Printable View
Good news, I picked up a CoolIt Freezone Elite brand new for $330 CAD ($265 US), which is a very decent price for this unit. Bad news, it is just sitting in a box right now because somebody forgot to include the LGA1366 retention bracket with it :shakes:
CoolIt is shipping one out to me but now I have to wait for almost another week before I can fire it up, and I freaking live in Vancouver, BC...province right next to their HQ location.
Darn it NCIX!
Prepare to watch that thing explode. How many watts is it rated for?Quote:
Originally Posted by dejanh
keep us updated dejanh - this looks like it could be a good answer to someone like me who wants better than air performance but without the hassle of a full water cooling system.
I am curious about this also. I have thought of trying the boreas - adding it into my existing loop. Not sure how much it would help though, my water system is doing a pretty good job already. See - http://www.frozencpu.com/products/71...ol_Center.htmlQuote:
Originally Posted by dejanh
My best 3dmark06 below. Interesting to note that I can get a higher score running 4 threads vs 8 - because I can oc higher with 4.
Following my last thread re 19*200 (3.8Ghz) before http://www.xtremesystems.org/forums/...postcount=2013 I wanted to bring details of this setup running 19*205 (3.9Ghz)
Ive had to bump voltages somewhat, and unfortunately the run was performed in a room which was several degrees higher so I wouldnt want to compare the temperatures directly to the previous run because the voltage will not account for all that delta.
Here's the new data (same format as before to make comparison easier)
Again, this system is 100% stable so far, running dual runs of Linx in parallel with Prime95 blend for a couple of hours thrashes it to death. Without this torture test I can drop some volts and still appear stable.
i920 'A' batch, Asus p6t-D, Thermalright 120 push only Sythe fan, corsair 620HX PSU, Win Vista 64bit
6GB Corsair Dominator 1600 ram @ 7-8-7-7-20
Bios settings.
CPU 1.38125v (1.31v indicated in CPUZ in windows under full system load)
QPI 1.3250v
Ram 1.64v
PLL 1.92v
IOH 1.24v
Load line calibration = off
CPU/IOH clock skew = normal
CPU Differentail Amplitude = auto
CPU / PCIE spread spectrum = disabled
c1e,virtualization tech, execute diable bit, a20m, c-state tech all off
HT tech, HW prefetch, adjacent cache line prefetch all on
Will push on to 19*210 next.......
Nice dude, you have me by around 1k but I used vista for it. Just picked up another GX2 for $70 lol, gonna give quad a go.Quote:
Originally Posted by SteveRo
I'm looking for another GX2 as well. May I ask where you got the GX2 for $70? PM me to keep this thread on topic....:)Quote:
Originally Posted by ReverendMaynard
Wow reverend, 70$ - that's a great deal. quad 9800 -can't wait to see your results.Quote:
Originally Posted by ReverendMaynard
The pair for $140...not bad lol. I might not have the PSU for it though :down:Quote:
Originally Posted by SteveRo
On topic:
I wanted to see the impact that QPI/PLL had on cpu temperature and going from 1.4v to 1.6v (lowest possible increase) I gained 7C on idle and load.
Food for thought.
Geez this is pissing me off. I CANNOT get 3.8GHz stable to save my life.
Oh how spoiled I was with my 4.4GHz E8600 Air OC...
These i7's are different animals...
Well, I am up and running but I did not do any tests yet. I want to let it run for a few hours before I decide to run tests. I did not wait for a LGA1366 bracket. Instead I remembered that the difference in the hole spacing between LGA775 and LGA1366 is very minor so I just made some minor adjustments to the LGA775 bracket by applying some force in the right spots. It works like a charm :)Quote:
Originally Posted by SteveRo
Now, this guy is 6 TECs, 72W each, so rated for 432W. It performs quite well so far. At 60% (basically inaudible) fan the idle temps are around 27C - 29C. I did not burn it in yet, as I want to get a few hours of it running before I decide to push the CPU and the cooler. However, considering that this managed to adequately cool a QX9770 running at 1.62V and 4.5GHz I think it will do just fine here.
The fan gets noisy past 90% (just tried it), so I will personally likely cap the fan at 90% and make do with what that gets me.
I like it so far though. Quite a bit cooler than air. So far, about 15C-20C less under normal load with the fan running at low speeds in my warm room. Air required me to control the room temps as well in order to be able to keep the OC stable, as under longer LinX runs when the room is normal temperature (23C) temps get up to 96C in Vista 64-bit which is unhealthy. I had to keep my room below 16C to keep the temps below 90C on air at 1.4V 4.2GHz for longer LinX runs.
