I'm pretty confident we'll see this in the 15" rMBP. If you remember Iris Pro which Apple nearly single handedly asked Intel for, I can see them being the ones to push for this as well.
Wonder if that means a redesign, if not even just more batteries in the same space would be nice, or more room for cooling. And, cough, SD card. Most likely none of that, but the move up from the meager 80GB/s feeding the current GPUs would go a long way, even aside from the extra execution resources.
Based on the TDPs and looking around at the market, they seem tailor made for VR work on imacs, possible mac minis (though I think that line is dead as soon as the stock is sold off).
Actually, the 15" throttles a whole lot if you do anything 'Pro' with it. When I do video encodes, it begins slowing down after the first 10 solid minutes or so. I love the machine, the build, the look, BUT, you are wasting money if you get a CPU above the base, because you are rarely going to get full speed under sustained load. If they move to this model and can actually maintain speed, I will most certainly upgrade. The Kaby Lake MBP has been a disappointment to me.
FWIW apple has shipped plenty MBPs where the charger isn't quite sufficient. These will drain the battery a little even if plugged in when running at full tilt (and at least some of them also have the habit of running really really slow if the battery isn't just old but completely dead because they will be forced to low power states). Albeit I agree for a 89W charger a 100W cpu+gpu is probably too much, since together with the rest of the system that might amount to a sustained power draw of over 110W, which would drain the battery too fast. But if apple wants a 80W version of it, I'm pretty sure intel would just deliver that, those limits can be easily changed.
For what it's worth, MS SQL Server on Linux/Docker is fairly limited, and the mgt software is still windows based, though you can do anything you need via sql execute statements... it's not the friendliest. I usually treat my database as mostly dumb anyway.
I just don't get this marriage. It seems the graphics power is juuust a wee bit over intels cores, so what graphics need would this push? "Oh look I get 5 more fps in Minecraft!"??
I thought it would be faster than that, so i did some back of the napkin math. 24 CUs = 43% of a Vega 56, but 19% slower on the GPU and 50% slower on the HBM. Seems reasonable to guess it will offer about 25% the performance of a Vega 56.
Vega56 gets 20k in 3DMark, 25% of that is 5k. The fastest iGPU I could find on futuremark's site is the 6700HQ, which scored 7910. So... it's slower than the fastest intel GPU from 2 years ago. Is that right?
Yeah that wasn't right, futuremark switched it from GPU to CPU when I searched for Intel. The fastest GPU score I could find for an Intel iGPU was the Iris Pro 6200 from the Broadwell generation. It got 1630. Skylake improved the iGPU quite a lot but I can't find the benchmarks offhand.
Hm? It's definitely a fair shot over the Iris Plus 650, and the Pro line seems dead after the Pro 580. This will absolutely be 3-4x over an Iris Plus 650, let alone the eDRAM-less Iris HD 630 thrown in there.
What did you mean by barely above the Intel part? I see nothing close.
What's the point if it doesn't remotely make sense? That's not an exaggeration, it's just not the truth, the dGPU is significantly better than any Intel iGPU.
Umm. Lol can you read? Perhaps you need your eyes checked? Because I really can't fathom how you ended up at THAT conclusion. This is a near GTX 1060 level part we're talking about here (and well beyond the 1050Ti). As others have said, it'll fall right around the MaxQ version of the 1060. That's damn impressive considering the size & power envelope.
In page 4 ("Intel’s Performance Numbers") the last table's headers are wrong (should be GH vs 1060).
I wonder how FreeSync would work if they're powering down the pGPU. One guess is some sort of mode switching to the iGPU's display engine as for light workloads the iGPU is more than capable to maintain max display frame rate. But then, there's some protocol stuff being done in FreeSync/G-Sync so it could also be that the pGPU's display engine would always be on with the iGPU feeding into it.
I don't get it, AMD GPU has never been good for efficiency, but now they claim to beat NVdia 1050/1050 ti in terms of efficiency.... sounds too good to be true.
EMIB + HBM2 + a multichip module with Intel is what it takes apparently. With all those edges, I'm sure it can manage to edge out 1050 perf/watt like they say, it's not all in the architecture.
