Intel Demonstrates 65W Broadwell-K Socketed Processors at GDC 2015

[Idontcare]

I also feel that there is too much uncertainty this late after 14 nm started going into production. With the rumored Skylake delay, I keep wondering if it's for marketing purposes, of if a the high performance quad's required a respin to meet yield and throughput requirements (or some other reason).

I suppose the real bottom line is that I don't like the new Intel model where the enthusiast crowd gets much less information in advance than b/4.
Change isn't always for the better - at least for our purposes. Instead of discussing the pros and cons of upcoming features or architectures - we end up in circular debates about speculative information from questionable sources.

That presumes the decisions makers at Intel know what the market needs or wants 90 days from now (standard cycle time for the fab)...I question whether we should be so confident in their confidence right now.

 

[Ajay]

That presumes the decisions makers at Intel know what the market needs or wants 90 days from now (standard cycle time for the fab)...I question whether we should be so confident in their confidence right now.

Yes, goog observation, the change in Intel's stance may well be based on a 'real' change to their business model. This is still distruptive to our habits regarding discussions of CPUs, but at least one can respect Intel's descision.

 

[Fjodor2001]

That presumes the decisions makers at Intel know what the market needs or wants 90 days from now (standard cycle time for the fab)...I question whether we should be so confident in their confidence right now.

Are you suggesting Intel actually would completely turn around their CPU plans within a 90 day time frame? I.e. prioritize mobile at one point, and then 90 days later do the opposite and prioritize desktop? Or decide to provide SKUs with huge iGPU and eDRAM at one point, and then 90 days later decide to skip it?

And why would Intel be less confident in what the markets needs now than they were a few years ago?

 

[DrMrLordX]

Not sure where you got the whole "provide SKUs with huge iGPU and eDRAM at one point, and then 90 days later decide to skip it" thing, since Intel has been talking about unlocked Broadwell Iris Pro in slides for months now, and they show no sign of skipping it. Assuming that's the product to which you are making reference. Yeah I know you're referencing Idontcare's post but I don't think that was the point he was trying to make. If anything, Intel has been more consistent in telling us about that one SKU than anything else related to Broadwell or Skylake.

 

[coercitiv]

I would rather see a mainstream quad mobile chip with iris pro instead of the desktop.

Please don't let common sense go to your head.

 

[Fjodor2001]

Not sure where you got the whole "provide SKUs with huge iGPU and eDRAM at one point, and then 90 days later decide to skip it" thing, since Intel has been talking about unlocked Broadwell Iris Pro in slides for months now, and they show no sign of skipping it. Assuming that's the product to which you are making reference. Yeah I know you're referencing Idontcare's post but I don't think that was the point he was trying to make. If anything, Intel has been more consistent in telling us about that one SKU than anything else related to Broadwell or Skylake.

It was not I who made the claim - I was merely asking a question. I just gave some examples of possible changes, asking if that was what's intended, since I didn't know myself what was intended.

But what other types of drastic changes do you see happening in 90 day time frames that would cause Intel to not be able to provide any technical info prior to release like they used to some years ago?

 

[DrMrLordX]

Well *I* don't see anything like that per se, but that's kind of the problem: if Intel's hiding it, then we don't see it. My guess is Idontcare is really saying that Intel may not be 100% sure of what it is the consumer market really wants right now. Tablet sales are actually flagging, PC sales are about level . . . the future's a bit hard to read. And so Intel is maybe kinda sorta unsure about which SKUs to launch when and where and to what platform, which is why they are not leaking a whole bunch of early tech specs. It doesn't (necessarily) mean they're having overly-many problems producing Skylake (or what have you). It's that their hyper-segmented SKU strategy is designed to hit the market in just the right places with just the right features to maximize profit, and they're struggling to make sense of exactly how to populate their next-gen lineup of products due to market uncertainty. Or something. Maybe Idontcare can present the point better than I. Regardless, Broadwell Iris Pro is one product about which they seem to be very certain. That's been in the pipe for awhile, and now it's been demoed.

Personally, I'm still wondering if Intel has two slightly different 14nm processes going, the earlier one being used for Broadwell U/Y and Xeon-D and the more-recent one being used for Skylake.

 

[coercitiv]

Personally, I'm still wondering if Intel has two slightly different 14nm processes going, the earlier one being used for Broadwell U/Y and Xeon-D and the more-recent one being used for Skylake.

They better have, because it seems to me whatever advantage 14nm Broadwell U holds over 22nm Haswell U in terms of power consumption tends to evaporate going towards 3Ghz.

 

[witeken]

My guess is Idontcare is really saying that Intel may not be 100% sure of what it is the consumer market really wants right now.

Intel interviews 200,000 people every year.

 

[Enigmoid]

They better have, because it seems to me whatever advantage 14nm Broadwell U holds over 22nm Haswell U in terms of power consumption tends to evaporate going towards 3Ghz.

