Surface Book GPU Confirmed 10/20/15 UPDATE

[manikyath]

well i was thinking more "out in the field work" if i have a desktop close, id rather sit in a comfortable chair with a good mouse and keyboard and work off that anyway

i'm *that* guy that spent networking class trying to get steam "in-home" streaming to work over a geographical distance of 25km.

and as long as your connection isnt as unstabile as a reliant robin its surprisingly manageable.

[LukaP]

i'm *that* guy that spent networking class trying to get steam "in-home" streaming to work over a geographical distance of 25km.

and as long as your connection isnt as unstabile as a reliant robin its surprisingly manageable.

yeah of course one could setup a VPN to emulate a LAN over WAN, but thats just cheating

[Sauron]

i wonder how much faster the 940 is over an IRIS GPU with DDR4

 

not much. but it doesn't really matter since neither is really suited for heavy loads. someone pointed out it can be useful for cuda development, which is true, but I still think it's a pretty niche use case.

[randomhkkid]

The fuck kind of drawings are you editing that you cant just hide most components to make it workable?

 

and twice the slices iirc

 

Xeon* 

Yes and no. Inventor is only certified to run on Xeons and Quadros. That doesnt mean it wont run on your run of the mill i5, or that the Xeon version of that i5 will run it any better, just that its been tested to work on that, and not tested on consumer products.

 

A xeon is not inherently faster at CAD than a Core i. It all depends on the amount of cores and their speed.

 

Engine visualisations or full automotive models is where I see slowdown, a dedicated GPU usually means I can do a full vis, also helps when modelling the material properties.

 

All the material I can find on Skylake-U points to it maxing out at GT3e with 48EUs.

[LukaP]

Engine visualisations or full automotive models is where I see slowdown, a dedicated GPU usually means I can do a full vis, also helps when modelling the material properties.

 

All the material I can find on Skylake-U points to it maxing out at GT3e with 48EUs.

well thats not exactly the kind of work this is suited to. this is meant for on the field, explaining stuff to engineers and workers, or quickly busting together a rough part that needs to be made from measurements you take there. not to render entire engines or stuff

 

Example from your engine: You work at VW, and do CAD for engines or something. The engineer calls you down to the machineshop to discuss what needs to be done to reduce the emissions of the diesel engines. you open up the 1.6 TDI on your surface book, hide parts that arent relevant, and then do changes to those that are, together with the engineer. 

 

Oh yeah, Sk-H is 72EU... Well then no luck

[ZetZet]

not much. but it doesn't really matter since neither is really suited for heavy loads. someone pointed out it can be useful for cuda development, which is true, but I still think it's a pretty niche use case.

problem with iris pro in this form factor would the that it's on the cpu, which would mean it would be on the tablet part, which would mean insane amount of heat in one tiny place, by doing it this way they split the heat.

[Commander Llama]

Seems reasonably powerful for the form factor and purpose. Remember the thing should want to stay cool, so no point cramming a high-end mobile GPU in there.

[patrickjp93]

I've probably get plenty of time then, nothing first year is going to be too demanding I would imagine. What would be a good yet cheap processor to get into it when things get a bit more demanding?/what platform(I have never really looked into xeons so I feel a bit dumb).

E3 Xeon if you need it on Z180 since it will be out in half a year.

[Doobeedoo]

Would like to have top spec model with 1TB though
But price is way high for specs to. Still a great product! I'd like to see better spec one in future, like quad core and better dGPU but for cheaper price, please!

Really want something like this. 

[Curufinwe_wins]

Iris pro would have been ideal imho, but I was expecting at least a 950 equivalent. 

 

a 940m (aka the same gpu on the nvidia shield tv) is ok I guess, but ehh I won't be buying it anyways.

[Curufinwe_wins]

problem with iris pro in this form factor would the that it's on the cpu, which would mean it would be on the tablet part, which would mean insane amount of heat in one tiny place, by doing it this way they split the heat.

Not necessarily. We don't actually know how much iris pro consumes and being a -U part.

 

Note according to a arstechnica story a while back the iris pro 6200 is a 28W part. So I'd expect the skylake GT3e to be about 20-25W (they didn't increase EU's).