EMIB adds nothing performance or efficiency-wise over a regular interposer like AMD uses. It's simply thinner/cheaper. Also, people drastically undervalue how power efficient Vega is when clocked in it's efficiency "sweet spot". The reason it looks so bad in Vega 56/64 (the latter especially) is because the clocks have been pushed right up to the process' limits, and well, well beyond said "sweet spot". The clock's being used here, otoh, fall right within it (for obvious reasons). I think people are going to be very surprised by both the efficiency here, and in the Vega Mobile dGPU AMD announced today (which could very well be based off this graphics part, but we'll know soon enough. I'd bet my lunch it ends up in discrete desktop cards as well at some point to replace at-least the 560 & 570, and possibly 580).
Where do they claim that it will beat the GTX 1050 in terms of power efficiency? They show some select benchmarks that imply a certain efficiency in those specific cases, but I didn't see that they mentioned general power efficiency or price at all.
This package from Intel does have HBM, which is more power efficient than GDDR5. That will help. But overall, my expectation is that Intel's new chip will be less efficient in graphics intensive tasks than a system with a latest generation discrete NVIDIA GPU. The dynamic tuning should help in cases where both CPU and GPU need to draw significant power, though.
We probably know how Vega performs. Assuming that the chips aren't TDP constrained, the more powerful of the two variants should probably perform somewhere between a 560 and 570 in games. The lesser variant should perform around a 560, less or more depending on how memory bandwidth plays into things. We'll have to see how power constraints factor into to things though.
Another thing to keep in mind is that for most of its lifetime, this chip will probably be going up against NVIDIA's next generation of GPUs and not their current generation. Intel did benchmark it against a 950M, but I wouldn't put it past them to ignore price differences in a comparison they release. The new chips will probably be expensive enough that they will have to go up against the latest generation of their competitor's chips.
This does leave room for Intel produce a slimmer GT1 or even omitting a GPU entirely for mobile when the know that it will be paired with a Radeon Vega on package. That'd permit Intel to decrease costs on their end, though this would up to Intel to pass onward to OEMs.
Vega is actually pretty efficient, just not when they try to chase high performance, then the power requirements jump exponentially in response to the higher clocks and voltage. Also, AMD has had the fficiency crown multiple times, just not recently. The Radeon 9700 pro, 9800 pro, 4850, 4870, 5850, 5870, 7790, 7950, and 7970 all say hello when compared to their Nvidia counterparts of the time.
In mobile the only difference was the core count, it doubled when coffee lake was released, but this kaby lake has similar core counts for some reason.
So will we see a 24 CU Vega-chip with HBM2 as a discrete chip for laptops too? It's essentially designed as a lower tier RTG/AMD GPU, kinda specced like a replacement for Polaris 11.
AMD announced Vega Mobile today which is pretty much exactly that. It's a freaking tiny package, and for that reason I expect it to be pretty successful vs Nvidia's traditional GDDR5 designs. (Though no word yet if the Vega Mobile and the semi-custom chip here are the same as far as CU count and what not, but it wouldn't be surprising).
Same Z-height as KBL-G at least. Vega mobile was expected, is definitively related to the chip on the KBL-G package but might be a chip with a slightly different CU/SP count but renders make it look exactly the same.
The one thing missing from this article, which I think if fair in the context of the fury of news last week, are Spectre and Meltdown. Intel is currently facing multiple law suits about it, especially in the context of bringing to market with a security flaw. These security flaws maybe the death blow to Cannon Lake which was originally to be a late 2018 part and already had the desktop parts removed form the line up in favor of Coffee Lake.
There is another Lake part coming in late 2018 called Whiskey Lake and the rumors are pointing toward it being yet another Sky Lake based 14 nm part (see Kaby and Coffee Lake). I have no idea what these parts could provide other than potential fixes for Meltdown and Spectre minus the offhand possibility of an updated GPU. Next actual CPU core design is set to be Ice Lake which may also be their first 10 nm chip. Intel has shied away from doing to many firsts as once due to the difficulty of isolating problem (is it process? design? packaging? interconnect?) but Intel may have no choice.