And where are you getting this?

(No, the fact that high-frequency chips haven't been launched yet does not count)

 

[coercitiv]

And where are you getting this?

(No, the fact that high-frequency chips haven't been launched yet does not count)

This has been discussed in the Broadwell thread, you were a part of the conversation actually: notebooks with BW U, while showing better sustained speeds under load and better idle power consumption, fail to consistently show significant improvements in load power consumption and overall battery life.

The i5 models tend to get better results, with some notable impressive examples, while the i7 have little to show for themselves. We should be seeing improvements across the board.

 

[dahorns]

This has been discussed in the Broadwell thread, you were a part of the conversation actually: notebooks with BW U, while showing better sustained speeds under load and better idle power consumption, fail to consistently show significant improvements in load power consumption and overall battery life.

The i5 models tend to get better results, with some notable impressive examples, while the i7 have little to show for themselves. We should be seeing improvements across the board.

which is mainly an indication that SoC power consumption is a relatively small part of overall device power consumption. The tests I've seen show better idle and better heavy load battery life. Battery life under light load is usually about the same.

 

[Enigmoid]

This has been discussed in the Broadwell thread, you were a part of the conversation actually: notebooks with BW U, while showing better sustained speeds under load and better idle power consumption, fail to consistently show significant improvements in load power consumption and overall battery life.

The i5 models tend to get better results, with some notable impressive examples, while the i7 have little to show for themselves. We should be seeing improvements across the board.

What? Max power usage? That's capped at 15W (or something a little higher depending on the OEM) and under heavy load that is what the SOC will use (hence no improvement over HW). Haswell throttled to 15W. Broadwell maintains turbo (higher clocks) and still hits the same 15W ceiling (unless configured otherwise). They both use 15W but one is more efficient that the other.

Battery life is hugely an effort on the part of the OEM to optimize the system.

The difference between the i5 and i7 models in terms of clockspeeds is insignificant (couple hundred mhz).

Your reading a lot into data with a tremendous amount of variability. Why you may reason (and you would be right) there is no guarantee that this is actually what will happen.

http://www.anandtech.com/show/8941/gigabyte-gbbxi7h5500-broadwell-brix-review/7

The slightly higher base clocks in the Core i7-5500U (compared to the Core i7-4500U) are probably the reason for the Haswell-based unit appearing more power efficient than the Broadwell counterpart - but, make no mistake here - the Broadwell unit wins the performance per watt test quite easily.

 

[coercitiv]

which is mainly an indication that SoC power consumption is a relatively small part of overall device power consumption. The tests I've seen show better idle and better heavy load battery life. Battery life under light load is usually about the same.

What BW i7 notebook have you seen with better heavy load battery life than it's HW i7 predecessor?

 

[coercitiv]

What? Max power usage? That's capped at 15W (or something a little higher depending on the OEM) and under heavy load that is what the SOC will use (hence no improvement over HW). Haswell throttled to 15W. Broadwell maintains turbo (higher clocks) and still hits the same 15W ceiling (unless configured otherwise). They both use 15W but one is more efficient that the other.

Max power usage is not the same as TDP, you know that. HW uses up to 25W in standard config for limited periods of time. I haven't had a BW to play with until now, but my impression is max power has been increased slightly... maybe 30W? (again, for the ~30s limited window of time)

If Broadwell is more efficient than Haswell in both idle and moderate heavy usage scenarios, why are we not seeing healthy battery life increase across the board like we saw with Haswell vs Ivy Bridge?

 

[ehume]

. . . if Intel's hiding it, then we don't see it. . . .

Didn't BK himself say that Intel would not be as forthcoming about upcoming products as they were in the past? I think it was in the context of his saying how Intel had overdisclosed info on their upcoming products.

 

[Enigmoid]

Max power usage is not the same as TDP, you know that. HW uses up to 25W in standard config for limited periods of time. I haven't had a BW to play with until now, but my impression is max power has been increased slightly... maybe 30W? (again, for the ~30s limited window of time)

If Broadwell is more efficient than Haswell in both idle and moderate heavy usage scenarios, why are we not seeing healthy battery life increase across the board like we saw with Haswell vs Ivy Bridge?

Because the big gains have already been made. Ivy -> haswell integrated the PCH on package, probably dropping system power by something like 1-2W. Ivy -> haswell on the desktop increased power usage remember. Different implementations. For mobile, by far the actual implementation is going to be the thing that makes a difference.

The max boost and power is really up to the OEMs to implement. Some OEM's want a strict 15W, others allow max turbo (and hence power usage). Intel has some control here but the vast majority of this sort of thing is implemented by the OEM.

