Widescreen Gaming Forum

I think I've finally settled on a Core i7 overclock. It's a little warm, but I think I can shave a few degrees off buy taking the voltage down a few notches. We'll see. Failing that, I can get some stronger fans, as while quite, the fans I'm currently using don't push a whole lot of air...

For those who read the Latest Acquisitions thread, you'll remember that I had some problems getting 3.6GHz stable, and so pretty much chickened out for the time being and backed off a mile.

Since then, I got bored, and got tweaking.


1.35v (BIOS) 1.31v (CPUz) 1.298v (Real) = 4.18GHz boot into Windows. Well, actually, 1.24v (CPUz) got me this as well, but it'd crash as soon as I opened Paint. This BSOD'd with LinX after about 15 seconds... and as I don't want to push the volts any higher until I slap I waterblock on, I backed it off. Actually, come to think of it, it might be stable here if I just disabled all the throttling features. However.


3.8GHz is solid. A bit warm, but solid.


3.8GHz loaded with Prime95. Hour shot. Got bored after that and wanted to check LinX stability. It's about 3*C warmer than I really want to see it, but if I take Vcore back a little it should drops temps a bit.


30 minutes of LinX (will do longer tomorrow) testing the CPU/RAM. To note: my 3.6GHz wasn't stable because LinX would crap out after 9-11 minutes. This is apparently because there isn't enough QPI voltage, as when I tweaked that, it didn't BSOD on me. Also, the QPI multipliers are quite finnicky - once I realised that quirk, my 3.6GHz was stable. But 3.8GHz is more interesting. :D

...

Might put it back at 4GHz, see how it handles the temps (and stability). I'll try 4.2GHz if I watercool it I think.

But there we are. Once the QPI multiplier thing was realised, clocking Core i7 became a lot simpler. The temps aren't quite as bad as they look as they're core temps, which are higher than the reported 'Tcase' of the CPU (hell, the Gigabyte Easytune monitoring utility and the BIOS report the CPU temp as 16*C at idle and 54*C loaded with LinX.

So. There we are. 3.8GHz.

With all that talk about fans, I gotta ask what aftermarket CPU cooler you're using? I didn't notice you mention one at all.

Here's a pretty good video guide on OCing an i7 CPU btw. It gets kinda funny at the end too. It's taken from some pretty knowledgeable guys at forums like XtremeSytems.org.

http://www.youtube.com/watch?v=iwbzQblRQ7I

I'll probably go with a 920 or 940 i7 on my next build around year's end. Which one I get will depend on the pricing on the 940. I don't think I'll go above 4GHz, and I know the 920 can go that high, but the 940 is a higher bin chip, so it should remain stable longer at that speed.

I'm still stock at 2.6GHz. I think I'll have to tweak mine soon. I wish someone made a low profile fan for the i7. Can take anything over 100mm tall in my Antec case.

Yeah, the heatsink is a TRUE120 with fans in push/pull. This thread was only really made so I didn't keep cluttering up the Latest Acquisitions thread. :lol: I lapped the heatsink - not a perfect lap, but infinitely better than the stock finish. It was keeping it really cool on the bench (10*C above ambient) but in my P180 the heat just builds up. As I said, the fans I've got in it just don't shift air fast enough.

Thing I've found is... you can take advice from others all you want... try their settings out with identical kit and everything. Every time I tried using someone elses 'known good' settings, all that would happen is the board would fail to even POST.

Unless you're desperate for that 22 multiplier instead of the 20 of the 920, I don't think the 940 is really worth it... at least in the UK where it's twice the price of the 920 for an extra 266MHz.

i7 puts out a lot of heat when it's overclocked. Well, actually, if you load it at max it puts out a lot of heat at stock, too. On an open bench it's controllable a lot easier than in a closed case situation. Any situation that rides the RAM hard pushes CPU temps up quite a bit too, due to the IMC. The Athlon64 never had temps increase from hard RAM use as much as this in my experience of those chips. Volts needed to get i7 to 3.6GHz is barely any above stock - some chips will do 3.6GHz on their stock volts. 3.8 needs a little push from what I've seen, and the required volts (and therefore temps) really start to climb above that.

As a note - I could take 10*C off those temps by turning off HyperThreading. I tried it. However, that has an impact on encoding times (a fairly significant one as it happens) so turning of HT is an imperfect solution for lowering temps.

