Why isn't Intel fixing the 'delidding' issue?

[2is]

No, enthusiasts won't.

 

[JimmiG]

You increased voltage and frequency, which means you did increase thermal output. You stated that measured core temp remained consistent.

Yeah, that's my point. In both cases (before delid at a low voltage and after delid at a much higher voltage), my temps are in the high 80's in stress tests, but the amount of heat dumped into the room differs significantly. So the temperature of the chip is unrelated to the amount of heat it dumps into the room, and by not delidding, you're putting a hard cap on how much heat the CPU can actually generate.

 

[TerryMathews]

Yeah, that's my point. In both cases (before delid at a low voltage and after delid at a much higher voltage), my temps are in the high 80's in stress tests, but the amount of heat dumped into the room differs significantly. So the temperature of the chip is unrelated to the amount of heat it dumps into the room, and by not delidding, you're putting a hard cap on how much heat the CPU can actually generate.

You're also putting a hard cap on performance, because if that heat can't be dissipated via heat sink it will remain in and around the core and once the processor reaches it's thermal limit it will throttle until it's under again.

Also, if you dropped your frequency and volts back to your pre-delidding level you would see core temp go down. I'm certain of it.

 

[alyarb]

Some 32, let alone 22nm users are hitting 5+ GHz on air cooling. As we approach the limits of silicon's usefulness, that is a large amount of frequency headroom we are taking for ourselves. The choice to move from solder to TIM is both a cost-cutting measure in the short term and it limits the value of the K-series to protect and protract the steady, however unsatisfyingly slow upwards march of base clocks.

I imagine below 14nm intel will be eager to sell us the 5 to 6 GHz space, not let the K-series from a previous node give it away for cheap. If you were a chip company and you owned 100% of your fabs, and you see people getting 2 GHz overclocks, I think you, too, will say "alright, enough is enough." Each node is more expensive than the last. I mean, you see the kind of heatsink they give you with the chip right? They do not care how high the K will go for you.

 

[jpiniero]

5 Ghz on 32 nm is pretty rare on air. 5 Ghz on 22 nm is extremely rare.

I imagine below 14nm intel will be eager to sell us the 5 to 6 GHz space

If anything, clock speeds could be going down.

 

[videogames101]

someone link me an example of a haswell chip throttling at stock

 

[mindbomb]

it is fixed on the 4820k (ivy-e) i heard, in that it uses a solder like sb-e chips.
I wonder why there isnt more hype around that thing.

 

[Pray To Jesus]

Money
&
Lack of competition

They might as well use that resource to make a competitive SoC line.

 

[bunnyfubbles]

it is fixed on the 4820k (ivy-e) i heard, in that it uses a solder like sb-e chips.
I wonder why there isnt more hype around that thing.

because the clockrates aren't much better (if at all) vs. SB-E, and it would need a really high clockrate to beat the 4770K, which also enjoys the benefits of Z87 (and thus there is basically 0 reason to go for a quad on X79 over Haswell)

Only way Ivy-E could have made a splash is if the chips could easily hit 5+GHz, or if they all were 6-8+core parts (including the 4820K)

 

[Meghan54]

it is fixed on the 4820k (ivy-e) i heard, in that it uses a solder like sb-e chips.
I wonder why there isnt more hype around that thing.

Maybe because of this:

The Extreme desktop parts come from the Xeon family. Sandy Bridge E was nothing more than a 6-core Sandy Bridge EP variant (Xeon E5), and Ivy Bridge E is the same.

Nothing new at all.

 

[tolis626]

For people living in countries with cold climate,delidding is the way to go.You delid your CPU,apply a big ass cooler and overclock the damn thing enough to heat the room you're in,to the point you don't have to use heating for that room.Profit!
In a more serious note,I think asking Intel to use soldering on all CPUs is kinda far-fetched.However,I can't see why they shouldn't do it for only their K-Series parts.I mean,these are parts targeted to overclockers,to that >0.1% of users (Regardless of the fact that average users buy them too because they think a K-series CPU is better than a non K one).And they charge a premium (like 40€) over the non-K parts for unlocking that multiplier.Yeah,it's bad for profits,but I see it as cheating us from their part.Practically,all they do is slap a K on the CPU name,unlock the multiplier,restrict the use of some virtualization stuff and voila!They make 40€ more!

 

[videogames101]

For people living in countries with cold climate,delidding is the way to go.You delid your CPU,apply a big ass cooler and overclock the damn thing enough to heat the room you're in,to the point you don't have to use heating for that room.Profit!
In a more serious note,I think asking Intel to use soldering on all CPUs is kinda far-fetched.However,I can't see why they shouldn't do it for only their K-Series parts.I mean,these are parts targeted to overclockers,to that >0.1% of users (Regardless of the fact that average users buy them too because they think a K-series CPU is better than a non K one).And they charge a premium (like 40€) over the non-K parts for unlocking that multiplier.Yeah,it's bad for profits,but I see it as cheating us from their part.Practically,all they do is slap a K on the CPU name,unlock the multiplier,restrict the use of some virtualization stuff and voila!They make 40€ more!

we're the ones choosing to pay for it

 

[mindbomb]

because the clockrates aren't much better (if at all) vs. SB-E, and it would need a really high clockrate to beat the 4770K, which also enjoys the benefits of Z87 (and thus there is basically 0 reason to go for a quad on X79 over Haswell)

Only way Ivy-E could have made a splash is if the chips could easily hit 5+GHz, or if they all were 6-8+core parts (including the 4820K)

4820k performs pretty much exactly like a 4770k (which wasn't even all that faster than a 3770k anyway) and doesnt have heat issues. I think the average person has 1-2 SSD's, so I don't think the 6 sata 6 ports of z87 is a big deal. Similiarly, I suspect people aren't really making use of their usb 3 ports. Meanwhile on x79, you get more ram support and quad channel mem, allowing you to make a sick ram disk.

The only real disappointing thing imo is the lack of avx2 support.

From my perspective, I'd rather have a 4820k.

 

[JimmiG]

4820k performs pretty much exactly like a 4770k (which wasn't even all that faster than a 3770k anyway) and doesnt have heat issues. I think the average person has 1-2 SSD's, so I don't think the 6 sata 6 ports of z87 is a big deal. Similiarly, I suspect people aren't really making use of their usb 3 ports. Meanwhile on x79, you get more ram support and quad channel mem, allowing you to make a sick ram disk.

The only real disappointing thing imo is the lack of avx2 support.

From my perspective, I'd rather have a 4820k.

http://www.anandtech.com/bench/product/836?vs=551

4770K is faster than the 3770K in many cases. More RAM lanes and a little more cache doesn't make much of a difference meaning the 4820K performs close to the 3770K.

Only reason to go with the 4820K over the 4770K and Z87 is if you're planning to go triple-SLI.

 

[cytg111]

someone link me an example of a haswell chip throttling at stock

mine is, have a thread on it some time back and there is others on this forum as well, just search for them.