Intel's CEO is set to take the stage at CES tomorrow so we'll probably get some answers to their roadmap as well has some groveling about Meltdown and Spectre.
Never underestimate the power of hatred against a mutual enemy. It worked for the allies in World War II, at least until that nice little cold war bit that came after :).
Good Jobs. but still dissapointed. intel not use hbm2 as l4 cache and share processor graphics when user wants to use graphics card and acording to ark.intel . And it is still 4 cores not 6 cores like i7 8700k
It's a laptop chip first and foremost, and the best & latest Intel has in it's mobile line is 4c/8t Kaby Lake for power reasons (and the max 100W power envelope here precludes 6c/12t Coffee Lake already, even if a hypothetical mobile CL part existed). Not to be rude, but your expectations were totally unreasonable considering the primary target market (thin & light gaming laptops & mobile workstations).
"It's a laptop chip first and foremost" ??? ----------------------------------------------- It may have been presented that way initially but there were hints for other products from the very beginning
Once the process is optimized over the next few years, we may start seeing some very capable 4K TV's without the need for thunderbolt graphics cards
Now, about that latency problem....... Whats new for gaming TV's at CES?
6c/12t would have been perfectly possible with Vega M under 100W. The catch is both the CPU and GPU wouldn't coexist well under full load. The end result would be a base clock lower than what Intel would have liked on both parts for that fully loaded scenario. Though under average usage (including gaming were 4c/8t was enough), turbo would kick in and everything would be OK.
The more likely scenario is that Intel simply didn't have enough time in development of this product to switch Kaby Lake for Coffer Lake in time and get this validated. Remember that Coffee Lake was added to the road map when Cannon Lake desktop chips were removed.
Out-earning AMD by far in that corner...I have a feeling this isn't a super high margin product, AMD just needs sales, look at what they sold 'firepros' to Apple for.
Seems to me, had AMD denied their GPU to Intel, Intel would have no decent SOC product to launch. Meanwhile AMD could release their Zen/Vega APU and be the only guy in town. Apple would have took notice.
Hi Ian, apparently most websites seem to have failed to notice that the standard height of this product class is the same as the height of AMD's own Vega Mobile, which is set at 1.7mm.
Intel clearly states its z-height is 1.7mm so where's the advantage ?!
Therefore, it appears that Intel's EMIB talk is just talk (in this current implementation) and saves no "height" , as correctly pointed out by SemiAccurate.com :
"note that the Z-height, a critical factor in modern notebooks, is the exact same 1.7mm as a Vega-M discrete GPU. Why is this important? It looks like EMIB saves ~0mm in Z-height versus a much simpler to manufacture interposer. Interesting, no?"
On the "Intel with Radeon vs i7-7700HQ + GTX 1060 Max-Q Data from Intel, not AnandTech" table, you have the second-from-left column mislabeled ("i7-8550U + GTX 1050")
That would finally be a mac mini that was worth a damn again, although honestly their place has been taken up by the multitude of small NUC like pc's that are now available. But it would beef up the low end on the mac line, not that Apple cares about the low end, or most of its customers, but i digress.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
66 Comments
Back to Article
Hixbot - Sunday, January 7, 2018 - link
Pigs just flew.tipoo - Sunday, January 7, 2018 - link
I'm pretty confident we'll see this in the 15" rMBP. If you remember Iris Pro which Apple nearly single handedly asked Intel for, I can see them being the ones to push for this as well.Wonder if that means a redesign, if not even just more batteries in the same space would be nice, or more room for cooling. And, cough, SD card. Most likely none of that, but the move up from the meager 80GB/s feeding the current GPUs would go a long way, even aside from the extra execution resources.
tipoo - Sunday, January 7, 2018 - link
That, uh, was meant to be a new comment.Comments, but editable.
nico_mach - Monday, January 8, 2018 - link
Based on the TDPs and looking around at the market, they seem tailor made for VR work on imacs, possible mac minis (though I think that line is dead as soon as the stock is sold off).willis936 - Monday, January 8, 2018 - link
Not even Tim Cook's Apple is crazy enough to put a 65W part in a 15" laptop. Hopefully at least.tyaty1 - Monday, January 8, 2018 - link
It is less than the combined TDP of CPU and the dGPU on the latest MBP.WinterCharm - Monday, January 8, 2018 - link
The MacBook Pro's current TDP is 87W.I'm amazed that they top out at 85ºC and barely throttle. It's kind of insane.