Do any of you i7 owners own GTAIV for PC? Can you see any difference in FPS before and after? My frames are chugging in Triplehead and I'm really frothing at the mouth for a new CPU.

http://www.pcgameshardware.com/aid,672004/GTA-4-PC-Intel-Core-i7-far-ahead-of-Core-2-Quad-in-CPU-benchmarks/Reviews/


Doesnt; lsit y our q6600 but it gives a pretty good idea of what you coudl expect.

Thing I've found is... you can take advice from others all you want... try their settings out with identical kit and everything. Every time I tried using someone elses 'known good' settings, all that would happen is the board would fail to even POST.

Unless you're desperate for that 22 multiplier instead of the 20 of the 920, I don't think the 940 is really worth it... at least in the UK where it's twice the price of the 920 for an extra 266MHz.

Well you generally get more similar results to what most others do if you start out with water cooling or a case with high air flow. The P180 was good for it's time, but you need something more like their P193 with that kind of gear. The case I'm liking right now is the Thermaltake ElementS. It's very well made, very affordable, and the airflow is so good there is very little change between idle and load temps. It's also pretty quiet and has dust filtration.

Also, the 940 may be physically the same chip as the 920, but being a higher bin spec means it's passed more stringent thermal imaging testing, so it doesn't leak current through it's transistors as bad and thus less heat produced. Chances are a 940 will run more stable at higher speeds. That's why they cost more, not due to a mere 2 digit higher multiplier. The TRUEs work pretty well to manage that heat though. I think it's more your case than your CPU cooler.

BTW, when you ramp up your bus frequency are you doing it in no more than 10MHz steps?

Also, the 940 may be physically the same chip as the 920, but being a higher bin spec means it's passed more stringent thermal imaging testing, so it doesn't leak current through it's transistors as bad and thus less heat produced.

The 940 is NOT the same chip as the 920 right now, Its getting phased out and getting replaced by the 950 for the same price with a higher clock speed, using the same chip that the 920 has now.

Well, I just took the Vcore down again after my attempts to get 4GHz. 3.8GHz is currently stable (although it's only after fifteen minutes) of LinX at the same votage I had 3.6GHz at. 1.275v (BIOS) 1.21v (CPUz, idle) 1.184v (CPUz, load). I found LinX tends to BSOD after <2 minutes if there isn't enough Vcore, so I'm hopeful this is good.

The other thing the votage drop has done is take 10-11*C off the load temps, and a good 8*C off the idle temps.

I'll post a screenie when it's done 30 runs... if it makes it. It looked like it will. GFlops has gone up - 49.87 is the top now, as opposed to the 48.2 or so it was yesterday. So far, the whole system is cooler, although I don't think temps have maxed yet. Still, I very much doubt they'll break 75*C on the cores, as it's really floating on 70-71*C right now.

...

All the 940's are C0 stepping chips. The 920's used to be C0 step, but are now D0. The 940 is being replaced by a D0 step 950 (@ 3.06GHz stock IIRC) and the 965 is being replaced with a D0 stepping 975 (@3.3GHz IIRC)

I guess binning procedures have changed since I last really looked at them, then, really.

Well, that's it... unless I feel desperate for another 200MHz, I'm gonna stick at 3.8GHz.


Shot during. LinX hammers a system so hard that Windows 7 struggles to keep the screen refreshed - there was a minute or so in that 50.05GFlop run that I thought it had frozen... each time I got a really high reading, the screen would either get really sluggish, or not update for a few seconds. Shows just how hard LinX can push, really.


LinX run complete without errors in 1h20m35s. The 50GFlops is an 18.5% jump from what I was getting at 3.2GHz... which is an 18.75% clock boost. RAM was running at the same speed for both, so that probably accounts for the not quite linear scaling with frequency.

System is now idling @ 41-41-43-38 on the cores, which is about 6*C hotter than stock with all the throttling functions enabled. I don't think 41*C is too bad at 3.8GHz, when it's 35*C at 1.6GHz. ;) Idle temps go up to 47-47-49-44 if I turn all the fans right down to minimum. However, I'm not going to do that often as the fans don't annoy me that much (a lot quieter than the fans in my old system at max) and the GPU fan is always at 100% so I can hear it if I turn all the others down. As the 'whoosh' of air is less annoying than the GPU fan whine, I'll stick to the 'whoosh'. :lol:

Despite RealTemp saying that temps maxed at 74-74-72-70, I didn't actually see it hit 74*C - that occurred in one of the 'slow refresh' moments and it usually floated around 72*C +/- 1.