I'll update with more results later :)
Sounds like we traded places. I had an E6420 that would do 3.2GHz and hit a brick wall. My i7 920 is doing over 4GHz under 1.3V.Quote:
Originally Posted by Brahmzy
Boy, what a difference disabling HT makes. I have my 920 running 50 passes Intel Burntest/12-hr Prime 95 stable @4.0Ghz 1.31vcore. I don't use a single program that would make use of Hyper-threading so I decided to see how low I could go with it turned off.
I'm going to be testing even lower voltages later but tonight I have a date with Fallout 3(already took care of the wife :D).
well here are preliminary results...
Its nothin much yet, but Im on the stock cooler.
Its been stable now for over an hour, doing random stress and priming here and there...
This is my 3836B043 =)
Finally took care of my video card dying problem. Turns out , I guess that PCIE power connector from my PSU must have a bad connector or something. Cuz, I went to a different PCIE cable and works fine now. This is on a single rail PSU too...so it wasnt a rail sharing issue.
Good luck. Mine boots into Vista low volts at 4GHz but isn't prime blend stable no matter what vtt,pll,vcore,vram, etc.Quote:
Originally Posted by GenTarkin
It sounds more like a possible memory issue. Have you tried running Memtest yet?Quote:
Originally Posted by derektm
Here's another 3837B073 gem. It can go over 4GHz with under 1.3V, but I'm on the stock cooler and temps get bad with Intel Burn Test. I think these will be close to my 24/7 air settings when I get the Crossbow retention bracket installed on my Vendetta 2. I should be able to drop the Vcore slightly because I took a shot in the dark for this run. Natalie Portman says hi.
http://img294.imageshack.us/img294/5167/ssssy8.png
Got an address where I can order one of those (international)?Quote:
Originally Posted by YukonTrooper
..& miss Hershlag ROXX! :yepp:
I would make Natalie Portman my permanent buffet but that's irrelevant. :)Quote:
Originally Posted by YukonTrooper
You're going to need more than a stock cooler to keep 4Ghz stable under load. You'll need good H20 to stay 4.0 stable for any length of time under significant load. These chips run hot, I've learned that over the past 10 weeks. Whether you choose to enable HT or not will have a huge impact on your final settings. I'm stable at 1.36v with HT on but, so far, 1.31v is stable with it off and I don't use any programs, and probably never will, that make use of it.
Yep. Not a memory issue.Quote:
Originally Posted by Classic Satch
Aahh, so the joy of overclocking tests continue.... :) I tried...Quote:
Originally Posted by derektm
Wow - dejanh - nice idle temps! Post back load temps when you have them. I am most interested in 100% load at high vcore - after the systems has come to thermal equilibrium. What warranty does coolit provide on these?Quote:
Originally Posted by dejanh
After a couple of weeks of tinkering, I have found my 24/7 settings:
4.0GHz. 4 cores, 8 threads, Prime (small FFTs) & LinX stable. 1523MHz DDR low latencies.
CPU Multiplier = 21x
BCLK = 190
VCore = 1.356v (allowing for 0.01 vdroop)
QPI = 1.35v
PLL = 1.89v
Vdimm = 1.65v
HT = on
Turbo = off
QPI = Lowest selectable value
UCLK = Lowest selectable value
http://img292.imageshack.us/img292/2...e95smalhg9.jpg
http://img90.imageshack.us/img90/805...mhzlinxsr2.jpg
Also passes Memtest and 4096FFTs in Prime (really stresses RAM) for over an hour.
Yeah... i too wanna a 3837B :)Quote:
Originally Posted by rabor
rem - good job! 24/7 temps under 80 should be fine - max temp is 100 on these.
my latest result :D
System specs
- Intel Core i7 920 3836A756
- Thermalright Ultra 120 Extreme + LGA1366 Bolt Thru kit + 2x 120x38mm Scythe UltraKaze 2000rpm 87cfm fans in push/pull
- DFI LP UT X58-T3EH8 R.A51 - 01/06 beta bios
- 128MB Gainward FX5200 PCI
- 2x1GB Corsair DDR3-2133C9DF Dominator + 1GB G.Skill PC3-12800HZ (3x Samsung HCF0)
- Ram cooling: 3x 60x25mm Sunon 23.5cfm Maglev fans
- 80GB Hitachi 7K80 SATAII
- Pioneer DVD-RW
- 1KW Corsair HX1000 psu
- WinXP Pro SP3 full nLite updated hotfixes / Windows 7 beta dual boot
i7 920 3836A756 @4769Mhz at 1.59375v bios
http://valid.canardpc.com/cache/banner/486620.png
http://fileshosts.com/intel/DFI/DFI_...z_valid_tn.png
Max screen capture capable @4680Mhz at 1.5875v bios
http://fileshosts.com/intel/DFI/DFI_...12517/cpuz.png
Super Pi 1M, 32M, Pifast results http://www.xtremesystems.org/forums/...d.php?t=214761 :D
eva - wow, highest I have ever seen on air I think - what do you think your temps where?