cditty - Sunday, January 21, 2018 - link
Actually, the 15" throttles a whole lot if you do anything 'Pro' with it. When I do video encodes, it begins slowing down after the first 10 solid minutes or so. I love the machine, the build, the look, BUT, you are wasting money if you get a CPU above the base, because you are rarely going to get full speed under sustained load. If they move to this model and can actually maintain speed, I will most certainly upgrade. The Kaby Lake MBP has been a disappointment to me.skavi - Monday, January 8, 2018 - link
You have to realize that we're already well past that point. There's a 45W CPU and 35W GPU in the current MacBook Pro.lefenzy - Monday, January 8, 2018 - link
The 15" rMBP ships with an 89W charger. So only the 65 W solutions would work here.mczak - Monday, January 8, 2018 - link
FWIW apple has shipped plenty MBPs where the charger isn't quite sufficient. These will drain the battery a little even if plugged in when running at full tilt (and at least some of them also have the habit of running really really slow if the battery isn't just old but completely dead because they will be forced to low power states).Albeit I agree for a 89W charger a 100W cpu+gpu is probably too much, since together with the rest of the system that might amount to a sustained power draw of over 110W, which would drain the battery too fast. But if apple wants a 80W version of it, I'm pretty sure intel would just deliver that, those limits can be easily changed.
Kevin G - Sunday, January 7, 2018 - link
And MS SQL Server is available for Linux. I think hell has frozen over.tracker1 - Monday, January 8, 2018 - link
For what it's worth, MS SQL Server on Linux/Docker is fairly limited, and the mgt software is still windows based, though you can do anything you need via sql execute statements... it's not the friendliest. I usually treat my database as mostly dumb anyway.Zingam - Sunday, January 7, 2018 - link
But does it melt down?haukionkannel - Monday, January 8, 2018 - link
Yes it does.B166ER - Sunday, January 7, 2018 - link
I just don't get this marriage. It seems the graphics power is juuust a wee bit over intels cores, so what graphics need would this push? "Oh look I get 5 more fps in Minecraft!"??schizoide - Sunday, January 7, 2018 - link
I thought it would be faster than that, so i did some back of the napkin math. 24 CUs = 43% of a Vega 56, but 19% slower on the GPU and 50% slower on the HBM. Seems reasonable to guess it will offer about 25% the performance of a Vega 56.Vega56 gets 20k in 3DMark, 25% of that is 5k. The fastest iGPU I could find on futuremark's site is the 6700HQ, which scored 7910. So... it's slower than the fastest intel GPU from 2 years ago. Is that right?
schizoide - Sunday, January 7, 2018 - link
Yeah that wasn't right, futuremark switched it from GPU to CPU when I searched for Intel. The fastest GPU score I could find for an Intel iGPU was the Iris Pro 6200 from the Broadwell generation. It got 1630. Skylake improved the iGPU quite a lot but I can't find the benchmarks offhand.JohnPec - Monday, January 8, 2018 - link
Linus said it will be as good as 1060maxq or better.tipoo - Sunday, January 7, 2018 - link
Hm? It's definitely a fair shot over the Iris Plus 650, and the Pro line seems dead after the Pro 580. This will absolutely be 3-4x over an Iris Plus 650, let alone the eDRAM-less Iris HD 630 thrown in there.What did you mean by barely above the Intel part? I see nothing close.