 

So total consumption at peak goes from 7-15W to 35-40W which is still super easy land (all of the HQ-HK parts are 47W if I remember correctly).

 

Now would this have required a better cooling solution? obviously. Does that mean it still wouldn't have been a better idea? Not in my opinion.

[randomhkkid]

Iris pro would have been ideal imho, but I was expecting at least a 950 equivalent. 

 

a 940m (aka the same gpu on the nvidia shield tv) is ok I guess, but ehh I won't be buying it anyways.

The 940m is much faster than the Tegra X1. 384 cores vs 256 and 5ghz memory vs 1.6ghz.

[ZetZet]

Not necessarily. We don't actually know how much iris pro consumes and being a -U part.

 

Note according to a arstechnica story a while back the iris pro 6200 is a 28W part. So I'd expect the skylake GT3e to be about 20-25W (they didn't increase EU's).

 

So total consumption at peak goes from 7-15W to 35-40W which is still super easy land (all of the HQ-HK parts are 47W if I remember correctly).

 

Now would this have required a better cooling solution? obviously. Does that mean it still wouldn't have been a better idea? Not in my opinion.

Yeah, exactly all the HQ parts are 40w+. That's why none of them are in slim laptops or tablets. 20w for something thinner than a cm would be insane.

[ionbasa]

This actually looks quite promising. SInce we don't know how well the new Iris Pro performs and microsoft needed an product to push, I think this was the best tradeoff. An 940M won't by any means give you the best performance, but it will enable users to at least use some CUDA accelerated apps (vs none at all).

[Curufinwe_wins]

Yeah, exactly all the HQ parts are 40w+. That's why none of them are in slim laptops or tablets. 20w for something thinner than a cm would be insane.

Here is the actual part they should have used.

 

http://ark.intel.com/products/91163/Intel-Core-i7-6560U-Processor-4M-Cache-up-to-3_20-GHz 15W TDP (downclockable to about 10W) or

http://ark.intel.com/products/91497/Intel-Core-i7-6650U-Processor-4M-Cache-up-to-3_40-GHz Same TDP

End of story.

 

Intel Iris Graphics 540 is GT3e (and is basically a 930m)

 

Had they actually wanted to use decent cooling, this is what I would have put on:

http://ark.intel.com/products/91167/Intel-Core-i7-6567U-Processor-4M-Cache-up-to-3_60-GHz 28W TDP (downclockable to 23W)

 

BTW in case you were wondering. They used the 6650U for the Surface Pro 4, but used the 6600U for the Surface book which is this poopsicle

http://ark.intel.com/products/88192/Intel-Core-i7-6600U-Processor-4M-Cache-up-to-3_40-GHz

[patrickjp93]

Yeah, exactly all the HQ parts are 40w+. That's why none of them are in slim laptops or tablets. 20w for something thinner than a cm would be insane.

Do you consider the Macbook Pro slim? It's thinner than 1cm from keyboard to base, and it has a 4980HQ and now the 5950HQ in it, both 47W parts.

[Curufinwe_wins]

Do you consider the Macbook Pro slim? It's thinner than 1cm from keyboard to base, and it has a 4980HQ and now the 5950HQ in it, both 47W parts.

And thermal throttles like a bitch. But still a decent enough point. Yet as the 6650 and 6560 show not even a needed tradeoff.

[ZetZet]

Here is the actual part they should have used.

 

http://ark.intel.com/products/91163/Intel-Core-i7-6560U-Processor-4M-Cache-up-to-3_20-GHz 15W TDP (downclockable to about 10W) or

http://ark.intel.com/products/91497/Intel-Core-i7-6650U-Processor-4M-Cache-up-to-3_40-GHz Same TDP

End of story.

 

Intel Iris Graphics 540 is GT3e (and is basically a 930m)

 

Had they actually wanted to use decent cooling, this is what I would have put on:

http://ark.intel.com/products/91167/Intel-Core-i7-6567U-Processor-4M-Cache-up-to-3_60-GHz 28W TDP (downclockable to 23W)

 

BTW in case you were wondering. They used the 6650U for the Surface Pro 4, but used the 6600U for the Surface book which is this poopsicle

http://ark.intel.com/products/88192/Intel-Core-i7-6600U-Processor-4M-Cache-up-to-3_40-GHz

We don't know what they used yet.

