Haswell Preview Benchmark - Page 34

nucleon · 2013-10-20, 06:02

I confess - I didn't get first go. Yes I saw the vice method and thought no way. But as crazy as it sounds the vice method is the safest option.

I tried the razor method, it's incredibly hard. I chipped the silicon. The vice method is cake. The only gotcha here - make sure you hold the block of wood firm against the vice and don't let it slip across the chip. I also did this too and knocked off one of the surface mount capacitors.

All in all, instead of buying multiple chips and failing, I'd rather have bought 2x systems and underclocked.

-- Craig

nucleon · 2013-10-20, 06:43

Quote:

Originally Posted by Prime95

My two Haswells are still not stable. I get a reboot every few days. I have no clue where the problem lies. I'm trying increasing various voltages so far to no avail. If you achieve a fully stable system, please share the settings that worked for you.

What you might be hitting is the failsafe that you can't disable. When the CPU gets to 100degC and all usual power states are disabled, then the cpu reboots itself.

-- Craig

VictordeHolland · 2013-10-20, 10:18

Quote:

Originally Posted by Mark Rose

To put that in perspective, a copper atom is about 140 pm in radius. So at 22 nm, features are only ~150 atoms wide.

Actually, the space between 2 silicon atoms is much larger than the atoms themselves.
Wikipedia:

Quote:

In its crystalline form, pure silicon has a gray color and a metallic luster. Like germanium, silicon is rather strong, very brittle, and prone to chipping. Silicon, like carbon and germanium, crystallizes in a diamond cubic crystal structure, with a lattice spacing of 0.5430710 nm.

Also, companies are not entirely transparent with their process nodes. xx nm usually means only the smallest structures on the die are of that dimension, but some/many can be are larger.

pepi37 · 2013-10-20, 11:54

Quote:

Originally Posted by nucleon

What you might be hitting is the failsafe that you can't disable. When the CPU gets to 100degC and all usual power states are disabled, then the cpu reboots itself.

-- Craig

http://www.coollaboratory.com/en/products/liquid-ultra/

Try this: is is ultimate thing , have 50-60 % of gallium metal and cannot be used with aluminum heat sinks, but give extreme good results. Temperature will drop at least 10-15°C ( on air cooling systems)

nucleon · 2013-10-20, 12:55

Quote:

Originally Posted by pepi37

http://www.coollaboratory.com/en/products/liquid-ultra/

Try this: is is ultimate thing , have 50-60 % of gallium metal and cannot be used with aluminum heat sinks, but give extreme good results. Temperature will drop at least 10-15°C ( on air cooling systems)

Isn't the IHS Aluminum?

-- Craig

nucleon · 2013-10-20, 22:37

Quote:

Originally Posted by nucleon

Isn't the IHS Aluminum?

To answer my own question. The IHS is nickle coated copper.

-- Craig

TheMawn · 2013-10-20, 23:08

Is there any need to worry about the electrical conductivity of the thermal paste applied directly to the die? Does Arctic Silver 5 work or do I need a special ceramic based paste or something?

Also, do most heatsinks or waterblocks fit when the IHS is removed, since the setup loses a few millimeters in thickness? Does anyone know if applying a better kind of thermal paste under the IHS and re-mounting with the IHS on provides decent results? I know it would be kind of a shame to go to all that trouble just to re-mount the IHS but it would be a lot safer and maybe easier if the heatsinks can't reach.

nucleon · 2013-10-20, 23:29

No, this is the problem I had.

Due to the mechanical properties of the socket, without the IHS, there are a few mm gap between silicon surface and the cooler.

If anyone knows a suitable cooler for a cpu sans IHS, I'm all ears.

Generally you'll get better results even with replacing the IHS. (from what I've read). In my mind, if you have a suitable cooler, I'm thinking silicon -> thermal paste -> cooler surface would be more efficient.

-- Craig

nucleon · 2013-10-20, 23:42

Aaah scratch that.

What I need to do, is remove the 'ILM', the locking mechanism.

Page20:
http://www.intel.eu/content/dam/www/...cket-guide.pdf

The IHS gives extra height to overcome the ILM.

Cool, something to try.

-- Craig

TheMawn · 2013-10-21, 00:18

I watched some video of a guy de-lidding. He didn't know what he was doing going into the de-lidding (admittedly) and ended up breaking his IMC so only one of his two DIMM slots worked.

He had to machine the standoffs for his cooler down a few mm as well as taking off the locking mechanism. Some heatsinks go on with an awful lot of pressure but I don't know that many / any of them can do all the way down to the die itself.