B166ER - Sunday, January 7, 2018 - link
Exaggeration to emphasize a point..tipoo - Monday, January 8, 2018 - link
What's the point if it doesn't remotely make sense? That's not an exaggeration, it's just not the truth, the dGPU is significantly better than any Intel iGPU.Cooe - Monday, January 8, 2018 - link
Umm. Lol can you read? Perhaps you need your eyes checked? Because I really can't fathom how you ended up at THAT conclusion. This is a near GTX 1060 level part we're talking about here (and well beyond the 1050Ti). As others have said, it'll fall right around the MaxQ version of the 1060. That's damn impressive considering the size & power envelope.OEMG - Sunday, January 7, 2018 - link
In page 4 ("Intel’s Performance Numbers") the last table's headers are wrong (should be GH vs 1060).I wonder how FreeSync would work if they're powering down the pGPU. One guess is some sort of mode switching to the iGPU's display engine as for light workloads the iGPU is more than capable to maintain max display frame rate. But then, there's some protocol stuff being done in FreeSync/G-Sync so it could also be that the pGPU's display engine would always be on with the iGPU feeding into it.
neblogai - Sunday, January 7, 2018 - link
On 4th page, last two charts (and also text between them) should have GH series part, instead of i7-8509G Vega M GL.StevoLincolnite - Sunday, January 7, 2018 - link
"Coffee Lake processors, using Intel’s latest 14++ process and running up to 8 cores."I could have sworn they only topped out at 6-cores.
nerd1 - Sunday, January 7, 2018 - link
I don't get it, AMD GPU has never been good for efficiency, but now they claim to beat NVdia 1050/1050 ti in terms of efficiency.... sounds too good to be true.tipoo - Sunday, January 7, 2018 - link
EMIB + HBM2 + a multichip module with Intel is what it takes apparently. With all those edges, I'm sure it can manage to edge out 1050 perf/watt like they say, it's not all in the architecture.tipoo - Sunday, January 7, 2018 - link
Also states power sharing saves 18 watts, so that's a lot of handicaps for the more efficient Pascal to catch up with.Cooe - Monday, January 8, 2018 - link
EMIB adds nothing performance or efficiency-wise over a regular interposer like AMD uses. It's simply thinner/cheaper. Also, people drastically undervalue how power efficient Vega is when clocked in it's efficiency "sweet spot". The reason it looks so bad in Vega 56/64 (the latter especially) is because the clocks have been pushed right up to the process' limits, and well, well beyond said "sweet spot". The clock's being used here, otoh, fall right within it (for obvious reasons). I think people are going to be very surprised by both the efficiency here, and in the Vega Mobile dGPU AMD announced today (which could very well be based off this graphics part, but we'll know soon enough. I'd bet my lunch it ends up in discrete desktop cards as well at some point to replace at-least the 560 & 570, and possibly 580).haukionkannel - Monday, January 8, 2018 - link
This. Vega is very effisient in low clock rates!Yojimbo - Sunday, January 7, 2018 - link
Where do they claim that it will beat the GTX 1050 in terms of power efficiency? They show some select benchmarks that imply a certain efficiency in those specific cases, but I didn't see that they mentioned general power efficiency or price at all.This package from Intel does have HBM, which is more power efficient than GDDR5. That will help. But overall, my expectation is that Intel's new chip will be less efficient in graphics intensive tasks than a system with a latest generation discrete NVIDIA GPU. The dynamic tuning should help in cases where both CPU and GPU need to draw significant power, though.
We probably know how Vega performs. Assuming that the chips aren't TDP constrained, the more powerful of the two variants should probably perform somewhere between a 560 and 570 in games. The lesser variant should perform around a 560, less or more depending on how memory bandwidth plays into things. We'll have to see how power constraints factor into to things though.
Another thing to keep in mind is that for most of its lifetime, this chip will probably be going up against NVIDIA's next generation of GPUs and not their current generation. Intel did benchmark it against a 950M, but I wouldn't put it past them to ignore price differences in a comparison they release. The new chips will probably be expensive enough that they will have to go up against the latest generation of their competitor's chips.
