4790k Delid Summer 2014 Bash!!

Sounds about right. My delid job is real half assed. Just using nt-h1 on the die and then more paste between ihs and coldplate of the smallest AIO you can get. Still made a massive difference.
 
lol my Asrock Z97 maxed out my 4790K at 4.3GHz using the auto overclock feature... here is hoping its an Asrock issue and not a historically bad 4790K. Manually its fine at 4.5 GHz on air, but that is hardly a feat..
 
Pitbull said:
I'm using this one.
https://mega.co.nz/#!3tAGnAqZ!QbCz2r1fG0WjM8DgGYeExngGypaHftAzPUgTSn2kAdk
Click to expand...

Do you happen to know which encoder that test is using, or if there is a front-end GUI for the included encoder?

I'm back to testing out Avidemux on this build, but only getting 10-20% CPU utilization during the 1st pass (will only do 4 threads), and only 40-50% CPU usage max during the 2nd pass.

The x264 test you posted uses 100% of the CPU while encoding, and at decent temps.
 
AFD said:
The x264 test you posted uses 100% of the CPU while encoding, and at decent temps.
Click to expand...

It's actually pretty good so far... It did force me to lower my Ring multi down to 45X. I'm sure I've got more tweaking to do, but you can tell when a chip is stable at a given frequency, even if you might need to add a little voltage to get it passed something nasty like p95. If you are having trouble even getting into windows, that's not good. But when everything works fine, you are getting good benchmarks, and only p95 or x264 brings you down, you are close to full stable.
 
It's actually a cool stresser, unlike synthetic prime and others.
I settled on 4.7ghz on lower voltage, runs much cooler and stable.

Bluesun311 said:
It's actually pretty good so far... It did force me to lower my Ring multi down to 45X. I'm sure I've got more tweaking to do, but you can tell when a chip is stable at a given frequency, even if you might need to add a little voltage to get it passed something nasty like p95. If you are having trouble even getting into windows, that's not good. But when everything works fine, you are getting good benchmarks, and only p95 or x264 brings you down, you are close to full stable.
Click to expand...
 
Got my 4790K from MC. It's a L419B538, and I thought I lucked out. My case was out of stock (Corsair 550D), so I tried the stock cooler just to see how it did. Fail! Hits 100C at 4.3Ghz and throttles running Rosetta@Home (Boinc). My i7-860 overclocked ran that for years 24/7. I do have a H110 for it, but it will not fit in the new case like Corsair said it would. Not happy. :mad:
 
The Reaper said:
Got my 4790K from MC. It's a L419B538, and I thought I lucked out. My case was out of stock (Corsair 550D), so I tried the stock cooler just to see how it did. Fail! Hits 100C at 4.3Ghz and throttles running Rosetta@Home (Boinc). My i7-860 overclocked ran that for years 24/7. I do have a H110 for it, but it will not fit in the new case like Corsair said it would. Not happy. :mad:
Click to expand...

Haswell and DC are both terrible designs in the thermal category. Not only is it a bad design but the resultant gap between the IHS and die varies a lot it would appear. Some people get chips that can run fine without being delidded, some need to be delidded to run normal temps stock.

But it's actually a LOT easier to delid them than I thought, but it's something that isn't much fun the very first time and I wouldn't want to do it without a decent vice. Now that I've done a few chips I'm not nervous about it and it even makes it easier. I have this piece of wood I use with a knot in it, it's very very hard right on that spot and I line it up right with the chip and just bang on it for 30 seconds. It will start to give and then you don't whack it so hard. They come apart a bit easier if you run a stress test on em for a few minutes and then pull the chip out, it stays warm for a good 5 minutes. I didn't really think a hair dryer helped at all.

I have learned how to bare die mount but it's really not necessary. Just put the IHS right back on there with some NT-H1 and you will see 20C less or even better.
 
Zangmonkey said:
I call BS on this.
Do you have a link?
Click to expand...

A link to my personal findings? Here: link. I started out with a 4770S that would throttle itself in AIDA64 @ 3.9Ghz. Dropped in a 4770k directly after and it didn't have thermal problems like that. It was stuck at 4.2Ghz. I've tested a lot of Haswells. 10 now, counting that S chip.

Sometimes the gap between IHS and die is larger than others. This isn't rocket science.
 
Bluesun311 said:
A link to my personal findings? Here: link. I started out with a 4770S that would throttle itself in AIDA64 @ 3.9Ghz. Dropped in a 4770k directly after and it didn't have thermal problems like that. It was stuck at 4.2Ghz. I've tested a lot of Haswells. 10 now, counting that S chip.

Sometimes the gap between IHS and die is larger than others. This isn't rocket science.
Click to expand...

You're suggesting that the reference design hits thermal throttles while in its designed operating spec... is that right?
 
Zangmonkey said:
You're suggesting that the reference design hits thermal throttles while in its designed operating spec... is that right?
Click to expand...

If the IHS isn't making good contact with the die... Yes. And Intel promptly replaced my throttling 4770S with a retail packaged one. Which I dumped on eBay.

The overclocking potential of the chips has nothing to do with their thermal characteristics, but both vary widely on the Haswell. In my admittedly very limited experience some chips are just hotter than others but that one did throttle at stock speed. With a Hyper 212+. Multiple seatings all similar temps.

I bought a 4770k to replace it with and it would hit 85C on the same setup on the same day at 4.2Ghz in AIDA64. Big difference.