I'm happy. That's faster than I wanted it, with temps I think are acceptable (<75*C) on the cores. My Core 2 Quad hit 70*C on the cores when it was running at 3.4GHz, so I don't think that's bad at all. I finally seem to have got my hands on a half decent overclocking chip. :lol:

...

Did check out that Thermaltake Element S case... I'm firmly in the 'plain simple understated cases' camp (like the P180 and the Silverstone TJ07) I think - it's not as bad as some cases I've seen, but I still thought it was ugly. It's got some interesting design ideas, but also I need dust filters everywhere on this rig, as the room it's in is fairly dusty.

Looking good! Temps seem to be pretty much at the threshold of acceptable. It's a pretty fine overclock as it is, anything more I guess would be 'just to see' rather than running it like that 24/7.

Gonna be keeping an eye on my monies anyway ... :P

The 940 is NOT the same chip as the 920 right now, Its getting phased out and getting replaced by the 950 for the same price with a higher clock speed, using the same chip that the 920 has now.

If that's the case, then you are comparing the 950 to the 920, not the 940, which IS the same chip, just higher bin/settings spec. Also, I said the same PHYSICALLY. If a chip is physically the same, but is clocked and multiplied differently, it still gets back to the point I'm making, that it's achieved because it's passed thermal testing better, NOT because it's any different in physical design.

In fact, if anything, what you just said further validates the point I made. Many have complained about the 940 being double the price of the 920, but this tends to verify it's more capable than they were originally specing it to, and likely more stable for a longer period of time than the 920.

To be honest though, I really don't think that's the reason Intel is making this move with both the 940 (and 965 as I've just read whilst writing this). With such a minor increase in speed (3.06GHz, up from 2.93GHz), it's more likely they're avoiding the counter productive mindset that the 940 wasn't even a full 3GHz. Pushing it slightly over that is like consumer psychology I tend to think.

So you have what is essentially the same chip clocked ever so slightly higher, and suddenly Intel gets a sales boost from higher advertised spec and vendors being able to keep the price longer without dropping it due to this so called "new" chip,...very clever. One has to wonder with an odd speed like 2.93GHz on the 940 if they planned this all along.

There is a new 920 stepping out that is supposed to run a bit cooler and has shown better overclocking results.

Bad for old owners but good for those looking to buy now, its the same stepping the new top i7 model will use (the extreme) but only the 920 so far has gotten it.

For cooling your OC down, not sure what cooler that is you have but the Thermalright Ultra 120 was one of the best coolers I could find when I built my rig and it keeps my Q6600 ice cold even in a case without a whole lot of airflow (P182 SE) fans are important.

Also there is a new cooling compound on the block! ICD7 or ICD24 (innovative cooling diamond) it has shown very good results so far, better than AS5 by a good margin and its also non conductive. So maybe look into a tube of this stuff. I have a tube to install on my GPU//CPU in my new laptop when I get in in a couple of weeks.

Hey Vicious, not seen you about for a bit. ;)

I've got the D0 stepping i7 920. I'm using a lapped TRUE120, too. :) I didn't know about that new thermal compound. I might give it a shot when I stick this chip under water.

...

I was coming back to this thread anyway, as I've just pulled the Akasa Amber fans I was running and replaced them with Scythe Kaze Jyuni 1900RPM ones. Noise at full speed has obviously gone up, but ambient temps have gone up about 5*C here in the last week, and my load temps were a lot higher than I liked because of it. At full speed the new fans are fairly loud... although the sound is simply the 'whoosh' of air rather than anything else. When I turn 'em down to minimum, they're quieter than the other fans were, while maintaining similar temps to the old fans. But at full clip, load temps are down a full 15*C... which says that the biggest cooling problem for the old fans was simply CFM.

Yeah innovative cooling started giving away some of the compound to get there name out. They happened to let me be one of the few to get the big tube the ICD24, and the stuff is pretty amazing, every poster so far has reported huge drops in temp from stock TIM and even semi large drops over AS5.