Kevin G - Monday, January 8, 2018 - link
This does leave room for Intel produce a slimmer GT1 or even omitting a GPU entirely for mobile when the know that it will be paired with a Radeon Vega on package. That'd permit Intel to decrease costs on their end, though this would up to Intel to pass onward to OEMs.nico_mach - Monday, January 8, 2018 - link
AMD wasn't good at efficiency mostly due to fabbing. That's easily fixable with a deep-pocketed and suddenly desperate partner like Intel.artk2219 - Wednesday, January 10, 2018 - link
Vega is actually pretty efficient, just not when they try to chase high performance, then the power requirements jump exponentially in response to the higher clocks and voltage. Also, AMD has had the fficiency crown multiple times, just not recently. The Radeon 9700 pro, 9800 pro, 4850, 4870, 5850, 5870, 7790, 7950, and 7970 all say hello when compared to their Nvidia counterparts of the time.jjj - Sunday, January 7, 2018 - link
Ask AMD for a die shot so we can count CUs lolshabby - Sunday, January 7, 2018 - link
8th generation... kaby lake? Have i been sleeping under a rock?evilpaul666 - Sunday, January 7, 2018 - link
Is there a difference between Skylake, Kaby Lake and Coffee Lake that I'm unaware of?shabby - Sunday, January 7, 2018 - link
In mobile the only difference was the core count, it doubled when coffee lake was released, but this kaby lake has similar core counts for some reason.extide - Sunday, January 7, 2018 - link
Yeah, for U/Y (and now G) series 8th gen is 'Kaby Lake refresh, not Coffee Lake)skavi - Monday, January 8, 2018 - link
Do we even have 8th gen Y?evilpaul666 - Sunday, January 7, 2018 - link
So is this all Intel has for CES?extide - Sunday, January 7, 2018 - link
So, here we have a MCM with chips from three different fabs on it. 1 from Intel, 1 from GF, and one from Samsung or SKHynix.Have we ever seen something like that before?
Penti - Sunday, January 7, 2018 - link
So will we see a 24 CU Vega-chip with HBM2 as a discrete chip for laptops too? It's essentially designed as a lower tier RTG/AMD GPU, kinda specced like a replacement for Polaris 11.Cooe - Monday, January 8, 2018 - link
AMD announced Vega Mobile today which is pretty much exactly that. It's a freaking tiny package,and for that reason I expect it to be pretty successful vs Nvidia's traditional GDDR5 designs. (Though no word yet if the Vega Mobile and the semi-custom chip here are the same as far as CU count and what not, but it wouldn't be surprising).
Penti - Monday, January 8, 2018 - link
Same Z-height as KBL-G at least. Vega mobile was expected, is definitively related to the chip on the KBL-G package but might be a chip with a slightly different CU/SP count but renders make it look exactly the same.flashbacck - Sunday, January 7, 2018 - link
"Intel’s internal graphics, known as ‘Gen’ graphics externally, has been third best behind NVIDIA and AMD for grunt."Grunt? huh?
Holliday75 - Monday, January 8, 2018 - link
I thought the same thing.Kevin G - Sunday, January 7, 2018 - link
The one thing missing from this article, which I think if fair in the context of the fury of news last week, are Spectre and Meltdown. Intel is currently facing multiple law suits about it, especially in the context of bringing to market with a security flaw. These security flaws maybe the death blow to Cannon Lake which was originally to be a late 2018 part and already had the desktop parts removed form the line up in favor of Coffee Lake.There is another Lake part coming in late 2018 called Whiskey Lake and the rumors are pointing toward it being yet another Sky Lake based 14 nm part (see Kaby and Coffee Lake). I have no idea what these parts could provide other than potential fixes for Meltdown and Spectre minus the offhand possibility of an updated GPU. Next actual CPU core design is set to be Ice Lake which may also be their first 10 nm chip. Intel has shied away from doing to many firsts as once due to the difficulty of isolating problem (is it process? design? packaging? interconnect?) but Intel may have no choice.
Intel's CEO is set to take the stage at CES tomorrow so we'll probably get some answers to their roadmap as well has some groveling about Meltdown and Spectre.
Oh, for the curious, you can look up what codenames Intel does have the horizon here and see Whiskey Lake for yourselves:
https://www.intel.com/content/www/us/en/design/pro...
