvious question of overclocking as the processor’s maximum TDP is quickly surpassed. A priori, Intel will finally not set any limitations here and it will be possible to go beyond this value, but of course the Turbo mode will not be in effect.
Only variations of Core i7 “Extreme Edition” models will enable modifying the TDP ceiling. However, note that these modifiable parameters on the Core i7 XE will only concern the Turbo mode. Should the need arise, it will thus be possible to modify the maximum multiplying coefficient as well as the TDP but this will not mean that the processor will function the entire time under these parameters.
So how are you guys finding out that the 21x multiplier sticks and is enabled at all times under full load of all cores? Is this something that someone found out during testing? How did they test?
Because for my Gigabyte EX58-Extreme this is what is says on the box:
All I know is that it's easier to get 20x190 w/ Turbo stable than it is 20x200 w/ Turbo disabled. And I doubt that it's simply due to a 10Mhz increase in the BLCK which would raise the QPI, Uncore, and Memory a little but still well within stable spec.Quote:
Intel Turbo Boost Technology
Dynamically reroutes power to improve performance
For situations where all cores are not needed for a particular workload, Intel Turbo Boost Technology powers down the idle cores and dynamically reroutes the power to the active cores, boosting their performance without wasting power. This allows the active cores to achieve significant overclocks.
Don't lean on articles from before the cpu was actually released.
Well like I said, link me to where you guys are reading that the 21x multi sticks just like a normal multiplier on the Gigabyte or Asus boards.Quote:
Originally Posted by ReverendMaynard
Yeah, that was the case initially, but then Gigabyte decided to permanently scrap sticking to the specs for TurboBoost and just have it on all the time, no matter TDP limts or anything. Hence, Giga boards have this on all the time.Quote:
Originally Posted by burningrave101
This is not "written" anywhere. People own these boards, and they know, they have tested it time and time again. I have an Asus board and mine shuts off Turbo as soon as temps get into the 80s. Asus - TDP limit enabled and cannot be disabled, Gigabyte - TDP limit permanently disabled.Quote:
Originally Posted by burningrave101
Someone corrected me earlier on this. The way turbo boost works is it gives you stock multi +1 for -all- cores if the thermal limits aren't reached, and the processor is loaded enough. If 3 of the 4 cores are doing nothing and switched off, the processor multi will go to stock multi + 2 for all cores but only one of them is active so effectively it's one.
The reason it's easier to get stable at 4.0ghz @ 20x190 with turbo instead of 4.0 ghz @ 20x200 without turbo is that the memory system and Uncore aren't being run at as high of a speed due to the reduced BCLK.
See this document:
http://www.intel.com/support/process.../CS-029908.htm
burningrave101, here is a TON of real world results provided by people who own the P6T Deluxe/RIIE and i920:
http://www.xtremesystems.org/forums/...d.php?t=212389
The Multi is depended on Temp in ASUS, they implemented it differently than other vendors like Gigabyte.
TDP is determined by the board, not the cpu so it is up to the manufacturers discretion.