Plus I like the fact its non conductive. The stuff is like 97% diamond dust or something, and since diamond is like the most dense substance on earth it has very good thermal transfer properties.

Its also very thick, so people have used it to replace not only TIM on cores, but thermal pads on ram chips. So for my notebook I plan to give it a shot and replace the pads & core tim with it.

I gotta look up this cooler your both using, see if its similar/better than the Ultra 120 I have.

Im out of funds so no i7 for me but I am still very happy with my 3.6ghz Q6600 and dual 4850 setup. I think it can hold out for 2 years so I can get a larger upgrade.

Im looking forward to SSD prices droping to something more sane too. I have been tempted to pick up a OCZ Vertex or G. Skill Falcon 120gb model for $300 for my notebook, but talk myself out of it before I hit "buy".

Oh the new notebook :P

Q9000 @ 2.7ghz (mobile quadcore) and dual 4870s :P lovin it. Now I can game and encode on the go as if I was on my desktop. Its not a big ugly brick like most other high performance laptops eather, and its cool & quite.

Asus W90VP-A1

Edit: Nevermind about the TRUE 120, I found out what it was, its the same one I have just never seen it abbreviated as TRUE 120 before. I was lucky and got the extreme right when it first came out pretty cheap.

Mine is paired with a 1600rpm s-flex fan seems to be a popular choice, and it blows right out into another one that is my top exhaust.

I've got the D0 stepping i7 920.

This is no doubt why it's common to see people running the 920s at 4GHz with no problems now. I don't like that there's no clear cut way of telling whether you'll get a CO or DO stepped CPU though.

What Intel SHOULD have done is put DO stepping on all 3 i7 models, renamed them 930, 950, 975, and clocked them at 2.9Ghz, 3.1GHz, and 3.4GHz.

the DO is supposed to be a bit better at overclocking, but by no means a land slide from what I read.

Basically you woul want one if you were buying now but if you had the older one already you shouldnt lose any sleep over it.

But yeah a 4ghz OC sounds so nice, given that the i7 is clock for clock faster than my q6600 and an extra 400mhz OC that would increase my video encoding time by a good degree.

I cant wait to see what comes next and hope it overclocks as well as these do.

The only slight negative I've ever read about the i7s is that they can leave data lingering in cache momentarily in some calculations, keeping them from achieving their full potential. I suspect that may have been partly to do with the drivers and/or developer threading for them needing to be perfected just after their release though.

To tell you the truth, I'm not sure even the 32nm versions will be much better in performance, except to offer a bit more overhead for extreme OCing. As far as speed goes, I'd rather they offer higher stock speeds in the non extreme versions like we saw in the Core 2 Duos. I don't really like the idea of cranking them up a lot.

One would think that once quads become the norm, we won't have to keep paying a MHz penalty for the extra cores.

For now they wont because they are still cashing in on it, many people out there are still in the mindset faster dual core = better than slower quad.

Once the slower people catch on and start overclocking the quads maybe it will change :P but thats never going to happen, members of a tech enthusiast forum like ourselves are the small minority.

Actually I think quads will become the norm soon enough, maybe about the time video stores rent more Blu-ray than DVD. One of the reasons I don't like or get into the OCing obsession much is it sets a trend for manufacturers to let the consumer clock them up, rather than doing it themselves. On the one hand, with the OCing trend you get better, more lenient warranties across the industry, esp on GPUs, but you pay for it in heavily dumbed down core speeds on CPUs.

IMO you should if you want be able to go out and pick parts that match up just fine without having to OC to avoid bottlenecks or sacrificing speed of the GPU just to match the abysmally clocked multi core CPUs. To be honest I'm seriously considering getting a 3.6GHz dual core CPU and spend a lot less dough because besides this kind of annoyance there's the fact that more and more PC games are just written and/or ported so lousy. It makes me feel stupid sometimes for taking part in PC gaming so much.

But hey, EA claims PC is the fastest growing platform for gaming lately, mostly due to digital download sales. Hmmm, wait a minute, that IS EA talking, the ones that make idiotic decisions for such type of marketing then go out of their way to try and recoup what they lost. You know, like the 100s of thousands of copies of Crysis that were stolen off their nifty predownload servers that didn't even work properly for paying customers. :roll:

Sorry to sound like a stick in the mud, but I was anticipating 2009 to be a great year for PC titles, and so far it's been anything but that for me.