When do you think we will see the first AMD High power APU for consumer PC?

[cbn]

By high power consumer APU I mean any APU with at least 150W TDP. This primarily for laptops like the 14" Razor Blade which has 45W i7-7700HQ and 120W ~100W GTX 1060 6GB (2014 version shown below as I don't have teardown pictures of the 2017 14" Razor Blade)

P.S. Here is another laptop (MSI GS63VR Stealth Pro) with 45W CPU (i7 6700HQ) and 120W ~100W dGPU (GTX 1060 6GB) in a 15.6" package at sub 4 lbs:

EDIT: Corrected dGPU TDP (See post #8)

 

[cbn]

There are actually laptops with higher TDPs than the ones I have shown in the opening post. The Razor Blade Pro has a overclocked 45W Intel quad core (?? TDP) and GTX 1080 (180W ~160W TDP) fed via a 250W power brick:

EDIT: Corrected dGPU TDP (see post #8)

https://www.razerzone.com/gaming-systems/razer-blade-pro

(It is 17.3" and 7.69 lbs though....almost twice the weight of the 14" Razor Blade mentioned in the OP)

 

[DrMrLordX]

Until we start seeing HBM2 in consumer-level products (read: not HPC-oriented Nvidia dGPUs), I don't see it happening.

 

[w3rd]

I expect to see mobile APU from AMD, featuring Zen + Vega in laptops as soon as 2017 Holiday season.

You would no longer need "alienware" style gaming laptop. Freesync2 for everyone..!

 

[sm625]

By all rights it should be this year. But knowing AMD... who knows.

 

[Phynaz]

Never.

150w consumer CPU is a non-starter. No OEM will touch it.

But your example is flawed, the 1060m is 80w, not 120w.

 

[NTMBK]

We already have, it's called the PS4.

 

[cbn]

But your example is flawed, the 1060m is 80w, not 120w.

Nvidia dropped the M naming so it is just GTX 1060. I couldn't find a TDP listed here, but this article is claiming around 100W. This would put the combination of 45W CPU and GTX 1060 6GB GPU (in laptop) at around 145W Total.

Nvidia hasn't revealed the TDP of the new laptop GPUs, but notes that the GTX 1080 sports a similar TDP to the GTX 980 (160W), the GTX 1070 to the GTX 980M (125W), and GTX 1060 to GTX 970M (around 100W).

 

[cbn]

One comment I have seen regarding these lightweight laptops with high power CPU and dGPUs is that they run loud and hot. The following is an excerpt from the second link I have in post #8:

The tradeoff for all that performance is noise and battery life. While much larger and thicker laptops like Clevo's P870DM or Asus' ugly G752VY laptop remained surprisingly cool and quiet under load in demos, both the Razer Blade and the MSI GS63 ran loud and hot.

However, looking at the pictures below of the 2014 Razor Blade and MSI GS63, I can imagine a HBM2 150W APU offering a lot more room for larger fans (Razor Blade, first picture) or an additional fan in the MSI GS63 (imagine a fourth fan occuping the space used by the CPU in the second picture below):

 

[Rifter]

November 15, 2013

 

[cbn]

We already have, it's called the PS4.

November 15, 2013

Changed the thread's title to read consumer PC rather than just consumer. This to avoid confusion with the console APUs.

 

[Phynaz]

Nvidia dropped the M naming so it is just GTX 1060. I couldn't find a TDP listed here, but this article is claiming around 100W. This would put the combination of 45W CPU and GTX 1060 6GB GPU (in laptop) at around 145W Total.

80w - https://www.notebookcheck.net/Mobil...060-Laptop-Benchmarks-and-Specs.169547.0.html

75w - https://www.pcper.com/reviews/Graph...GTX-1080-1070-and-1060-Enter-Gaming-Notebooks

80w - http://gpu.specshero.com/en/nvidia-geforce-gtx-1060m-power-consumption/model-541-3

 

[Atari2600]

There is/are some small minded folks around.

It is blindingly obvious that a high performance APU will find significant traction in the market as a number cruncher. Since that market is quite lucrative, it is worth designing for. After designing/building for that lucrative market, it is a small step to use drivers to limit compute performance (in a similar vein to how consumer and professional GPU drivers work) and release the same part as a consumer product.

