Ivy Bridge update may disappoint...

[bill-p]

Image

The intel i7 2600k is the high performance Sandy Bridge CPU which is slower than then i7 3770K ( Ivy Bridge ). Yet the Ivy Bridge CPU has less Power drain.

TDP of 3770K and 2600K are different. 3770K is binned at 77W compared to 2600K, which is 95W.

So comparing TDP alone, there should be that difference. But it's worth noting, and I repeat, that mobile Ivy Bridge doesn't have lower TDP compared to mobile Sandy Bridge. In that case, would this translate over? I don't think so.

Also to note, Anand has some funny stuffs going on in his previews. The 2500K in particular, sometimes outperform or match the 2600K. In that case, I think you can safely compare the 3770K vs the 2500K in some of those workloads. And also to note, compared to 2500K, the power saving of 3770K isn't that large.

 

[Quash]

The probable reason Ivy Bridge on mobile platforms has the same TDP is the bigger GPU. If you are not using the gpu intensively the power consumption of just the cpu will probably be significant lower.

On the desktop side the gpu is a less significant part of the total power consumption so they can lower the tdp. On mobile it's a bigger part of the the tdp so they can't. It will probably still mean that an ivy bridge macbook pro will run significantly cooler than a sandy bridge one.

 

[dusk007]

The probable reason Ivy Bridge on mobile platforms has the same TDP is the bigger GPU. If you are not using the gpu intensively the power consumption of just the cpu will probably be significant lower.

There is the GPU Turbo. If the GPU is in heavy use clocked to max speed the CPU has to sit back a little and vice versa. The TDP is not GPU TDP+ CPU TDP. It is max TDP and GPU TDP + CPU TDP > max TDP.
The effekt might be a lot smaller than you'd think.

On the desktop side the gpu is a less significant part of the total power consumption so they can lower the tdp. On mobile it's a bigger part of the the tdp so they can't. It will probably still mean that an ivy bridge macbook pro will run significantly cooler than a sandy bridge one.

Could be true but it clashes with Intel's claims of a 100% power efficiency improvement of its 22nm Intel HD 4000 EUs. If they are twice as fast at the same power and some 50-60% faster that should mean the GPU 22nm GPU sucks overall 25% less power than the HD 3000.
Ergo if the GPU sucks 25% less and the CPU 30% (up to 40% is 22nm claim) we should hit 35W at about the same clock rate.

I agree though it probably just doesn't show because the numbers are too close and new process nodes there are too many chips that are hotter than they should be. Intel would either have to bin more aggressively or use them all and sell them in a higher than real TDP class. The can afford the latter with no real competition, they'd be fools not to.

 

[Fleet would mac]

What TDP measures is how much heat the chip puts out on average, or how much heat Intel "thinks" it puts out on average.

Are you sure about that, Bill?

My understanding is that TDP is closer to being the maximum amount of thermal power the chip is designed to output**. How much thermal power the chip outputs on average (whatever that means ) is another matter.

(** In practice, a TDP of 35W for the non-ULV Ivy Bridge CPUs for mobile systems simply tells system designers that they need to supply a cooling system capable of dissipating 35W of heat. Hopefully, the fact that this is unchanged from Sandy Bridge won't encourage Apple to stick with the same relatively noisy fans that the current pro range is using.)

It's also going to be interesting to see whether Apple is going to make use of the configurable TDP (cTDP). For example, it could automatically switch to a lower TDP ("cTDP down") when running off the batteries.

And also to note, compared to 2500K, the power saving of 3770K isn't that large.

Yes, in Anand's preview, the power consumption of the IB i7-3770K under load (x264 HD Pass 2) was only slightly lower than that of the SB Core i5-2500K (128.3W vs 131.6W). However, don't forget that this test utilizes all cores and the i5-2500K does not support Hyper-Threading. So the i5-2500K is only running 4 threads concurrently, compared to the i7-3770's 8, which probably largely accounts for why the i7-3770 was nearly 43% faster than the i5-2500K in this benchmark.

 

[Freyqq]

speed-wise, perhaps. However, the biggest improvement will be in battery life, intel graphics, and standard USB 3.0

 

[bill-p]

Are you sure about that, Bill?

My understanding is that TDP is closer to being the maximum amount of thermal power the chip is designed to output**. How much thermal power the chip outputs on average (whatever that means ) is another matter.

(** In practice, a TDP of 35W for the non-ULV Ivy Bridge CPUs for mobile systems simply tells system designers that they need to supply a cooling system capable of dissipating 35W of heat. Hopefully, the fact that this is unchanged from Sandy Bridge won't encourage Apple to stick with the same relatively noisy fans that the current pro range is using.)