Well, mixed opinions so far...here's why...Quote:
Originally Posted by SteveRo
1. Best cooling is achieved when the TECs and the fan are running at 100%. However, the noise of the Panaflo pushing the full 110CFM - 120CFM is unbearable. Reminds me of the one time many years ago when I picked up a Delta Screamer. That f-in thing sounded like a jet lifting off, and this is not much better at 100%. My room is quiet and this rips my ears. However, the temps in Prime Small FFT at 4.2GHz, 1.4V, HT on max out at about 77C - 78C max on the hotter cores (1-2). This however is not an acceptable setting for me for day-to-day use. Some bench yeah, but day-to-day no way I could run this ever, not ever.
2. Keeping (1) in mind I went searching for what is "more acceptable" noise level. Quite frankly, for me that caps out at about 80% (70% preferred) as this is basically silent, or no louder than all of my Antec 900 stock fans combined at medium speed. This however also forces the TECs to run at the same 70% - 80% power (any reason why CoolIt?) and then I am seeing performance only marginally better (considering the cost) than air over extended periods of time (LinX caps out at about 90C, Prime at about 87C, 3DMark06 CPU tests at about 84C) which for extended runs is about 5C-7C better than TRUE. However, key here being that for instance 3DMark06 CPU test would most likely fail eventually under air, where as here it can keep running forever and the temps stay the same and this is all without lowering my ambient temps of 23C. For air I had to lower ambient temps past 4GHz otherwise LinX runs would climb past the 97C mark and I would get into the thermal throttling/shutdown area.
So, I am not sure what to say ATM. Is it worth it $320 CAD? Well, maybe. If you are like me and your average ambients are pretty hot, then it is probably worth it as it has more consistent performance than air. Also, if you do not mind the noise (which I do big time, for me > 80% is not a usable setting), then it is definitely worth it as you will see about 15C - 20C gains it seems compared to air at full loads.
Also, as funny as this will sound, this system will probably be better for the future 32nm chips and you do also get some pretty high quality components with it that could be reused in a decent custom WC system.
A few things to note still is that I did not use AS5 under the block but the stock paste that came splattered on the block. I am also using my modified LGA775 mount instead of the LGA1366 mount until that one gets here. It is possible that changing these two things could make some difference even at lower speeds. Also, the fan is in a "pull" configuration with the radiator here and I will investigate whether it can be changed into a push configuration to see what effect that has on the temps.
I kind of have a feeling that if this thing could be convered to be used with a even just a dual rad somehow you could be getting 50C - 60C temps with it easily. One guy posted a mod over at the Coolit forums where he just introduced an external reservoir in the loop, and changed the flow a bit and that caused his temps to drop significantly.
Quote:
Originally Posted by eva2000
Great job, EVA! I was wondering...previously, you reported 4610Mhz as your highest effort. What did you do to raise it to 4765? To go from 4610 to
4765 with just a vcore increase seems unlikely. 4765 is LN2 territory, and then only with a top 1% chip.
Question, does raising IOH core hurt anything? stock its 1.1 I put it at 1.120v that ok?
thats your NB voltage, i've had mine till 1.4v, still aliveQuote:
Originally Posted by GenTarkin
Wow, I had an original Freezone on my Opty 165 and it did a good job. I almost waited for a Freezone Elite to use on my new rig but impatience forced me to build a water cooled set-up.Quote:
Originally Posted by dejanh
Given your reports, I'm glad I did. The highest I've ever seen, regardless of clock speeds and voltage during a Linpack test, is 75C. At my 24/7 settings of 4.0Ghz, I usually don't crack 70C. I had a feeling that the small coolant volume in the Freezone units would present a problem and from what you've posted, it looks like I'm right.
eva - awesome bclk on i920. I am stuck at 215, which I can do at fairly low qpi/vtt. I am trying to get 217 to windows...with high volts I can sometimes get 217 boot, but cant even get screen shot. Any voltages/settings that help other than qpi/vtt or qpi pll?
Well, I guess my 3836B aint that great. It wont even do 3.4ghz @ 1.125v =(
Its stable for about 15mins in prime95 blend then BSOD's
Temps maintain 70-73c ... that too hot?
