Intel Broadwell Thread

[witeken]

14 nm at this point has to be very expensive, not to mention all the money spent on the specific process for Core M. The bigger problem is that there is no market for >400 non-Apple tablets, and even then the tablet market is being eclipsed by phablets.

According to Skaugen, customers of 2-in-1s upgrade 1 full year earlier, and more than 50% of them considered buying a tablet, but went with the 2-in-1 anyway and bought it because they wanted it instead of having to, and Intel has 10% market share of all 9 to 12 inch products, including iPad and Android devices. Every device that is sold that has Intel inside is a loss for Apple or another company. Intel can only win by addressing this market with Core M; they have an excellent value proposition. Core M is more than 30% smaller, so that helps, and I don't think the process is any expensive because it's just one of the many, many process flavors they have. 14nm doesn't perform so well in terms of cost because it has to catch up to a very, very high yield bar. If they don't match those yields -- 22nm has 2 year headstart in yield improvements -- the process will be less of an improvement than otherwise possible.

 

[Abwx]

14nm doesn't perform so well in terms of cost because it has to catch up to a very, very high yield bar.

At what height is this bar exactly, i mean how much are thoses yields, at 22nm for instance.?

 

[Idontcare]

At what height is this bar exactly, i mean how much are thoses yields, at 22nm for instance.?

Intel claims that 22nm is their highest yielding node ever in the history of the company.

For a logic node, going by historical numbers, I'd estimate that would mean 22nm is yielding at least 90% if not 95%.

This makes the bar for 14nm yields all the higher for all the obvious cost and margin reasons.

 

[Abwx]

Intel claims that 22nm is their highest yielding node ever in the history of the company.

For a logic node, going by historical numbers, I'd estimate that would mean 22nm is yielding at least 90% if not 95%.

This makes the bar for 14nm yields all the higher for all the obvious cost and margin reasons.

Although you re surely aware of some numbers since you re working in the industry i m skeptical of thoses 90-95% unless thoses take accounts of the harvested die parts, yields for 100% functional 100mm2 chips are about 65% at a 28nm mature node, i would expect a 22nm mature process to be at about 50%, and this seems correlated by Intel s 5.6M/year 300mm equivalent waffers output, if 2 millions of these 5.6M are needed for their CPU business then it means that yields can be even lower, not counting that given their capacity they can litteraly scrap parts that would be harvestable if ever a brutal rise in demand would let them capacity constrained.

 

[krumme]

Although you re surely aware of some numbers since you re working in the industry i m skeptical of thoses 90-95% unless thoses take accounts of the harvested die parts, yields for 100% functional 100mm2 chips are about 65% at a 28nm mature node, i would expect a 22nm mature process to be at about 50%, and this seems correlated by Intel s 5.6M/year 300mm equivalent waffers output, if 2 millions of these 5.6M are needed for their CPU business then it means that yields can be even lower, not counting that given their capacity they can litteraly scrap parts that would be harvestable if ever a brutal rise in demand would let them capacity constrained.

I agree yield numbers are meaningless without a context and i think its bs giving that signal without proper interpretation.

But it doesnt change the fact that intel can tailor and integrate eg design, production, segmentation, pricing whatever in a way and to a degree nobody else can. And it shows in the end result.

 

[Abwx]

I agree yield numbers are meaningless without a context and i think its bs giving that signal without proper interpretation.

But it doesnt change the fact that intel can tailor and integrate eg design, production, segmentation, pricing whatever in a way and to a degree nobody else can. And it shows in the end result.

Indeed, and i didnt say otherwise when pointing that they have the means to completely reject parts that could had been harvested, currently their cost must be about 7500$/waffer, adjusting for yields it should be twice this value, to compare with the packaged resulting price of 65000$ or so per eqivalent fully functional waffer...

 

[mikk]

http://www.notebookcheck.com/Test-HP-Enyy-x2-15-c000ng-Convertible.130835.0.html


HP Enyy x2 test.

