Ivy Bridge (3770K & 3570K) Results and Discussion thread

[Archer0915]

got a screen shot of IBT at max on your oc,with the specs just so i can compare

I have no idea why it was reading 4.4 in Core temp it was clocked at 4.4. This is from a set voltage run.

 

[Qatar]

IBT max 4.7G 1.375V LLC 75% 16GB 9-11-11-28-1t max core 83c

 

[Archer0915]

IBT max 4.7G 1.375V LLC 75% 16GB 9-11-11-28-1t max core 83c
View attachment 110826

Looking good. If water was not against my faith I would be a bit cooler.

 

[RGE]

are you using prime 2.77? and do you test with IBT at max,i needed i think for 4.6 think it was 1.25v LLC 75%

yep, prime 27.7 for stability used it first, also ran 26.6 just to put in database on one forum that for unclear reasons is using 26.6, and can decrease vcore 2 notches and still be stable with 26.6 and no whea erros. I can run 27.7 at 2 notches below 1.314 for 15hrs, but get whea errors, had 8, 5 of them on 40kffts run. 2 notches higher at 1.314 real load, I get no whea errors on 40k ffts, then easily passed 27.7 again with no whea errors. Now i use 40kffts and look at whea errors for screening stability before long runs.

yep, I ran 4.7 with linx 5gb mem, I use 1gb less than max free, here is pic at load near end. I have 8 gb ram. But I can run linx at lower vcore than prime 27.7, mainly because linx doesnt seem to generate whea erros.

 

[RGE]

rge, what I am seeing with my 3570K at 4.8 is looking much like thermal instability. 4.7 isn't too much a problem getting stable (so far) but 4.8 is a totally different story. Like you, I am seeing the need for a much greater bump in vcore, which then compounds the heat problem with the processor due to Intel using a very inferior TIM between die and IHS. It seems like I can almost get it stable, but then temps climb and stability goes down the crapper yet again. I even tried up to 1.43 vcore (under load) and using a pretty decent loop for cooling and last P95 run I tried at that setting locked up sometimes after Core 1 hit the 100 C mark. With the crappy TIM they are using, there is just no way to effectively cool IB (ambient cooling) once it hits the spot where vcore starts ramping quickly. And that is different from what I see with my 2500K. I was using it at 4.7 as a 24/7 clock, but I can get well past 5.0 when benching with acceptable temps and stability. Also, I am seeing that with my 3570K and my 3770K, vcores when overclocking to past 4.5 run no lower than what I was having to use with my 2500K and 1 of my 2600K procs. I am thinking that Intel needs to spin through a few more steppings to refine 22 nm before we see decent ambient results on this process.

yeah I agree, mine may be thermally related as I get up there in volts as well... will be interesting to see the improvements with tech refinements...and maybe if intel gets tired of reading about tim issues and delidding, they will consider solder again for Haswell, at least for K chips. One can hope anyways.

I did my small part of protesting non-solder tim via intel support a few days, I asked support via chat "when I remove the IHS on my cpu to cool my 3770K, since intel went with cheaper paste instead of thermally better solder, do I need to use non-electrically conductive tim for bare die cooling?" I went through two guys, before someone told me something other than, "that is not recommended." Finally I got passed off to someone that said they thought it was better to use non electrically conductive, and he added that it voided my warranty. I asked him why they used cheaper paste on the overclocking K chips, and whether they were going to switch back to solder, but at that point he was getting annoyed and said he couldnt discuss those subjects via this outlet, so I thanked him for his help and said goodbye. Everyone was polite, as was I.....I just hope the story gets passed around and up the chain like... you wont believe what this one idiot kept asking me.

 

[Qatar]

yep, prime 27.7 for stability used it first, also ran 26.6 just to put in database on one forum that for unclear reasons is using 26.6, and can decrease vcore 2 notches and still be stable with 26.6 and no whea erros. I can run 27.7 at 2 notches below 1.314 for 15hrs, but get whea errors, had 8, 5 of them on 40kffts run. 2 notches higher at 1.314 real load, I get no whea errors on 40k ffts, then easily passed 27.7 again with no whea errors. Now i use 40kffts and look at whea errors for screening stability before long runs.

yep, I ran 4.7 with linx 5gb mem, I use 1gb less than max free, here is pic at load near end. I have 8 gb ram. But I can run linx at lower vcore than prime 27.7, mainly because linx doesnt seem to generate whea erros.

nice oc mine wont do that,what do you have your memory running at?

