4770K 90C with H100i @ stock speeds?

Budwise said:
Yea, i tested it with the stock HSF and temps were only 5C hotter. Its just a poorly done IHS...
Click to expand...

If this is the case, that would rule out the other coolers for contact or broken pumps etc, as it shows it is not a problem with the HS/Rad dissipating the heat, but rather the heat being removed from the cores. Which means it is time to contact Intel and request a RMA.
 
To those that read all of my posts thank you for the input. For those that didnt and kept asking the same questions that I already answered thanks anyway. It was not a troll thread, I've been overclocking for over 10 years on various platforms and cooling and I was just having a hard time believing a chip this frickin hot so I made a thread to have some other people double check my logic. Pretty simple.

TLDR: 4770K ran super hot under H100i, I remounted 3-4 times, same results, I tried a Megahalem, same results, I then tried stock HSF and it was a bit warmer. In other words, Intel royally botched that particular 4770K's TIM job. I cant technically RMA because it runs warm since it was after all stable at stock speeds. The plan was give my Grandfather my 3770K+Z77 rig and take this one but alas I gave him the 4770K/Z87 instead since he wont be overclocking or doing anything remotely pushing it (stock market trading mostly).
 
Well what store did you get it from? For instance I know Microcenter would let you take it back.

Also, I don't think seeing only a small difference between another air cooler and the stock one is necessarily a huge deal. What is strange is seeing a huge difference between an AIO and either one of those.Are you sure that h100i is working properly?
 
Yea, I tested the H100i on my 3770K and it worked fine.

I got the chip from Newegg.
 
http://help.newegg.com/app/answers/detail/a_id/181

It's not refundable, but as long as you haven't damaged it, it looks like they may give you a shot at getting a replacement. I'd probably go for it. Just tell them it runs far too hot or something. Newegg has been getting better stuff as of late or so I've heard. But for max cred, try running Intel's testing program and showing that the temps from that are too high, too.

Also is this LinX program as taxing as Prime 95 28.1 from a few months back? I tried running that thing and even on my Kraken X40 I almost instantly broke and bluescreened at like 100C with my overclock. Even at stock I was hitting 83 fairly quick.

Here's the reference thread for Prime 28.1:
http://hardforum.com/showthread.php?t=1793932
 
I will also admit I returned multiple Haswell chips to Amazon before I found an "acceptable" one (not even a very good one). And I'm using a H100 push pull with a Corsair Air 540 case so I have a lot of ventilation and airflow.
 
Delid or get a new one, the thermal paste on a 4770k is crap but it sounds like you have a worse one than usual.
 
StoleMyOwnCar said:
http://help.newegg.com/app/answers/detail/a_id/181

It's not refundable, but as long as you haven't damaged it, it looks like they may give you a shot at getting a replacement. I'd probably go for it. Just tell them it runs far too hot or something. Newegg has been getting better stuff as of late or so I've heard. But for max cred, try running Intel's testing program and showing that the temps from that are too high, too.

Also is this LinX program as taxing as Prime 95 28.1 from a few months back? I tried running that thing and even on my Kraken X40 I almost instantly broke and bluescreened at like 100C with my overclock. Even at stock I was hitting 83 fairly quick.

Here's the reference thread for Prime 28.1:
http://hardforum.com/showthread.php?t=1793932
Click to expand...

Holy shit... That thread's a pretty good read if you have any interest in Haswell or its features. This is somewhat off topic but I'm following your lead. I was there at the outset of that thread and kind of inadvertently started a technical argument between a few other people that goes on for 8 pages more or less and it's surprisingly civil and informative. The gist of it is this: Is prime 95 28.1 for 24 hours (it's for HASWELL ONLY, folks) the ulitimate indicator of stability and what do you do if you can't actually even run it without throttling?

For the second question... the answer is unequivocally to delid. Whether or not you go water you should delid this chip if you want to do this prime95+Valley for 24hours shit. That thread's where I got that idea from though I did not remember. Here's the thing about the second question: I don't WANT to torture my proc. I want to baby it but be reasonably confident that it is stable.