 

[jpiniero]

http://www.notebookcheck.com/Test-HP-Enyy-x2-15-c000ng-Convertible.130835.0.html


HP Enyy x2 test.

It's tough to compare the battery life to the ARM tablets but you would think the difference would be more between the Envy and the U parts esp since it has a 50Whr battery. It sure looks like it is drawing far more than the typical ARM tablet does.

 

[frozentundra123456]

I dont know, I dont think the battery life looks that bad, but it certainly is not outstanding. In any case, the biggest problem I see is 800 euros and a 5400 RPM mechanical hard drive. Seriously?? And I also question the utility of a 15 inch convertible. When detached, the tablet is just too big.

 

[witeken]

I dont know, I dont think the battery life looks that bad, but it certainly is not outstanding. In any case, the biggest problem I see is 800 euros and a 5400 RPM mechanical hard drive. Seriously?? And I also question the utility of a 15 inch convertible. When detached, the tablet is just too big.

How Kirk looks at this is that, basically, if you want to buy a laptop, it should be the best laptop, and then you get a tablet for free. When you want to buy a tablet, it should be the best tablet in the world, and then you have a keyboard. Being able to use it as a tablet doesn't mean you lose the laptop function...

I also don't think a 15 inch tablet is too big. There are use cases where sizes doesn't matter.

 

[frozentundra123456]

How Kirk looks at this is that, basically, if you want to buy a laptop, it should be the best laptop, and then you get a tablet for free. When you want to buy a tablet, it should be the best tablet in the world, and then you have a keyboard. Being able to use it as a tablet doesn't mean you lose the laptop function...

I also don't think a 15 inch tablet is too big. There are use cases where sizes doesn't matter.

Well, there may be some cases where a 15 inch tablet is ok, but there are others, including any time you want to hold it in one hand, where a 15 inch tablet is definitely unwieldy. I mean, do you see any stand alone 15 inch tablets? I personally would prefer a nice ultrabook for laptop use and a cheap windows tablet for light browsing and content consumption. Or a cheaper solution like the Acer atom transformer, where at least you dont have to pay a premium price for a compromised solution.

 

[Abwx]

http://www.notebookcheck.com/Test-HP-Enyy-x2-15-c000ng-Convertible.130835.0.html
HP Enyy x2 test.

The apparently not that high power numbers wont hide that this is definitly the defeat of the big cores approach for mobile items, in term of perf/watt this thing is litteraly crushed by a simple Beema at the rate of about 50% at stock conditions for GPU + CPU, and a modest 100% when max throughputs are equalised, the estimation hold also for Bay trail on the CPU part but to a lesser extent, although substancial, for GPU + CPU.

 

[mikk]

The apparently not that high power numbers wont hide that this is definitly the defeat of the big cores approach for mobile items

What is your definition for mobile?

in term of perf/watt this thing is litteraly crushed by a simple Beema at the rate of about 50% at stock conditions for GPU + CPU, and a modest 100% when max throughputs are equalised, the estimation hold also for Bay trail on the CPU part but to a lesser extent, although substancial, for GPU + CPU.

Beema is 15W product. Having compared the gaming benchmarks with Beema, I can say that HP Enyy x2 does slightly better than a Beema top dog. Doesn't bode well, a 15W Broadwell should destroy this.

 

[Abwx]

What is your definition for mobile?

It s a large definition, in this case it look like a huge tablet also, not very mobile and hand helding friendly with about 2kg weight, neverless it s supposed to be mobile, no.?.

Beema is 15W product. Having compared the gaming benchmarks with Beema, I can say that HP Enyy x2 does slightly better than a Beema top dog. Doesn't bode well, a 15W Broadwell should destroy this.

http://www.notebookcheck.net/HP-15-g005ng-Notebook-Review.126027.0.html

You have a comparison here, i dont see the Envy as being better, the CB score, though not really representative, is 10% lower and the graphic scores are on par, the Envy CPU + GPU consume less but it s only an appearance, what is missing are the scores but we can estimate them as well as the resulting perf/watt.