 

[RGE]

nice oc mine wont do that,what do you have your memory running at?

thanks, havent played enough with mem settings yet, but
Prime stable settings so far on my memory
1866 at 9,9,9 27 1.5v
2000 9,10,9 27. 1.53v
2133 9,11,10,27 1.5v

currently running 2000 9,10,9 27 1.53v.

For benching spm1 I can run it at 2400 10,11,10, 1.68v. 32m havent played with enough yet, probably secondary timings would help, otherwise have to loosen prim timings up too much regardless of vdimm, vtt, imc v. best time so far with 32m was with 2000 9,9,9 27 which is very stable at 1.6v.

 

[Archer0915]

Well I think I have made my decision as it pertains to the IB CPU.

We got used to easy, high and relatively cool overclocking with the Sandy Bridge. Comparing this to Sandy Bridge is almost like comparing the BD to the PhII x6. If you have a SB don't upgrade unless you can use the extra features or you dont OC.

I got mine because I needed it not because I wanted it and though it was, is and will be fun to play with, it is still a little anticlimactic.

So I think those who live off of the IGP or game a bit with GFX bound games could make use of it and but as a replacement CPU it is not worth the money unless you want a challenging OC.

Just my two cents.

Upgrade? NO! Not for anything over a 2500K until the low power model duals and single core celery hit then the game changes. At stock this is the better of the two any day and those benefits may be a win for IB in the SFF/HTPC sector.

New purchase for non extreme overclocking? Yes and no, because you will get better performance at stock clocks all the way up to the limit which seems to be 1-500 less on air with reasonable temperatures. So 4.4-4.6 seems to be the max due to the temps on air.

New purchase for not overclocking? Yes and yes.

 

[sk1llz]

105C is the max... keep it below that however.

Thanks

Yes, will definately stay below 105C, i'll even stay below 90C just in case, that's hot enough for me, hehe, almost too hot!

 

[ktownhero]

Well I think I have made my decision as it pertains to the IB CPU.

We got used to easy, high and relatively cool overclocking with the Sandy Bridge. Comparing this to Sandy Bridge is almost like comparing the BD to the PhII x6. If you have a SB don't upgrade unless you can use the extra features or you dont OC.

I got mine because I needed it not because I wanted it and though it was, is and will be fun to play with, it is still a little anticlimactic.

So I think those who live off of the IGP or game a bit with GFX bound games could make use of it and but as a replacement CPU it is not worth the money unless you want a challenging OC.

Just my two cents.

Upgrade? NO! Not for anything over a 2500K until the low power model duals and single core celery hit then the game changes. At stock this is the better of the two any day and those benefits may be a win for IB in the SFF/HTPC sector.

New purchase for non extreme overclocking? Yes and no, because you will get better performance at stock clocks all the way up to the limit which seems to be 1-500 less on air with reasonable temperatures. So 4.4-4.6 seems to be the max due to the temps on air.

New purchase for not overclocking? Yes and yes.

I agree with your first and last statements, but not the one about new buyers. Even if you plan on being an OCer, Ivy is worth it. Yes, it's true, you cannot take it clock for clock as high as Sandy, but the performance difference works out in the end. And with Ivy you have additional benefits, and may very well have a more resilient chip.

Absolutely wouldn't recommend an upgrade from Sandy, but I would say any new buyer whether they are an OCer or not might as well just get it. Especially if you live by a Microcenter

 

[Tank Geek]

Well I think I have made my decision as it pertains to the IB CPU.

We got used to easy, high and relatively cool overclocking with the Sandy Bridge. Comparing this to Sandy Bridge is almost like comparing the BD to the PhII x6. If you have a SB don't upgrade unless you can use the extra features or you dont OC.

I got mine because I needed it not because I wanted it and though it was, is and will be fun to play with, it is still a little anticlimactic.

So I think those who live off of the IGP or game a bit with GFX bound games could make use of it and but as a replacement CPU it is not worth the money unless you want a challenging OC.

Just my two cents.

Upgrade? NO! Not for anything over a 2500K until the low power model duals and single core celery hit then the game changes. At stock this is the better of the two any day and those benefits may be a win for IB in the SFF/HTPC sector.

New purchase for non extreme overclocking? Yes and no, because you will get better performance at stock clocks all the way up to the limit which seems to be 1-500 less on air with reasonable temperatures. So 4.4-4.6 seems to be the max due to the temps on air.

New purchase for not overclocking? Yes and yes.

I have been debating and researching this topic and found to get the IB over the SB. I am building my first rig and want to OC. So are you saying to get the SB?