The fact is, I'm overclocking to the edge of my chip's capabilities at what seems to me a reasonable voltage bump--I can't get any more out of it even if I add more voltage than I'm willing to, so this is it. 4.6Ghz. I did prime95 28.1+ valley for about 30 minutes and that was good for me, adding a lot more variables and a little more heat for a long period of time seems like asking for blowback to me; I rarely hardcore straight strafe for more than 30 minutes, and even when I'm doing weird shit that pushes my GFX cards for long periods of time (I like stupid amounts of MSAA on emulators that don't play nice with FXAA) there is never a time where AVX2 comes into play during these activities and that is the ONLY difference between P95 v28.1 and v27.9, to my knowledge. So I actually used BOTH to help validate my overclock. But I know exactly that I have baseline stable settings without an OC on my proc and I know how to make it just barely UNstable at my OC, which helps me to be confident that it is stable when I can back it down to the point that things lock into place completely.

So how do you avoid the 24 hours of Prime95?

Use it to validate certain custom configurations that you know have caused a crash--recreate these same conditions (they will generally be either RAM speed/voltage related or CPU speed/voltage related) instead of starting the test all over again. Blend is like... set it forget it--well, I have a $50 watercooler sitting right on top of $1000 in GPUs. So, I'm not so into set it and forget it.

Here's the thread that will teach you how to use custom prime95 configs:
Prime95: A Quick & Dirty Guide To The Custom Settings

Generally speaking, if you are stable everywhere but prime95, you are looking at low numbers for the FFT size. Skip to those and if you can run them for a few hours without crashing I'd call it a day. The other trick I pulled so that I could run prime95 for 24 hours on my proc STOCK to validate the rest of the hardware, without worrying I'd spring a leak, was to use a good air cooler like a Hyper 212 and then sell it. I have 2 H50s that are both good and if the pump dies on one I have a backup.

As for P95 28.1 showing instability that other stress programs wont: In the thread linked above, I am on the other side of this argument from where I am now, because now I have a lot more experience with Haswell. It's true. p95 28.1 will find weird issues that AIDA64 and IBT any really short test wont find.
 
Budwise said:
To those that read all of my posts thank you for the input. For those that didnt and kept asking the same questions that I already answered thanks anyway. It was not a troll thread, I've been overclocking for over 10 years on various platforms and cooling and I was just having a hard time believing a chip this frickin hot so I made a thread to have some other people double check my logic. Pretty simple.

TLDR: 4770K ran super hot under H100i, I remounted 3-4 times, same results, I tried a Megahalem, same results, I then tried stock HSF and it was a bit warmer. In other words, Intel royally botched that particular 4770K's TIM job. I cant technically RMA because it runs warm since it was after all stable at stock speeds. The plan was give my Grandfather my 3770K+Z77 rig and take this one but alas I gave him the 4770K/Z87 instead since he wont be overclocking or doing anything remotely pushing it (stock market trading mostly).
Click to expand...

That might be the case with Newegg and others, but Intel has always been very good with RMA, and we are talking about a K chip that is meant to be OCed, I would hit them up anyway with the problem and would be willing to bet they give you a RMA for the chip.
 
Somewhat related rant...
You know, I regret getting this chip. The overclocking of these 4770k's is just so blah that it's not even funny. I don't mean speed as I know some people are getting these into the higher clocks just fine. I just mean the overall experience. The way it's pretty much a lottery through and through, and for some chips it's pointless to even try, and for others it's just easy as heck. And the way that testing the overclocks is so shaky unless you delid. IE doing one stability test doesn't mean you're stable, but unless you delid you can't even realistically complete the more demanding ones. Put simply, it's no fun. It's just stressful. Unless you've got a better chip (which is stress in itself), you've got to really try to get stable with mediocre clocks (but you never know because you can't run more demanding tests regardless of cooling?), so you're faced with a choice of putting up with mediocrity or delidding (which voids your warranty and is stressful in and of itself).

I think they really messed up on these. And heck, most of it was because of a stupid, sloppy design decision by Intel. All I know is my money on this Kraken X40 feels wasted; this was the first chip I really wanted to try OC'ing to its limit. I can't ever be 100% certain my overclock is valid because there's no way it could pass Prime 28.3 or Linpack no matter how much voltage I pump in because I know I can't keep the heat off because of some extra MM of spacing between the friggin top of the CPU and the chips. Ugh. I really should have just listened to what the reviews said...
 
You must log in or register to reply here.
Back
Top