The comsumption test use Furmark + Prime 95, the Core M run at 0.8 CPU and 0.45-0.5 GPU, the Beema above under the same conditions does 1.4 CPU and 0.8 GPU, knowing the absolute perfs/GHz of each SoC it s straightforward to estimate the throughputs at thoses frequencies and then weight the power comsumption accordingly.

Edit : You can see that if Beema is at 15W then the Core M is at 10W according to the 6W difference at the main, wich is about 4.8W once we account the losses in PSUs and VRMs, so much for the 15W Broadwell, doesnt bode well for their 14nm in its current form.

 

[Phynaz]

More made up numbers from abwx. Citations please.

 

[Sweepr]

More made up numbers from abwx. Citations please.

If he earned a dollar every time he talks about how great Mullins/Beema is in an Intel thread he would probably be rich by now.

 

[Abwx]

More made up numbers from abwx. Citations please.

In the stress power comsumption test the CPU throughputs in terms of Cinebench are 0.72 for the core M and 1.4 for the Beema 6410, the simultaneous GPU throughputs are 60% of the stock GPU frequency perf for core M and 100% for Beema since its GPU doesnt throttle.

 

[Abwx]

If he earned a dollar every time he talks about how great Mullins/Beema is in an Intel thread he would probably be rich by now.

Well, i could had cashed perhaps 100$, far from the rougly million you would had deserved for some BT thread...

Edit : What other brand Intel s CPUs should be compared with.?.

 

[krumme]

Obviously this is not fit for 4.5w tdp.

But still as the big cores are leaking the same my bet is a 15w tdp part will have more or less same battery life and give a substantial perf boost.

These products doesnt tell much about 14nm just that it slipped a bit as we knew. Leaks more in the order of hw. And is not fit for 4.5w tdp. The battery life is perhaps the sad thing but thats how it is with big cores.

 

[Sweepr]

Well, i could had cashed perhaps 100$, far from the rougly million you would had deserved for some BT thread...

Yeah, I know that thread bothers you and some of your friends from a certain forum a lot. While you are crying about contra-revenue and still using performance numbers from the old AMD Discovery Tablet preview cause Mullins is almost MIA in actual tablets there's tons of interesting BT designs wins out there. Keeping that thread updated is prettty hard, no wonder it has so many posts.

 

[Abwx]

Yeah, I know that thread bothers you and your friends a lot. While you are crying about contra-revenue and still using performance numbers from the old AMD Discovery Tablet preview cause Mullins is almost MIA in tablet there's tons of BT designs wins out there, keeping that thread updated is pretty hard.

Stop being paranoid, there s no thread that bother anyone here, i find it even amusing that ultimately you are promoting the good, although mediocre in absolute terms, horse, i m curious how you will do for this one, seems that what was expected as being some relief is now exposed as yet another annoyance, it s already not easy to be that enthusiastic with an outdated SoC, in some way you are either very courageous or blindly enthusiastic, or even both, who knows...

As for Mullins being MIA it s logical that, as good as it can be, at +40$ or even +30$ it cant compete againt a mediocre product whose price is -75$, we re not talking of shenanigans but of products perfs...

 

[mikk]

http://www.notebookcheck.net/HP-15-g005ng-Notebook-Review.126027.0.html

You have a comparison here, i dont see the Envy as being better, the CB score, though not really representative, is 10% lower and the graphic scores are on par, the Envy CPU + GPU consume less but it s only an appearance, what is missing are the scores but we can estimate them as well as the resulting perf/watt.

http://www.notebookcheck.com/AMD-Radeon-R5-Beema.122364.0.html


Despite the fact that Beema has a 3x larger TDP it isn't faster there, to be accurate slightly slower than the HP Enyy x2.