Dejanh - looks like mixed results - ok -and you bring up some good ideas for mods, assuming you are going to keep the cooler you should pursue the mods, some more ideas - perhaps change to a different fan or add a fan is a push/pull. Perhaps consider a better cpu block or add a radiator in front of the coolit. I plan to research this some more, I already have my temps <80 for 1.55vcore when multihour priming. This is down from 90-100 when I was on the nochua. If I could get another 10C off I might be able to get 4.5+ (HT off)? If I can find data to support this - a Boreas will be very tempting!Quote:
Originally Posted by dejanh
This is very interesting - look at this - http://www.coolitsystems.com/index.p...tid=18&id=3134
If that is just +1 up - it shouldn't be a problem.Quote:
Originally Posted by GenTarkin
Also, is 70-73 under full load?
What is your idle?
I have been able to run prime for many hours with very high temps (max temp at 97 - below - on air) - I would say 73 is a nice cool temp if that is at full load.
From what I can see - my own experience and that of others - nehalem does generate a ton of heat but it also seems to tolerate it surprisingly well!
Yes, maybe small amount coolant volume might be holding it back but what if you put a coolit in front of a 480 radiator in a custom loop?Quote:
Originally Posted by Classic Satch
max temps on these CPU's is 100ºc. You're fine.Quote:
Originally Posted by GenTarkin
Well now, that may be a different story! :p:Quote:
Originally Posted by SteveRo
I'm interested in the Freezone, as I have high ambient temps (around 23-25°C all year), and i LIKE them so high. I did look into water chilling in the past, but never found an easy and relatively cheap solution.Quote:
Originally Posted by dejanh
For one thing, did you change the cpu waterblock?
What about changing the fan with a better one like the scythe ultra kaze? My 3000 has absolutely bearable noise for me :)))Quote:
I kind of have a feeling that if this thing could be convered to be used with a even just a dual rad somehow you could be getting 50C - 60C temps with it easily. One guy posted a mod over at the Coolit forums where he just introduced an external reservoir in the loop, and changed the flow a bit and that caused his temps to drop significantly.
Here's a hour+ prime run at 1.187 vcore, 4011 Mhz, 75F ambient temp, notice the time stamps on worker threadsQuote:
Originally Posted by eva2000
Linpack anyone?
3836A977 i7 920
air cooler Tuniq Tower 120 w/handmade hold down
gigabyte UD5
Quote:
Originally Posted by eva2000
Here's a hour+ prime run at 1.187 vcore, 4011 Mhz, 75F ambient temp, notice the time stamps on worker threads
Linpack anyone?
3836A977 i7 920
air cooler Tuniq Tower 120 w/handmade hold down
gigabyte UD5
NICE. :up: Please post complete settings and bios revision! :)Quote:
Originally Posted by roller11
Edit: I see... 4 threads. Still, NICE! :up:, and I want to see the details. :p
dam that's one nice cpu as well.. week 36 seem to rule :DQuote:
Originally Posted by roller11
all info listed at http://www.xtremesystems.org/forums/...d.php?t=214381 :)Quote:
Originally Posted by rge
key was 4610 was bclk limited at 220 but i surpassed that limit to hit 228-230bclk and key was voltages listed at http://www.xtremesystems.org/forums/...d.php?t=214381Quote:
Originally Posted by roller11
thanks, 3rd highest overall on hwbot http://www.hwbot.org/quickSearch.do?hardwareId=CPU_1741 could be highest for air not sure hehQuote:
Originally Posted by SteveRo
Well my chip sucks or my PSU is holding me back.... either way my chip isnt doing 3.5ghz stable @ 1.182v in bios (1.156 cpuz) on stock cooler ...
ugh this is so discouraging =(
If I give it any more voltage my video card shuts down with its LED displaying error on 12V EXT...so, Im wondering if its my PSU holding me....
Right now, Im hoping that 3.32ghz will be stable @ 1.15v
Wow, how many amps are these things pulling when overclocked? Also you may be limited by the stock cooler just a bit there so don't be too sad yet!
well load temps on my highest clock of 3.6ghz @ 1.182v was only like 78C ishQuote:
Originally Posted by JWMc
Yeah. 3836B043 is sucking now! My first impressions of it were good, but upon further testing, I believe it sucks. :down: It wont do 3.8GHz stable at 1.296v, but my 940 will. On the other hand, the 3836B043 will boot into Vista at 4GHz with 1.296v, and the 940 won't! :rofl:Quote:
Originally Posted by GenTarkin
So far I've tested:
3836B043 (920)
3836A394 (920)
3837A728 (920)
3838A332 (940)
None of them will do 4GHz Stable without at least 1.4V+, but all will do 3.8GHz at <1.35v. I've tested this on eVGA X58 and Asus P6T with 4 different kits of Corsair Dominator 6GB 1600.