 

[Phynaz]

That's just what AMD needs, more products that nobody buys.

/s

 

[w3rd]

That's just what AMD needs, more products that nobody buys.

/s

Yeah, because you want people to buy those outdated $2k gaming laptops, with "dual" GPUs in them, instead of a $800 laptop, with a new APU in it, right..?

Is that what you are railing against?

 

[Arachnotronic]

That's just what AMD needs, more products that nobody buys.

/s

It's a good thing major corporations don't take product/SKU advice from people on internet message boards

 

[cbn]

Nvidia dropped the M naming so it is just GTX 1060. I couldn't find a TDP listed here, but this article is claiming around 100W. This would put the combination of 45W CPU and GTX 1060 6GB GPU (in laptop) at around 145W Total.

80w - https://www.notebookcheck.net/Mobil...060-Laptop-Benchmarks-and-Specs.169547.0.html

75w - https://www.pcper.com/reviews/Graph...GTX-1080-1070-and-1060-Enter-Gaming-Notebooks

80w - http://gpu.specshero.com/en/nvidia-geforce-gtx-1060m-power-consumption/model-541-3

Here is what the author in the first link wrote:

The TDP is slightly less than the old GTX 970M (predecessor).

(snip)

The power consumption should be around 80 Watt (TGP)

80W is slightly less than 100W of the linked GTX 970M? (Also note a different term "TGP" is being used for this 80W rather than the term "TDP")
.
Also it sounds like Nvidia was more specific to the author of the Arstechnica article I linked from post #8:

Nvidia hasn't revealed the TDP of the new laptop GPUs, but notes that the GTX 1080 sports a similar TDP to the GTX 980 (160W), the GTX 1070 to the GTX 980M (125W), and GTX 1060 to GTX 970M (around 100W)

 

[Atari2600]

https://www.top500.org/news/pondering-amds-ambitions-for-high-performance-apus/

AMD CEO Lisa Su reiterated the goal to build such a processor at the end of last week’s Q2 earnings call. “I think it’s fair to say that we do believe that we can combine a high performance CPU with a high performance GPU,” said Su, adding that “it’s an area where combining the two technologies makes a lot of sense.”

http://www.kitguru.net/components/a...n/amd-said-to-be-working-on-zen-apu-with-hbm/

It looks like we can expect an AMD APU based on the Zen architecture in the future, though rather interestingly, it will include stacked High-Bandwidth Memory (HBM) with bandwidth speeds of 128GB/s along with a larger onboard GPU for better gaming performance.

This information comes to light from a paper(via: wccftech), which was co-authored by AMD graphics engineer, Mike Mantor. The paper shows that this Zen-based APU would make use of an upgraded version of the memory technology found in Carrizo, which started off as an improved design over the memory interconnects found on the current generation consoles.

On top of that, the Zen APU is shown to use stacked HBM with 128GB/s of bandwidth. Aside from that, we don’t know much more at this time, though earlier reports have suggested that Zen-based APUs won’t arrive until 2017, following the release of desktop Zen CPUs in the fourth quarter of 2016.

http://www.tweaktown.com/news/54721/amds-next-gen-apu-zen-cpu-cores-vega-gpu-hbm2/index.htmlhttp://www.tweaktown.com/news/54721/amds-next-gen-apu-zen-cpu-cores-vega-gpu-hbm2/index.html

AMD has an exciting year ahead of itself in both the CPU and GPU divisions, with their next-gen Vega GPU architecture set to be unveiled soon, and their upcoming Raven Ridge APU that has some surprising tech inside of it. amds-next-gen-apu-zen-cpu-cores-vega-gpu-hbm2_01 According to the latest report from Bitsandchips.it, AMD is working on a few Raven Ridge APUs - with the higher-end model featuring a 1024-core Vega GPU with HBM2, 16 CUs, DDR4 support, with Zen CPU cores (4 x CPU cores, 4 x - and it all arrives in TDPs of between 35-95W.