They can't be binning the chips on anything but other than where they are on the disk, or whatever you call the silicon thing they stamp them out of all at once, right? Not like they test them do they? I have no idea about the binning process. But I test chips like a monkey so I can give you some observations. If I were in a different tax bracket I would be messing with X79 right now, but I'll wait for Haswell-E.
 
melk said:
How much of a hassle is it to do bare die?
Click to expand...

It's a real pain in the ass, unless you have someone else's example to follow or you just get it right the first time out of blind luck.

There's a really good thread about rigging your own mounting posts here. It's a test of buncha thermal pastes with an H100i and a 3770k iirc, but it has the best explanation of bar die mounting I've seen. Unfortunately, I don't have an H100i and using an H55/50 style mounting system is much different.

I finally got my H50 correctly mounted bare die... but it's stupid. I have an XSPC Raystorm CPU block sitting here I'm planning to use soon and if you just use the backplate from the Raystorm kit and the mounting pins from the H50, it works OK.

My 4770k is at 4.6Ghz p95 28.5 stable on blend and it stays in the 60s for the little time I let it run using the smallest AOI on the market (afaik). I already know what voltage it's stable at on p95 from running it on air so I don't long term test these chips anymore. Not after I had a leak on a H80i and Kyle Bennet's 4.8Ghz chip started degrading. I've never had a bluescreen from any regular application when my Haswell could run p95 28.5 low FFTs for even just 30 seconds. That's my experience, not a fact, but it's pretty much how I evaluate a chip quickly these days. I do spend a lot of time benchmarking the chips I keep, but to get a basic idea of what a chip can do I throw it on p95 28.5 small ffts. If it's ballpark stable it won't crash for 30 seconds... if it's not it will pretty much instantly crash. Nice and simple... but DC is so hot at just a tiny bit higher clocks 200-300Mhz more... and I couldn't do stress test like that on my DC at 4.8Ghz without throttling until I delidded. Yes. I let my DC throttle for a couple seconds. No big deal. After delidding it can do low FFTs but it's still nasty on an H55. I'm interested to see what a proper loop will do for the temperatures.
 
Bluesun311 said:
If the IHS isn't making good contact with the die... Yes. And Intel promptly replaced my throttling 4770S with a retail packaged one. Which I dumped on eBay.

The overclocking potential of the chips has nothing to do with their thermal characteristics, but both vary widely on the Haswell. In my admittedly very limited experience some chips are just hotter than others but that one did throttle at stock speed. With a Hyper 212+. Multiple seatings all similar temps.

I bought a 4770k to replace it with and it would hit 85C on the same setup on the same day at 4.2Ghz in AIDA64. Big difference.

They can't be binning the chips on anything but other than where they are on the disk, or whatever you call the silicon thing they stamp them out of all at once, right? Not like they test them do they? I have no idea about the binning process. But I test chips like a monkey so I can give you some observations. If I were in a different tax bracket I would be messing with X79 right now, but I'll wait for Haswell-E.
Click to expand...

So you had a defective item.... That doesn't point to a fundamental problem with the design.

And yes, I assure you, Intel does sample-test the retail products with the IHS attached.

They certainly don't test every unit (very few companies in very few industries do... .maybe Aircraft fuselages?)
 
Can a delidded 4790K be run with a CM Hyper 212 EVO directly on the core, or would it need a shim, or is it a bad idea too prone to chip damage?
 
Bluesun311 said:
So the 4770 and 4770k both boost to 3.9ghz and the 4790 does 4.0 but the k will boost to 4.5ghz? Sounds weird. And unlikely.

Also pretty sure the rumors that DC will only work on z97 are not rumors.
Click to expand...

Certainty doesn't get you as far as it used to :p
 
Bluesun311 said:
Wow this shit get's reaaaaaaly hot.

I delidded mine and it dropped 25 degrees at 4.7Ghz with Large FFTs in Prime. Then I started working with Prime95 v28.5 small FFTs at 4.8Ghz/46x cache. 1.3v Core 1.25v Cache. It hit 100C pretty much right away like that... but it didn't crash, I'll just quit pushing this chip until my loop is finished o_O

Devil's Canyon + tiny AIO cooler not good mix.

If I had to guess, this chip is 4.8Ghz stable.
Click to expand...

Purely 28.5. Test on 27.9 or something non synthetic ala x264. 28.5 is pure self limitation imo
 
Not dead guys. Appreciate all the activity in this thread. Been working on family fun this summer. Disappointed with haswell overall.
 
SonDa5 said:
Not dead guys. Appreciate all the activity in this thread. Been working on family fun this summer. Disappointed with haswell overall.
Click to expand...

Saw your post in the 5770k thread earlier today and figured you were alright ;)
 
Just watched a really cleanly done 4670k delid from one of the guys at Chiphell, and figured Bluesun might enjoy this..

6or5ATh.jpg


http://v.youku.com/v_show/id_XNzM4MTE3Njcy.html
 
FrostyAMD said:
Tried to watch video but no go using the link posted, Any help ???
Click to expand...

Just tried it again, and it gave me a weird error message in the video, until I turned off the Google translate thing that popped-up at the top.. played fine after that.
 
Got the ad to load using hola plugin for chrome. Saw a pretty awesome Chinese ad for Banished. But then the page auto-reloads before the video and then I have to watch the ads again....
 
looks like a neat way to de-lid might have to try that one. Does it work on ivy and older haswells too??. No risk of damaging pcb ??
 
You must log in or register to reply here.
Back
Top