I don't know how you guys are getting these things to do 4GHz at low volts like this. I am on a Noctua Air cooler, but my temps are only topping 74 C.
What happens is prime blend just freezes, and 10 seconds later BSOD.
Anyone know if Frys has any good steppings in right now? (Bay area).
It was purchased from NCIX. They share the same warehouse as NCIXUS and DirectCanada. That's all the info I have.Quote:
Originally Posted by rabor
Actually, I meant around 3.8GHz for 24/7 air settings. My Vendetta 2 should be able to pull it down at least another 10°C, so I'll be in the green. Add at least another 5°C drop because those were Linpack load temps, and I'll likely never see those under 24/7 use.Quote:
Originally Posted by Classic Satch
Its funny this sucks =( rofl....my prime acts the SAME exact way on my chip.Quote:
Originally Posted by derektm
Freezes into about 15mins of it, then BSOD's 10 seconds later with "an uncorrectable hardware error has occurred" Its that same message each time.
I have a 3836B043 chip.
I wonder if there is one or two settings for this batch that specifically need tweaked....and we just gotta find them.
Oh well, so much for B's being good chips...I went through hell to get this one with frys rofl. First I bought a 3839A chip, didnt even open it..just took it back and fought for the last B chip they had.
Same here. I've tested 4 chips so far. I've tried every setting / voltage combination imaginable to no avail.Quote:
Originally Posted by GenTarkin
If anyone's in the bay area and wants to test these CPU's, let me know. I'd be interested in what someone else can do with them.
Try higher VTT... or lower mem divider/laxer mem timings. At least that's the issue for me when that error shows up, not the Vcore.
whats max safe VTT and Im assuming this is QPI/VTT right?Quote:
Originally Posted by jcool
I've tried up to 2.1V on PLL and 1.5V on the VTT (QPI/DRAM). Lowest mem dividers as well. I'm going to sell a couple of these chips and buy a good watercooling setup and see if that helps me any.Quote:
Originally Posted by jcool
I may go to Frys and see if I can find a 3837 or 38B next. It would be awesome if I could find a chip that does 4GHz with <1.3v because then I could just stick with air cooling.
Just to add some confusion to the mix, running QPI on the "slow" setting makes my setup super unstable. I can attain much better clocks when it's left at x36. F4k BIOS on DS4.
They call is slow for a reason - I ran 3dmark06 on slow and my score dropped from 24k to 3k.:down:Quote:
Originally Posted by YukonTrooper
Yup. Just wanted to reflect on what some people were saying about running it on "slow" for benching higher clocks. It seems that it runs so slow that it causes instability. Even when I select the OS on boot, it clears the screen very slowly compared to normal.Quote:
Originally Posted by SteveRo
derektm can you try clocking it without HT enabled and see if clocks go higher?
Sure. I'll test in about an hour. Need to get some groceries!Quote:
Originally Posted by GenTarkin
Seems very unlikely that its the CPU/vcore as you are getting same errors with different chips.Quote:
Originally Posted by GenTarkin
Maybe higher QPI/vtt although below 200blck you usually don't need any increase from stock voltage.
Does higher vcore helps to get the chips stable ?
No one knows for sure yet. Intel says 1,375V max, most people use 1,5-1,55V for higher clocks. If you aren't prepared to lose your CPU I'd recommend staying at or below 1,45V for 24/7 air, but that's just a rough guesstimate and my personal opinion on the matter.Quote:
Originally Posted by GenTarkin
ok, cuz Ive only tried 1.325v or so...
Yeah bump it on up to at least 1.375v-1.4v to see if that helps any. Even the Corsair XMP memory profiles set the QPI/Vtt voltage at 1.35v.Quote:
Originally Posted by GenTarkin
Yeah I have the OCZ 1333 stuff and their XMP profiles both bump it up to 1.4v .. I was just makin sure because intel spec sheet says 1.35 the highest, and then I saw what XMP set it at and I was like hrm....dangerous or ok? heheQuote:
Originally Posted by burningrave101
So, I will try higher.
