AMD Radeon finally back into laptops?

D

Deleted member 13524

Guest
In this presentation I finally saw an announced laptop carrying a Polaris chip:

5BxVBOW.png


Judging by the picture, it's the next generation of the HP Omen, a relatively high-end gaming laptop (as in quite expensive because of premium looks and materials, not very high performance) which currently has a GTX 960M GPU.
http://www8.hp.com/us/en/campaigns/hp-omen/overview.html
http://www.notebookcheck.net/HP-Omen-Notebook-Review.134844.0.html

Seems like a high-profile design, though nothing compared to say a Macbook Pro refresh.

Polaris 11 is predictably going after GM107 models. I wonder what the performance boost will be comparing a full P11 to a full GM107, considering the latter is a first-gen Maxwell that's been pretty much unchanged for over 2 years.
 
Has anyone tried to get a sense of how the RX 470 might perform and how power-efficient it might be by playing with WattMan on the RX 480?
 
Has anyone tried to get a sense of how the RX 470 might perform and how power-efficient it might be by playing with WattMan on the RX 480?


In a reddit AMA, Robert Hallock has stated that while the RX480 gets 2-2.3x the performance/watt of Hawaii, the RX470 gets 2.8x. That's up to 40% better performance/watt than the RX 480, on a desktop product.
This means the absolute performance of Polaris 10 is definitely better in the RX 470 iteration for the power/performance curve.

Also, there are some weird things about the RX 480 though, like an unnecessarily high vcore and more 12V phases than it can chew (perhaps AMD wanted the OEMs to shine with their custom versions?). People lowering the vcore are probably getting better power consumption numbers, but I haven't seen any power measurements with the undervolting without overclocking.
 
Yeah I've seen those claims from AMD, which is why I was wondering whether anyone had put them to the test. Of course, the RX 470 might be binned differently (it probably is) but you could get a pretty good idea of the validity of AMD's claims by underclocking and undervolting an RX 480 down to about 110W.
 
AMD's claims of the RX 480m claims make me believe them. Fake claims about laptop parts would seem odd. AMD should have same board power rating for OEMs as they say to public for them. And being truthful there, or slightly overestimating is better right?
1000-1100Mhz clockspeed, RAM at effective 5-6GBps. See how much a Rx480 draws than. Cut in half about for Rx480m.
 
The most interesting thing was that the laptop parts are projected to be fantastically efficient. As in doubling the efficiency of previous generation Nvidia chips, which is so far better than what Nvidia's current chips have done by a solid amount.

It makes it weird, considering the RX 480 is so inefficient, but it hass hit the TDP and performance figures AMD stated (with the usual little PR fudges). So I see no reason these chips wouldn't be as potentially excellent as claimed. No wonder Apple gave the contract to AMD.
 
The most interesting thing was that the laptop parts are projected to be fantastically efficient. As in doubling the efficiency of previous generation Nvidia chips, which is so far better than what Nvidia's current chips have done by a solid amount.

It makes it weird, considering the RX 480 is so inefficient, but it hass hit the TDP and performance figures AMD stated (with the usual little PR fudges). So I see no reason these chips wouldn't be as potentially excellent as claimed. No wonder Apple gave the contract to AMD.

nV doesn't really care for Apple as they tend to squeeze margins out of suppliers. This can be seen for the last 3 years of Apple products where nV was really no where to be found even when they had better perf/watt in every segment. Also Apple is a big proponent of open CL, which AMD products tend to better in and Apple software is built for open CL acceleration. I wouldn't read too much into what Apple sells with their systems cause that doesn't' tell use anything about what other OEM's are thinking.
 
The most interesting thing was that the laptop parts are projected to be fantastically efficient. As in doubling the efficiency of previous generation Nvidia chips, which is so far better than what Nvidia's current chips have done by a solid amount.

