Another member of the Haswell club

[Kenmitch]

- Oh well, guess it is incentive to go for a better and perhaps more silent cooling solution ala megahelm.

I opted for the H100i as I dont feel comfortable hanging 2lbs off my motherboard.

 

[VirtualLarry]

Prime95 is not a power-virus program though. Prime95 is an actual program that people use to find Mersenne primes.

Intel should not have released haswell if the iVR algo's are this poorly tuned, or if haswell itself has a critical speedpath in its AVX circuits that necessitates such an elevated voltage in order to reliably operate at the target clockspeeds.

To me this reeks of an immature product, which either means a new stepping is in the offing that addresses the gating speedpath in the AVX circuits (and folks are advised to hold out for it before upgrading to Haswell) or it means Intel really phoned it in on Haswell's iVR (ala Silverthorne style) which would justifiably lead to all sorts of questions regarding the legitimacy of their TDP spec (no different than the scrutiny leveled against AMD and the FX-8350's TDP).

Didn't Intel's new CEO say (effectively), that they were going to move to a "release early, release often" schedule, rather than "make it perfect, before release" schedule?

Perhaps that is the meaning of "Haswell refresh" on the roadmaps? A fixed Haswell design?

 

[blackened23]

The context of that statement I believe is in regards to the mobile market. Intel basically cannot survive unless they release products rapidly, and using their latest process nodes. Consider the original atom has been around how long? Intel cannot do that and thrive in the mobile market, so they're going to release yearly or twice-yearly products for their low power line - this is beginning with Bay Trail.

Also, I could be wrong but from what I can tell, the Haswell refresh is actually "Haswell-E". I believe intel is moving most of its LGA CPUs over to the enthusiast platform due to the direction of the computing market (ie mobile).

 

[coffeejunkee]

- Oh well, guess it is incentive to go for a better and perhaps more silent cooling solution ala megahelm.

A better cooler never hurts, but Shintai is just wrong about mct causing the throttling. It only happens when you run avx applications, in which case the voltage gets an automatic 0.1V bump (for reasons known only to intel engineers).

Run something like handbrake to get a more realistic idea of temps under load. There aren't any useful avx programs anyway (yet).

 

[ShintaiDK]

A better cooler never hurts, but Shintai is just wrong about mct causing the throttling. It only happens when you run avx applications, in which case the voltage gets an automatic 0.1V bump (for reasons known only to intel engineers).

Run something like handbrake to get a more realistic idea of temps under load. There aren't any useful avx programs anyway (yet).

So why did IB throttle with it on then? And running at 80-85C with it off?

 

[coffeejunkee]

Can you link to some kind of source that shows it does? Until then, I rather go by my own experience (and common sense) that shows a 1 degree increase vs stock for mct enabled and a massive temp increase when one increases vcore with a substantial amount like 0,1V. Also, if those lousy few extra turbo bins cause so much heat, why do so many mobo vendors enable it by default? (probably shouldn't be asking rhetorical questions in your case)

 

[ShintaiDK]

Can you link to some kind of source that shows it does? Until then, I rather go by my own experience (and common sense) that shows a 1 degree increase vs stock for mct enabled and a massive temp increase when one increases vcore with a substantial amount like 0,1V. Also, if those lousy few extra turbo bins cause so much heat, why do so many mobo vendors enable it by default? (probably shouldn't be asking rhetorical questions in your case)

There is an example here:
http://forums.anandtech.com/showthread.php?t=2260263&page=5

Why? Because it will only throttle when you use a load, that people normally wouldnt experience. So mobo makers enable it as default to give an edge in for example games.

A debate about it here as well:
http://www.anandtech.com/show/6214/multicore-enhancement-the-debate-about-free-mhz

 

[cytg111]

I opted for the H100i as I dont feel comfortable hanging 2lbs off my motherboard.

And googling into it, that might just be the road ahead for me, thx.

edit : And if MCT is running out of spec, -as to say I get throttled under certain circumstances, that I would have liked to have known before hand.. but its nothing that sets me off, when this rig retires to server duty, MCT if off.

 

[coffeejunkee]

There is an example here:
http://forums.anandtech.com/showthread.php?t=2260263&page=5

Why? Because it will only throttle when you use a load, that people normally wouldnt experience. So mobo makers enable it as default to give an edge in for example games.

A debate about it here as well:
http://www.anandtech.com/show/6214/multicore-enhancement-the-debate-about-free-mhz

Well ok, mct might contribute but the testing in your first link is done at a very high ambient temp of almost 35 degrees. Even then, no throttling happens as it hits 'only' 103 degrees. So I still maintain the 0.1V avx bump is a much bigger factor. Also notice his vcore changes with mct enabled, from 1.016 without to 1.080 (actually I thought vcore stayed the same, but I didn't test very thoroughly).

Best way to test would be for op to change his vcore to fixed and set it to the exact same value as auto would give (without the bump). Then run avx Linx with mct enabled and see if throttling occurs.

Btw, H100i is complete overkill if you're not overclocking. There are many aircoolers that don't weigh a ton and still give excellent cooling at much lower noise levels. I recommend Thermalright True Spirit 140 or Macho or Noctua NH-U12S.