The lower-end version loses the HBM2 technology and kicks the Vega GPU from 16CUs on the high-end model, to 12 CUs on the lower-end model. The 170mm² part will have a TDP of 4-35W, with both versions of the Raven Ridge APUs being made on the 14nm FinFET process.

 

[Rifter]

I just hope when they do it they do it right, like make sure it can game at 1080p at a solid 60fps. This will sell like hotcakes.

 

[frozentundra123456]

https://www.top500.org/news/pondering-amds-ambitions-for-high-performance-apus/



http://www.kitguru.net/components/a...n/amd-said-to-be-working-on-zen-apu-with-hbm/




http://www.tweaktown.com/news/54721/amds-next-gen-apu-zen-cpu-cores-vega-gpu-hbm2/index.html

First quote sounds pretty vague "we do believe we can do it" is a far cry from saying they have or are going to do it for the consumer market.
Second quote, the ever reliable WCCF tech.
Third quote sounds more definite. lets see if it pan out.

 

[cbn]

First quote sounds pretty vague "we do believe we can do it" is a far cry from saying they have or are going to do it for the consumer market.
Second quote, the ever reliable WCCF tech.

FWIW, Here was an article from 2015 (Yes I know its very old and from WCCF tech/Fudzilla) with a claim about the Server APU using Zeppelin MCM to a Vega 10 dGPU with HBM2 on an interposer:

]

So if this is true, this high power Server APU will not need its own special die. It would just be a combination of off the shelf parts already being used for lesser products MCM and connected via fabric. Therefore I would think a high power laptop APU could be made the same way with the difference being different binning (eg, harvested and downclocked Zeppelin and Vega 10 dies, etc rather than the full power versions)

P.S. I suspect what we are seeing above will not be far from the truth (re: AMD has already revealed that the Naples 32C/64T Server CPUs will be made up of four Zeppelin dies MCM ......so that Server APU scheme doesn't seem much a stretch to me)

While not specifically mentioned in the announcement today, we do know that Naples is not a single monolithic die on the order of 500mm2 or up. Naples uses four of AMD’s Zeppelin dies (the Ryzen dies) in a single package. With each Zeppelin die coming in at 195.2mm2, if it were a monolithic die, that means a total of 780mm2 of silicon, and around 19.2 billion transistors – which is far bigger than anything Global Foundries has ever produced, let alone tried at 14nm. During our interview with Dr. Su, we postulated that multi-die packages would be the way forward on future process nodes given the difficulty of creating these large imposing dies, and the response from Dr. Su indicated that this was a prominent direction to go in.

 

[w3rd]

See^ at least somebody knows where this is all going. (ie: HSA = fabric)

I'm glad I am not the only one whos been listening an watching AMD stuff happening, over the years. The writing been on the wall for a long time & Dr Su knows where all of this is going..

And it isn't going to be a Ghz war..




I know a thing to two, because I've seen a thing or two...

 

[LightningZ71]

So, then, it looks like there are two products planned for four different market segments:

1) Zen based APU with fewer Vega CUs and no HBM. Single channel DDR4 enabled platform for bottom end of notebook market. Also for budget ultrathin convertible 2 in 1s and tablets.

2)Zen based APU with fewer Vega CUs and no HBM. Dual channel DDR4 enabled platform, potentially with additional x8PCIe dgpu on board. Middle of market for budget enthusiast, budget power user, etc. There is potential that there could be a high CU count part here too, but without HBM, it would be quite handicapped (could be recovered high end parts with bad HBM module) includes higher spec 2 in 1s and ultrathins.

3) Zen based APU with max Vega CUs and with HBM, dual channel DDR4 platform. High end ultrathins, 2 in 1s, tablets, space limited performance notebooks, etc.

4) Zen based APU with max Vega CUs and HBM, dual channel DDR4, and an x8 PCIE connected dgpu in crossfire/SLI for maximum performance laptops.

A cost effective solution for every segment using only 2 dies. I wonder if AMD has a specific dgpu part in mind for this? Perhaps a 16CU, HBM enabled VEGA unit with a 64 bit GDDR5 off chip memory channel? That would make it mostly comparable to the iGPU for the purposes of performance targeting.