 

Do you consider the Macbook Pro slim? It's thinner than 1cm from keyboard to base, and it has a 4980HQ and now the 5950HQ in it, both 47W parts.

Well did you see the surface book? 7.7mm on the screen side and the screen is glass with a touchscreen. It's HELLA thin.

[patrickjp93]

And thermal throttles like a bitch. But still a decent enough point. Yet as the 6650 and 6560 show not even a needed tradeoff.

No it doesn't, at all. The multicore boost is 3.4 for the 4980HQ, which it can sustain in a 47W TDP in the MBPr 15". That's not throttling, just as the 4790K only sustaining 4.2GHz on all-core load isn't throttling. It's the actual design spec.

[Curufinwe_wins]

 

We don't know what they used yet.

 

Well did you see the surface book? 7.7mm on the screen side and the screen is glass with a touchscreen. It's HELLA thin.

 

Actually we do. We know the Surface Pro 4 uses the i7-6650U, and we KNOW the Surface Book uses the i7-6600U.

 

http://www.pcworld.com/article/2989906/laptop-computers/here-are-the-details-of-the-surface-book-and-surface-pro-4-chips-and-why-they-matter.html

[ZetZet]

Actually we do. We know the Surface Pro 4 uses the i7-6650U, and we KNOW the Surface Book uses the i7-6600U.

 

http://www.pcworld.com/article/2989906/laptop-computers/here-are-the-details-of-the-surface-book-and-surface-pro-4-chips-and-why-they-matter.html

Well so it's good.

[Curufinwe_wins]

No it doesn't, at all. The multicore boost is 3.4 for the 4980HQ, which it can sustain in a 47W TDP in the MBPr 15". That's not throttling, just as the 4790K only sustaining 4.2GHz on all-core load isn't throttling. It's the actual design spec.

You are kidding right? look up Macbook Pro thermal throttling and LITERALLY EVERYONE SEES IT. It hits thermal limits in minutes and throttles massively.

 

 

 

After our one-hour stress test with Prime95 and FurMark (Windows), the CPU runs at only1.2 GHz, while the graphics card is also limited to just 400 MHz. Even though devices from Asus, Acer & Co. also throttle, none of the direct rivals loses that much performance. The performance of the MacBook was much better under OS X, but the tools (Cinebench and Unigine Heaven) are not that demanding for the hardware.

From this review. http://www.notebookcheck.net/Apple-MacBook-Pro-Retina-15-Mid-2015-Review.144402.0.html

 

This is literally the reason LTT made that watercooling the 2015 Macbook (and that little bitch uses an ATOM processor.) The issue isn't fixed on the rMBP even though they add fans because again the processor uses so much more heat.

[patrickjp93]

You are kidding right? look up Macbook Pro thermal throttling and LITERALLY EVERYONE SEES IT. It hits thermal limits in minutes and throttles massively.

 

From this review. http://www.notebookcheck.net/Apple-MacBook-Pro-Retina-15-Mid-2015-Review.144402.0.html

 

This is literally the reason LTT made that watercooling the 2015 Macbook (and that little bitch uses an ATOM processor.) The issue isn't fixed on the rMBP even though they add fans because again the processor uses so much more heat.

funny because my MBPr runs AIDA 64 for 5 minutes straight and remains at 3.4GHz, though it does start to scream.

[Curufinwe_wins]

Well so it's good.

Except they stuck a worse processor on the Surface Book Pro so they could add a gimmicky nvidia cpu. Afterall that additional space could have been used for batteries or something. (hell battery life would have been significantly better if they just didn't use the dGPU at all. (the 930m is a 25W gpu.)

[Curufinwe_wins]

funny because my MBPr runs AIDA 64 for 5 minutes straight and remains at 3.4GHz, though it does start to scream.

Dat 6000 rpm fan... Note the time differences I mentioned. 1 hr vs 5 minutes.