UtilityMax - Thursday, January 11, 2018 - link
This gotta be the worst time to announce a new Intel CPU. Sadly, I think the sheep that are the average consumers will gladly buy PCs with these CPUs.MFinn3333 - Monday, January 8, 2018 - link
Intel and AMD were forced to work together because they hated nVidia.This is a relationship of Spite.
itonamd - Monday, January 8, 2018 - link
and hate qualcomm works for windows 10artk2219 - Wednesday, January 10, 2018 - link
Never underestimate the power of hatred against a mutual enemy. It worked for the allies in World War II, at least until that nice little cold war bit that came after :).itonamd - Monday, January 8, 2018 - link
Good Jobs. but still dissapointed. intel not use hbm2 as l4 cache and share processor graphics when user wants to use graphics card and acording to ark.intel . And it is still 4 cores not 6 cores like i7 8700kCooe - Monday, January 8, 2018 - link
It's a laptop chip first and foremost, and the best & latest Intel has in it's mobile line is 4c/8t Kaby Lake for power reasons (and the max 100W power envelope here precludes 6c/12t Coffee Lake already, even if a hypothetical mobile CL part existed). Not to be rude, but your expectations were totally unreasonable considering the primary target market (thin & light gaming laptops & mobile workstations).Bullwinkle-J-Moose - Monday, January 8, 2018 - link
"It's a laptop chip first and foremost" ???-----------------------------------------------
It may have been presented that way initially but there were hints for other products from the very beginning
Once the process is optimized over the next few years, we may start seeing some very capable 4K TV's without the need for thunderbolt graphics cards
Now, about that latency problem.......
Whats new for gaming TV's at CES?
Kevin G - Monday, January 8, 2018 - link
6c/12t would have been perfectly possible with Vega M under 100W. The catch is both the CPU and GPU wouldn't coexist well under full load. The end result would be a base clock lower than what Intel would have liked on both parts for that fully loaded scenario. Though under average usage (including gaming were 4c/8t was enough), turbo would kick in and everything would be OK.The more likely scenario is that Intel simply didn't have enough time in development of this product to switch Kaby Lake for Coffer Lake in time and get this validated. Remember that Coffee Lake was added to the road map when Cannon Lake desktop chips were removed.
Bullwinkle-J-Moose - Monday, January 8, 2018 - link
You had it right the first timeCoffer Lake holds all the cash
Hurr Durr - Monday, January 8, 2018 - link
Come on, this particular cartel is quite obvious.Strunf - Monday, January 8, 2018 - link
This all proves that X86 wars are a thing of the past and NVIDIA is pushing these two into a corner...tipoo - Monday, January 8, 2018 - link
Out-earning AMD by far in that corner...I have a feeling this isn't a super high margin product, AMD just needs sales, look at what they sold 'firepros' to Apple for.Hixbot - Monday, January 8, 2018 - link
Seems to me, had AMD denied their GPU to Intel, Intel would have no decent SOC product to launch. Meanwhile AMD could release their Zen/Vega APU and be the only guy in town. Apple would have took notice.IGTrading - Monday, January 8, 2018 - link
Hi Ian, apparently most websites seem to have failed to notice that the standard height of this product class is the same as the height of AMD's own Vega Mobile, which is set at 1.7mm.Intel clearly states its z-height is 1.7mm so where's the advantage ?!
Therefore, it appears that Intel's EMIB talk is just talk (in this current implementation) and saves no "height" , as correctly pointed out by SemiAccurate.com :
"note that the Z-height, a critical factor in modern notebooks, is the exact same 1.7mm as a Vega-M discrete GPU. Why is this important? It looks like EMIB saves ~0mm in Z-height versus a much simpler to manufacture interposer. Interesting, no?"
Source : https://semiaccurate.com/2018/01/07/intel-kaby-g-n...
and
https://hexus.net/media/uploaded/2018/1/81819abc-2...
Is this a wrong assumption to make, or is Intel lying when they say EMIB is "better" than AMD's interposer ?
boeush - Monday, January 8, 2018 - link
Ian,On the "Intel with Radeon vs i7-7700HQ + GTX 1060 Max-Q Data from Intel, not AnandTech" table, you have the second-from-left column mislabeled ("i7-8550U + GTX 1050")
alumii - Tuesday, January 9, 2018 - link
3 words: New Mac Miniartk2219 - Wednesday, January 10, 2018 - link
That would finally be a mac mini that was worth a damn again, although honestly their place has been taken up by the multitude of small NUC like pc's that are now available. But it would beef up the low end on the mac line, not that Apple cares about the low end, or most of its customers, but i digress.