Yes, I am 100% sure that TDP does not represent maximum thermal dissipation. At least it doesn't with Sandy Bridge, and it most likely won't apply to Ivy.

Here's an older thread on MacRumors to deal with this whole TDP thing:
https://forums.macrumors.com/threads/1183742/

The problem here is that Intel doesn't take Turbo Boost into account when they post these TDP numbers, so when you factor in Turbo, TDP changes quite drastically.

Also, like I kept saying, the 3770K in that example had a lower TDP compared to the 2500K and 2600K. I'd say that's definitely indicative of lower power consumption and heat, which is good for desktop parts, but it's not the same case in mobile because TDP remains unchanged for mobile Ivy Bridge.

Furthermore, the x264 situation was one of the very few that showed a dramatic difference between Ivy and Sandy, and not all of us encode videos on a daily basis.

 

[dusk007]

Yes, I am 100% sure that TDP does not represent maximum thermal dissipation. At least it doesn't with Sandy Bridge, and it most likely won't apply to Ivy.

You just make it sound more complicated than it is.
TDP is the maximum thermal dissipation that the cooling system needs to be able to deal with. Sandy Bridge can exceed the TDP for short durations but that doesn't matter because it may not exceed it long enough to have an adverse effects.
Intel does take Turbo boost into account but it is designed so it has no effect on the maximum TDP. And TDP doesn't change because of Turbo. The TDP is a fixed value of sorts. That the chip can for short durations spike in power and heat changes nothing.
You may say that is just tautology but for all practical purposes it really doesn't make any differences.

That the chip can over clock just means it heats up quicker not that it actually gets hotter at the end of the day.

Some people think the chip under full load just consumes 45W with turbo speeding it up to the limit but a slower chip will consume less even if it is the same TDP class.
With a none mature process binning can negate some of the effect but with a mature process most chips are about equally cool at the same clock speed. They end up differently fast only because of factory settings.
A 2675 QM has 6 turbo bins for 4 cores. Thus it can clock at a max 2.8Ghz when all for cores are active. A 2860QM has 8 bins and starts at 2.5Ghz which means 3.3Ghz max. If the target was to run either chip at exactly 45W they'd probably both end up running a 10 min encode job at 2.6Ghz and be equally fast or some clock rate that would be really close.

TDP of 45W even with Sandy Bridge is no average power consumption and the internal Turbo power target is probably not simply 45W for any chip that is that TDP class.

 

[thundersteele]

TDP is the maximum thermal dissipation that the cooling system needs to be able to deal with. Sandy Bridge can exceed the TDP for short durations but that doesn't matter because it may not exceed it long enough to have an adverse effects.
[...]

TDP of 45W even with Sandy Bridge is no average power consumption and the internal Turbo power target is probably not simply 45W for any chip that is that TDP class.

I agree with the above. The TDP specifies much cooling the system has to provide to operate the CPU under load for extended time periods. This is of course correlated with the average power consumption under (full) load, since the thermal energy has to come from somewhere.
The maximal power consumption for short periods can be higher, that's how the turbo boost works, while it will be much lower e.g. when the laptop is idle.

Battery life has to do mostly with idle power consumption and with how the CPU reacts to small loads or to quick spikes (e.g. loading a website). This has very little to do with the TDP.

 

[mark28]

TDP of 3770K and 2600K are different. 3770K is binned at 77W compared to 2600K, which is 95W.

So comparing TDP alone, there should be that difference. But it's worth noting, and I repeat, that mobile Ivy Bridge doesn't have lower TDP compared to mobile Sandy Bridge. In that case, would this translate over? I don't think so.

Also to note, Anand has some funny stuffs going on in his previews. The 2500K in particular, sometimes outperform or match the 2600K. In that case, I think you can safely compare the 3770K vs the 2500K in some of those workloads. And also to note, compared to 2500K, the power saving of 3770K isn't that large.

TDP = recommended energy / heat that needs to be dealt with for the computer builder. It's not a measurement of the Power Drainage.

The 3770K is faster than the 2600K, so it's fair to compare the Power Drain to the 2600K, despite the 3770K have a lower TDP. If an Ivy Bridge of the same TDP is used for mobile CPU's, then the outperformance would be greater obviously, since a low TDP Ivy Bridge is already beating the high performance Sandy Bridge model.

Ivy Bridge = Faster while offering significant lower Power Drainage.