Also, question is LinX a decent program to stability test with? Ive been running it on max memory and like times to run of around 200
With HT off temps go down drastically, which may allow for more overclocking headroom. With that said, I still couldn't get 4GHz stable on it even with 1.4v. Hopefully you'll have better luck than me. I always get the worst of the worst chips! :p:Quote:
Originally Posted by GenTarkin
I am really not sure where the variation is coming with regards to clocks on batch numbers. Need to get a good database going.Quote:
Originally Posted by derektm
Mine is 3838A557, and it doesn't matter how much voltage I give it, it wont do 4GHz stable.
But it runs 3.8Ghz with 1.27V.
I am on water.
Eh, its good enough. I am not one to hunt around for a chip. Sure wish I would have waited to purchase though, down $70 from what I paid. If it fails I will RMA, but if not I will stick with it.
I don't think temps are the limiting factor on these i7's unless you are either running the Intel stock cooler or using a real high vcore. I'm at 3.8Ghz with 1.3v testing right now and Prime95 Blend isn't even pushing me over 65C with aircooling. I still stay under 80C with 4Ghz and 1.4v+ while running LinX as well.Quote:
Originally Posted by derektm
Personally I think the limiting factor is the memory controller and it's becoming unstable at high clocks. I know it has to be something like that in my case because I can get 190x21 stable but 211x19 won't work for me and the lockups and BSOD's look memory related.
I've also noticed that a lot of the people with the highest overclocks are running 3x1GB of RAM instead of 3x2GB. Would be interesting to compare overclocks between running 3x1GB and 3x2GB to see what if any difference there is. Also note that some people are running Vista x64 while others are running XP or Vista 32-bit, especially if they only have 3 gigs of RAM. That does make a difference because it's going to be harder to get stable on Vista x64 due to the 64-bit excecution.
Yeah... I agree entirely. Its almost not worth it just to have that extra 200mhz. I've been through 4 chips that won't do 4Ghz reliably, but all did 3.8 with ease. That chip you have is better than my B chip which takes just a bit over 1.3v for 3.8Ghz.Quote:
Originally Posted by Aberration
I agree with you 100%. This is why i hate that intel has put the on die memory controller on these chips. It just like when i was overclocking socket 939 AMD's. Alot of my overclocking problem were due to a weak memory controller. Only way to test this theory is set the multi to the max and mem to the lowest and see what happens.Quote:
Originally Posted by burningrave101
what saved AMD is that you can get black edition CPU's for $200-300, but if you wanted that on a i7, you'd have to spend $1000+ on the 965! I still have my AMD rig in my sig, and overclocking those things is a bit hard because the old gen phenoms dont clock very high, so it was a bit of a challenge getting a good clock out of it. But the unlocked multipliers were awesome, it was hard to overclock it, but easy to tell what was the problem. I simply upped the multi, while keeping the HT, and the NB at default clocks along with the ram since I was not raising the fSB. automatically you knocked everything out of the equation as to where your instability was coming from.Quote:
Originally Posted by liftedcj7on44s
Now with the i7, god knows what is causing this instability, this platform is still very young and alot is to be learned as time passes. however, blaming the IMC could be a good good guess!
I've seen Fugger himself remark how stable the new 'uncore' is (memory controller / L3 cache), especially at super high clocks, maybe it varies greatly between chips.
Not the first time i've seen that, there are many posts with people doing 3.2ghz - 3.8ghz with 1.1v-1.2v, but i wouldn't be surprised if half of them won't go much higher without tons of extra voltage, that seems to be the way with these chips and a better quality chip really means you have more room before you hit this wall / ultra-diminishing returns. I remember Anandtech stating in one of their pre-launch articles how it would be childsplay to take i7 to 4ghz but anything above that would be what 'seperates the men from the boys' so to speak, perhaps they were partly right but it may more come down to getting a chip that doesn't have that low ceiling.Quote:
Originally Posted by Aberration
In the table below I think max qpi v is vtta and vttd - :shrug:Quote:
Originally Posted by jcool
There's no VID on i7s, right?
Ive got problems getting my board to even post at anything above 196bclk, 21x196 is prime stable for me with HT on @ 1.39vcc 1.35 qpi/pll .
When i put 197 no post , im sure i have more juice in this proc ;/ Tried anything i can think of , lowest mem/uncore/qpi dividers - bumped qpi voltage up to 1.55 , dram to 1.7 , vcc to 1.55. Laxed memory timings , tried skew settings - nothing. No post at 197 and up at all.
There was one time that PC posted at 197 noticably slower (watching the detect dram/cpu on the little lcd asus board has)and then froze on post. Cant reproduce this again thou, no post.
Tried 0903 and 0904 bioses, dunno if cpu just cant do above 196 or its a dummie board problem.