As far as why you aren't seeing anything? The power usage is already so low that the screen is taking up the vast majority of the power. Apple gets a 1 hr increase in battery life. Intel claims 30% less power. That means 30% at most more battery life. If the CPU is only about 1/3 of the total power consumption then you will only see a battery life increase of 10% which is going to somewhat fall between the cracks with the variability of mobile platforms.

 

[frozentundra123456]

Max power usage is not the same as TDP, you know that. HW uses up to 25W in standard config for limited periods of time. I haven't had a BW to play with until now, but my impression is max power has been increased slightly... maybe 30W? (again, for the ~30s limited window of time)

If Broadwell is more efficient than Haswell in both idle and moderate heavy usage scenarios, why are we not seeing healthy battery life increase across the board like we saw with Haswell vs Ivy Bridge?

I am not convinced we have enough data yet to make a final judgement on BW battery life and power usage, but I think part of the problem could be that as cpu power usage gets less, it is a lower part of the total power consumption, so improvements have less and less effect on overall power consumption and battery life.

 

[cytg111]

Intel interviews 200,000 people every year.

And is 100.000 of those witeken wearing different shades, mustaches and wigs?

 

[Idontcare]

Didn't BK himself say that Intel would not be as forthcoming about upcoming products as they were in the past? I think it was in the context of his saying how Intel had overdisclosed info on their upcoming products.

If you, as a CEO, say that and then proceed to execute to a roadmap schedule like clockwork then the audience can justifiably assume your reasons for clamming up were legitimate.

If, however, as a CEO you say that on the heels of having no choice but to rollout a "100MHz refresh" instead of the expected "tick-tock" schedule of new products, combined with a healthy 6+ month delay in finally seeding the channel with desktop 14nm parts, then the audience is justified in assuming you are just using the "overdisclosed" excuse as smokescreen to cover for the fact your company is executing poorly and with no confidence in when it will resume executing with any certainty.

 

[Dave2150]

If you, as a CEO, say that and then proceed to execute to a roadmap schedule like clockwork then the audience can justifiably assume your reasons for clamming up were legitimate.

If, however, as a CEO you say that on the heels of having no choice but to rollout a "100MHz refresh" instead of the expected "tick-tock" schedule of new products, combined with a healthy 6+ month delay in finally seeding the channel with desktop 14nm parts, then the audience is justified in assuming you are just using the "overdisclosed" excuse as smokescreen to cover for the fact your company is executing poorly and with no confidence in when it will resume executing with any certainty.

Devil's Canyon was a 500Mhz clock speed increase, both a stock and at turbo clocks. 5x more than the 100Mhz you mention.

We've had Haswell-E also to satisfy the enthusiasts, and news of the upcoming 65W Broadwell-K with Iris Pro 128MB l4 cache is now here.

Though I am highly curious to see how much of a performance improvement it will be over the 4790K.

 

[DrMrLordX]

I think he's referring to the maximum average overclocks on Haswell Refresh vs the older 4770k. Yeah the stock clocks on the 4790k are nice, but it has almost no headroom.

Haswell-E is just moar cores. You'll be lucky to get past 4.5 ghz with either the hex or octacore chips.

Idontcare has a point, though it's hard to say exactly what, if anything, is implied by this phenomenon. Back in '05 we had Intel leaks of Conroe ES chips around 7-9 months before their historical '06 release. Now go over to XS and see what Fugger has posted most recently: a 6 ghz run of an i7-5960X ES under triple-stage phase. That's awesome, but it's also an ES of a chip that's already on the market.

Intel had a lot to prove back then, so the leaks were a masterful marketing ploy to get people ready for the Conroe roll-out in June/July of '06. This time around, Intel doesn't have so much to prove, though in a way they do: the Broadwell delay has some people a bit nervous. It would be great if there were at least a few Skylake ES chips out there, subject to enthusiast-grade testing. We're, what, maybe 3 months out from the projected Broadwell Iris Pro launch and ??? months out from Skylake, and so far all we've got is a canned Intel demo of the Iris Pro chip and . . . I don't know what for Skylake.

 

[Phynaz]

If you, as a CEO, say that and then proceed to execute to a roadmap schedule like clockwork then the audience can justifiably assume your reasons for clamming up were legitimate.

If, however, as a CEO you say that on the heels of having no choice but to rollout a "100MHz refresh" instead of the expected "tick-tock" schedule of new products, combined with a healthy 6+ month delay in finally seeding the channel with desktop 14nm parts, then the audience is justified in assuming you are just using the "overdisclosed" excuse as smokescreen to cover for the fact your company is executing poorly and with no confidence in when it will resume executing with any certainty.

Otellini sure knew when to walk away.

 

[erunion]

And is 100.000 of those witeken wearing different shades, mustaches and wigs?

I wonder what kinds of products he's secretly urging Intel to make.

 

[Ajay]

Otellini sure knew when to walk away.

Yes, he sure did. 'Interesting' times ahead for Intel.