Edit : You can see that if Beema is at 15W then the Core M is at 10W according to the 6W difference at the main, wich is about 4.8W once we account the losses in PSUs and VRMs, so much for the 15W Broadwell, doesnt bode well for their 14nm in its current form.

Core M is a ~5W Soc, Beema is 15W.

http://www.notebookcheck.com/fileadmin/Notebooks/Sonstiges/Prozessoren/Broadwell/dota2.png

 

[Abwx]

http://www.notebookcheck.com/AMD-Radeon-R5-Beema.122364.0.html
Despite the fact that Beema has a 3x larger TDP it isn't faster there, to be accurate slightly slower than the HP Enyy x2.
Core M is a ~5W Soc, Beema is 15W.

http://www.notebookcheck.com/fileadmin/Notebooks/Sonstiges/Prozessoren/Broadwell/dota2.png

On the two reviews the numbers are :

Core M idle 4.2W load 19.1W delta 14.9W

Beema idle 5W load 25.8W delta 20.8W


So with a 14.9W delta said Envy core M is 5W but it takes only 6W more for Beema to be a 15W part.?.

6W more is 40%, if you think that this core M is at 5W then the Beema is at 7W...

Edit : and Beema is twice as fast in this test, it run at 1.4 with 4 cores and 800MHz GPU while the core M is at 800MHz CPU and 450-500MHz GPU, do the maths, Beema would have full score on GPU and 1.4 on CB while the Core M would yield 60% of its max GPU score and 0.72 in CB.

 

[Enigmoid]

On the two reviews the numbers are :

Core M idle 4.2W load 19.1W delta 14.9W

Beema idle 5W load 25.8W delta 20.8W


So with a 14.9W delta said Envy core M is 5W but it takes only 6W more for Beema to be a 15W part.?.

6W more is 40%, if you think that this core M is at 5W then the Beema is at 7W...

Edit : and Beema is twice as fast in this test, it run at 1.4 with 4 cores and 800MHz GPU while the core M is at 800MHz CPU and 450-500MHz GPU, do the maths, Beema would have full score on GPU and 1.4 on CB while the Core M would yield 60% of its max GPU score and 0.72 in CB.

Because at this level the energy used by non performance parts is most definitely non negligible.

Idle and load also differ in screen brightness and wifi. Subtracting off about 5W would be a good start to an apples to apples comparison.

1080p touchscreen (core M) vs 768p (Beema) screens as well.

I have no idea how your math works the bolded above is absolutely asinine and would fail the logic test of a 12 year old.

IE)

Your 5 kW stove uses twice the energy of a 10 kW stove. However when combined with the energy usage of the house (100 kW) its only 4.7% more therefore the rating of the 10kW stove is wrong, its actually a 5.23 kW stove. And since its producing twice the amount of heat, its twice as efficient.

This type of logic is a disservice to other members on this forum.

 

[krumme]

Clearly there is hefty throtling going on for some usecases and therefore comparing this product to more normal cpu like eg beema and core ulv 15w parts is not straight forward.

You dont get the performance and low tdp at the same time with heavy work cases. And this test dont even dont show and meassure the throtling problems besides the games - no wonder. Its bad marketing and notebookcheck is a business too.

We have enough test now to see, what is not the sligtest surprising - that 4.5w tdp is way out of the effective area for this arch. What is so damn difficult about saying that? It goes all the way to perhaps 45w (?) so its actually quite a stretch and impressive. Why does everything have to be interpreted in a black and white, very good, very bad way.

This is a niche cpu. And the product we have seen so far is very niche like. But fine. For those that have the $ and do office work and want a convertible its a super opportunity to have because you get fast browsing and navigating in windows and office.

- and yeaa the mechanical hd is a bummer that destroys performance and makes it useless for this class of product but that is a problem seen before and it can probably be solved.