 

[Tank Geek]

I agree with your first and last statements, but not the one about new buyers. Even if you plan on being an OCer, Ivy is worth it. Yes, it's true, you cannot take it clock for clock as high as Sandy, but the performance difference works out in the end. And with Ivy you have additional benefits, and may very well have a more resilient chip.

Absolutely wouldn't recommend an upgrade from Sandy, but I would say any new buyer whether they are an OCer or not might as well just get it. Especially if you live by a Microcenter

ktownhero, You were the one who originally talked me into the IB so I expect this from you. I think I will stick with the IB

 

[KillrBuckeye]

I have been debating and researching this topic and found to get the IB over the SB. I am building my first rig and want to OC. So are you saying to get the SB?

The consensus seems to be that if you already have SB, going to IB is probably not a worthwhile upgrade. If you have something older, there's really no reason not to go with IB. I just upgraded my Q6600 system to a 3770K and I am very happy with it.

 

[Archer0915]

I have been debating and researching this topic and found to get the IB over the SB. I am building my first rig and want to OC. So are you saying to get the SB?

No I am saying it is not worth it to upgrade from SB if you want to high air OC unless it has features you really need. If I had to buy new I would get IB but if I were looking at replacing my 2500K I would wait until gets the issues worked out as far as heat and the like or even skip this generation all together.

Ivy is better than SB there is no doubt and unless you want to go for a high OC IB is the way to go.

 

[Tank Geek]

The consensus seems to be that if you already have SB, going to IB is probably not a worthwhile upgrade. If you have something older, there's really no reason not to go with IB. I just upgraded my Q6600 system to a 3770K and I am very happy with it.

Thanks!

No I am saying it is not worth it to upgrade from SB if you want to high air OG unless it has features you really need. If I had to buy new I would get IB but if I were looking at replacing my 2500K I would wait until gets the issues worked out as far as heat and the like or even skip this generation all together.

Ivy is better than SB there is no doubt and unless you want to go for a high OC IB is the way to go.

Thanks I will get the IB

 

[MasterRy88]

Ok so I finally got my rig together, never owend a quad core before. Whats the biggest temp difference I should see between cores. Mine generally sits with core 3 around 10c higher than core 0 with cores 1&2 falling inbetween them. Is this normal?

 

[Janus67]

That's pretty normal. My hottest core sits about 15C different from my coolest.

 

[MasterRy88]

ok good to hear that's normal

 

[RGE]

Playing with fluke multimeter on GB Z77 Udh5 w/ 3770K, looking at vdroop with LLC on HIGH vs LOW, and vdroop of prime 26.6 vs prime 27.7, (though typically use LLC on turbo which little higher than high). My crashes and whea errors are typically in small ffts 40K and 28K especially, so was looking at differences with different ffts as well. Surprised by how accurately the board keeps vcore with multiple reading over several minutes
(note looking at 1sec intervals here, not talking about 1-4 microsecond under/overshoots and fluctuations, just 1 second avg vcore is rock steady).



1) At LLC on LOW, prime stable at 26.6 at 1.310v load by multimeter small ffts, I need increase vcore 2 notches in bios to be stable on prime 27.7, vdroop increases so vcore decreases from 1.311v prime 26.6 to 1.302v prime 27.7 or .009v more with small ffts, (1 notch in bios = .005v).

2) AT LLC on HIGH, I need to increase vcore 1 notch in bios to be stable on prime 27.7 vs 26.6, vdroop is only .004 volts more, ie 1.314-1.310.

3) There is no difference in power consumption (kow or gb sensor), vdroop (fluke), or whea erros between running small ffts on die vs small ffts with 1600 mb memory vs small ffts with 5500 mb memory.

4) Larger FFTS (on die or with 5500 mb ram), incurs less power consumption and less vdroop than small ffts on IVY, example 2048ffts was 1.321v. Also, on a few larger ffts there was slightly less vdroop and power consumption when testing with larger amounts of memory ie 5500mb vs 1600mb or on die, though only .001 to .002v difference, so likely not an issue to use more memory, especially seeing that large ffts dont drop vcore near as much as smaller ffts.

5) The difference between prime 26.6 and prime 27.7 in vdroop and power consumption (10 watts) is primarily in small ffts. Larger ffts less difference between two, and went either way sometimes. But all large ffts way less vdroop hence higher load vcore than small ffts.

 

[I.M.O.G.]

You didn't mention in your post - are those results on the UD5 in your sig?