Any ideas ?
exactly my findings with mine.....Quote:
Originally Posted by Xello
3.6-3.8 very easy....
19*210 requires LOTS of V to stabalise (outside of TRUE cooling on airs ability co cool in a warm room)
Im going to try 20*200 and see if the slower memory timings let me get to 4ghz another way - without resorting to frying my CPU :)
I would think vid=vcore=Vcc (in the table) for 1.55v max?Quote:
Originally Posted by anzial
the 21 multi is the sweet spot (for me anyway) - can either of you boot at the 21 multi - maybe need to enable turbo to get the 21 multi?Quote:
Originally Posted by vikingboy
That crappy IMC theory certainly holds true for my 940. I can run prime, linx etc fully stable at 23x192 (4,4Ghz) with 6x ram mult, but even the 8x is no go. Boots into Windows but freezes with an evil Bluescreen after 10-20 minutes of prime. That is running 3x2gb with the most relaxed timings and subtimings possible, and 2T. Also tried up to 1,6V VTT to no avail. Uncore speed seems fine, I am currently running 6x dram and 20x uncore mult (3,8xxGhz uncore) so it's not that. It really has a crappy IMC.
Furthermore I had no trouble running DDR3-2000 with my previous 3838A 920 with the same mobo and rams, so I can rule them out. I guess I got a good CPU for high Mhz (does 4Ghz at 1,3V) but it has the crappiest IMC ever.
Unless it will run the mem better using other CPU multipliers - anyone compared that yet? Or maybe it'll simply need more Vcore to do it, even though that is kinda unlogical to me. Penny for your thoughts guys...
Yes, the spec claims max voltage of 1.35v to the Vtt analog or digital portions make no difference according to the spec.Quote:
Originally Posted by SteveRo
According to the spec, the analog and digital domains of the QPI should be fed by the same voltage.
I wonder if different voltages to the Analog domain and digital domain of the QPI can help to achieve higher QPI clocks.
I kind of have the same issue only at different speeds.Quote:
Originally Posted by jcool
I can run my CPU (x21 multi) with DDR 1800-1850 Mhz-> bClk 180-185Mhz with DRAM multi x10 benchable.
But it won't do even lower dram speeds at high CPU speeds, I couldn't pass 215 bClk with x21 CPU multi and x8 Dram multi which results in 1720 Mhz.
I tried to relax timings and raise QPI/VTT voltage to a high degree of 1.65v but none of this helped.
It was very frustrating as I could run higher CPU speeds (221blck x 21 CPU multi) at x6 Dram benchable.
I concluded that my IMC just can't run high DRAM speeds at high CPU speeds.
So you are saying that my 940 might run higher dram speeds if I go down on the clock? Might be.. I tried at stock clocks and got a no post using the 12x ram mult, after having to remove CMOS battery again I was fed up trying and just fired it up at 4,4Ghz again :DQuote:
Originally Posted by The-Fox
I mean it's fine, I don't really need more bandwidth than Tri-Channel DDR3 at 1200Mhz gives, but I feel kinda stupid running 2000Mhz Diablos in there while a cheap-ass set of 1333 whould do just fine :shakes:
I have really gotten no where with the 23 multi - 31 and 32 multi's work fairly well but the high multi's also increase the heat (higher temps) vs the x21 (seems to be the sweet spot for my proc). Hard to say what is holding back i7 - cpu vs imc. Do any of the x58 boards let you drop the qpi multi below 36? The ggbt slow setting is really too slow to be a reasonable test. Qpi multi's of 34, 32 and 30 would be very nice!Quote:
Originally Posted by jcool
No... not that I'm aware of.
QPI isn't the only problem though, high BCLK takes a toll on the CPU and expecially VTT requirement, that's why I prefer the 23x mult to the 21x.
Yahh, that is exactly my point.Quote:
Originally Posted by jcool
Try to do the following:
Lower CPU speed to about 3.8Ghz (preferably using lower bClk and high CPU multi) and then try to increase dram speed.
If you still get no POST/unstable, try to divide if it is RAM or Uncore by running the Uncore at the same speed but DRAM at lower speed.
Keep in mind that on some cpus, higher Uncore speeds require increase in DRAM voltage.
BTW, what M/B are you using ?
Read my post again ;)Quote:
Originally Posted by The-Fox
I already ruled out Uncore, because I can run 20x uncore mult fine no problems.
Specs see sig, it's a DFI x58.
how are you guys running your uncore speeds so high? I leave my uncore ratio on auto. which sets it at the lowest ratio there is. I have tried setting it to 16x but get a no post.