That coollaboratory ultra stuff looks to be completely bitchin'. I'm going to watercool my PC, probably as a Christmas project. I'm going to leave my GTX 660 Ti on air, though, since it's so cool and quiet and its only purpose is to TF. I'm seriously considering getting some of that Ultra and re-seating all the heatsinks I can find...

TheMawn · 2013-10-21, 05:12

Just to bring myself back up to speed, are the Haswell instabilities due to overheating in some regard or is the stress-test failing with incorrect residues or something? I like to keep up with the latest in computer hardware but this whole business is making me feel kind of iffy about Haswell in general.

You had mentioned something about new instructions as well. Is the old version without any instructions changes also unstable?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Haswell-E Prelim. Benchmark	sdbardwick	Hardware	37	2015-02-10 18:49
Prime95 and Haswell	Pleco	Information & Answers	22	2014-07-13 16:03
Haswell Rig	Mini-Geek	Hardware	64	2014-05-27 13:22
Prime95 version 27.1 early preview, not-even-close-to-beta release	Prime95	Software	126	2012-02-09 16:17
Missing mouse-over preview text	retina	Forum Feedback	1	2011-09-12 15:32

2013-10-20, 06:02	#364
nucleon Mar 2003 Melbourne 5·103 Posts	I confess - I didn't get first go. Yes I saw the vice method and thought no way. But as crazy as it sounds the vice method is the safest option. I tried the razor method, it's incredibly hard. I chipped the silicon. The vice method is cake. The only gotcha here - make sure you hold the block of wood firm against the vice and don't let it slip across the chip. I also did this too and knocked off one of the surface mount capacitors. All in all, instead of buying multiple chips and failing, I'd rather have bought 2x systems and underclocked. -- Craig

2013-10-20, 23:08	#370
TheMawn May 2013 East. Always East. 6BF₁₆ Posts	Is there any need to worry about the electrical conductivity of the thermal paste applied directly to the die? Does Arctic Silver 5 work or do I need a special ceramic based paste or something? Also, do most heatsinks or waterblocks fit when the IHS is removed, since the setup loses a few millimeters in thickness? Does anyone know if applying a better kind of thermal paste under the IHS and re-mounting with the IHS on provides decent results? I know it would be kind of a shame to go to all that trouble just to re-mount the IHS but it would be a lot safer and maybe easier if the heatsinks can't reach.

2013-10-20, 23:29	#371
nucleon Mar 2003 Melbourne 5×103 Posts	No, this is the problem I had. Due to the mechanical properties of the socket, without the IHS, there are a few mm gap between silicon surface and the cooler. If anyone knows a suitable cooler for a cpu sans IHS, I'm all ears. Generally you'll get better results even with replacing the IHS. (from what I've read). In my mind, if you have a suitable cooler, I'm thinking silicon -> thermal paste -> cooler surface would be more efficient. -- Craig

2013-10-20, 23:42	#372
nucleon Mar 2003 Melbourne 515₁₀ Posts	Aaah scratch that. What I need to do, is remove the 'ILM', the locking mechanism. Page20: http://www.intel.eu/content/dam/www/...cket-guide.pdf The IHS gives extra height to overcome the ILM. Cool, something to try. -- Craig

2013-10-21, 00:18	#373
TheMawn May 2013 East. Always East. 11·157 Posts	I watched some video of a guy de-lidding. He didn't know what he was doing going into the de-lidding (admittedly) and ended up breaking his IMC so only one of his two DIMM slots worked. He had to machine the standoffs for his cooler down a few mm as well as taking off the locking mechanism. Some heatsinks go on with an awful lot of pressure but I don't know that many / any of them can do all the way down to the die itself. That coollaboratory ultra stuff looks to be completely bitchin'. I'm going to watercool my PC, probably as a Christmas project. I'm going to leave my GTX 660 Ti on air, though, since it's so cool and quiet and its only purpose is to TF. I'm seriously considering getting some of that Ultra and re-seating all the heatsinks I can find...

2013-10-21, 05:12	#374
TheMawn May 2013 East. Always East. 11010111111₂ Posts	Just to bring myself back up to speed, are the Haswell instabilities due to overheating in some regard or is the stress-test failing with incorrect residues or something? I like to keep up with the latest in computer hardware but this whole business is making me feel kind of iffy about Haswell in general. You had mentioned something about new instructions as well. Is the old version without any instructions changes also unstable?