High toll on the cpu - I don't know.Quote:
Originally Posted by jcool
I am able to run 211 bclock all day with qpi at 1.335v (near intel max)
I thought that if I had a lower qpi multi - say 34, 32, ... I could push bclock higher.
I am trying to improve my understanding of i7 relative to cpu speed, hyper threading, vcore and the resulting temperatures.
This little experiment was executed on a 965, ggbt extreme mobo - on an open case (tech station) with water cooling (so, note – YMMV).
Starting with the cpu at default speed – 3.2ghz – the average temp under load is 46C.
Moving up to 4.0 HT on – the temp under load increases to 54-55.5C, depending on the cpu mutliplier selected.
Moving up again to 4.2 (close to the max this i7 can do) the average temps under load climb to ~75.4C.
Interesting to note then that the increase in speed from 3.2 to 4.0ghz costs ~8-9C, while the increase from 4.0 to 4.2ghz costs another 20C!
Another observation is comparing hyper threading on to hyper threading off.
At 4.2ghz, (remember - under water) the temp differences appear to be ~3C, HT off being less stressful.
Interesting that the difference was not as pronounced at 4ghz.
The last observation is that C1E seems to have more effect at higher multi/lower bclock (in lowering temps).
I suspect this might be due to C1E lowering the multi while keeping the bclock constant.
This can be seen in the realtemp 2.90 display when at idle.
That is all.
Looking at your sig, you are running 3x1gb of ram. Try 3x2 and you'll notice you'll need at least +0,15-0,2 VTT to run that kind of BCLK, if you can even get it stable.Quote:
Originally Posted by SteveRo
Yahh I saw you writing on the Uncore.Quote:
Originally Posted by jcool
Missed the DFi on your sig (are you sure you didn't just add it right now ? :p:).
So you are telling me that you couldn't run 12x memory at stock CPU speed ?
That is really weird.
i have been running 211x19 with 1.356 vcore with qpi set to +0.150 and DRAM voltage set to 2.0v for about a week now. Never had a crash while using the PC for what i normally do. Been playing games like crazy, Backing up DVD's and never had a problem. But the instant i start Linx or Prime i get a BSOD. Guess its not enough voltage somewhere but i have fiddled and fiddled with settings and cant get anything stable. Will not even stabalize at 1.44 vcore. I have come to 2 possible causes. 1 Being as stated the OMC. My ram should be able to do DDR3-2000 but it wont. The other issue i think is HT. I am pretty much fed up with trying to get 4ghz stable. Id rather just run it at default then any other speed than 4ghz.
I believe the ram has to be half of the uncore, if the uncore is 16x the ram will only run at 8x. I had the same issue. I at all default settings I would set it at 10x for the ram and it'll only run 8x due to the uncore default being 16x. Please correct me if i'm wrong, but I think that is why.Quote:
Originally Posted by The-Fox
I think there is a +1 in there some where - something like uncore/mem multi needs to be 2x plus 1 clock?Quote:
Originally Posted by Chicken Patty
When I am running high bclock I set multi's to qpi, uncore, mem: x36, x13, x6.
hmm, not sure where that +1 clock is though, the lowest I can set the uncore on mine is 16x.Quote:
Originally Posted by SteveRo
Well at least not on stock VTT (which is 1,21 with the DFI), no.Quote:
Originally Posted by The-Fox
And no, I didn't just change my sig, changed it last week already to i7 ;)
Not quite. Uncore speed has to be AT LEAST 2x as high as dram speed, or the system will be unstable. On the DFI board, you can't select uncore lower than 2x the dram frequency, for instance.Quote:
I believe the ram has to be half of the uncore, if the uncore is 16x the ram will only run at 8x. I had the same issue. I at all default settings I would set it at 10x for the ram and it'll only run 8x due to the uncore default being 16x. Please correct me if i'm wrong, but I think that is why.
However you can run uncore faster than 2x the dram w/o problems, like I am doing right now. Default uncore mult for 6x dram is 12x, however you can also run 6x dram 16x uncore or 6x dram 20x uncore etc. The Gigabyte has even more uncore mults available, some of them working better than others.
For general OCing, set uncore to 2x the mem speed because that setting will always work. You can always try to increase uncore later.
^^^thanks.
i am at 211x19
memory is at 4x and uncore is at 8x. any higher on the uncore and it will not load windows.
what is the lowest qpi multi that your board allows - the ggbt boards only go down to x36 :shakes:Quote:
Originally Posted by liftedcj7on44s
the only option i have is slow or fastest.
Ouch. You need to get rid of that garbage MSI board. :down:Quote:
Originally Posted by liftedcj7on44s