It makes it weird, considering the RX 480 is so inefficient, but it hass hit the TDP and performance figures AMD stated (with the usual little PR fudges). So I see no reason these chips wouldn't be as potentially excellent as claimed. No wonder Apple gave the contract to AMD.

Well, so far, Pascal parts have been clocked very high (compared to usual GPU clocks). It's hard to quantify without knowing what the voltage/frequency curve looks like for this particular architecture on this particular process, but still, at lower clocks it should be more efficient.
 
nV doesn't really care for Apple as they tend to squeeze margins out of suppliers. This can be seen for the last 3 years of Apple products where nV was really no where to be found even when they had better perf/watt in every segment. Also Apple is a big proponent of open CL, which AMD products tend to better in and Apple software is built for open CL acceleration. I wouldn't read too much into what Apple sells with their systems cause that doesn't' tell use anything about what other OEM's are thinking.

Indeed, nVIDIA left a sour taste in the mouth when they decided to neglect OpenCL. However, you would have to be insane to turn down millions in revenue, low margins or not.
 
Indeed, nVIDIA left a sour taste in the mouth when they decided to neglect OpenCL. However, you would have to be insane to turn down millions in revenue, low margins or not.

Depends on how Apple treated them, millions don't matter when you make hundreds of millions....

Have we seen Open CL come close to what CUDA has done for the industry so far? Open CL hasn't made headroom yet either outside of a few applications, notably Adobe and Apple stuff. The amount of money losing something than putting in the work and effort and not gaining enough from that work and effort has to be weighed.

The discrete GPU portion of Apple is small, most people don't by Apple products to use final cut or do work with the adobe suite and definitely not gaming lol.
 
Depends on how Apple treated them, millions don't matter when you make hundreds of millions....

Have we seen Open CL come close to what CUDA has done for the industry so far? Open CL hasn't made headroom yet either outside of a few applications, notably Adobe and Apple stuff. The amount of money losing something than putting in the work and effort and not gaining enough from that work and effort has to be weighed.

The discrete GPU portion of Apple is small, most people don't by Apple products to use final cut or do work with the adobe suite and definitely not gaming lol.

Not to argue but they still sell millions of computers, over 20 million last year alone. They also ranked 5th last year in marketshare, according to Gartner.
 
Depends on how Apple treated them, millions don't matter when you make hundreds of millions....

Have we seen Open CL come close to what CUDA has done for the industry so far? Open CL hasn't made headroom yet either outside of a few applications, notably Adobe and Apple stuff. The amount of money losing something than putting in the work and effort and not gaining enough from that work and effort has to be weighed.

The discrete GPU portion of Apple is small, most people don't by Apple products to use final cut or do work with the adobe suite and definitely not gaming lol.
I understand Nvidia getting out of Mac Pro to protect high margin sales of professional cards. I do not understand why not competing in laptops and for new iMacs is not okay. Maybe 1-2 million GPU units a quarter. I guess Nvidia on it's quest for higher margins decided sacrificing income is fine.
 
You really think its 1-2 million DISCRETE gpu's per quarter Apple sells in their notbooks? I think its a hell of a lot less then that. Its simple if AMD had 1-2 million DISCRETE GPU sales per quarter you think their notebook discreet would be in the toilet for the past 2 years? They should be well ahead of nV if that was the case.
 
Last edited:
You really think its 1-2 million DISCREET gpu's per quarter Apple sells in their notbooks? I think its a hell of a lot less then that. Its simple if AMD had 1-2 million DISCREET GPU sales per quarter you think their notebook discreet would be in the toilet for the past 2 years? They should be well ahead of nV if that was the case.
mis-remembered Apple's quarterly sales. So 500k GPUs. if 10% of sales are iMacs with dGPUs or rMBP 15 with dGPU. I think 10% is fair mix.

As Mac ASP is about 1200-1300 dollars.
 
500 k yeah that sounds good, so it ends up not being much at all. These will be the lowest margin units. Low end laptops GPU's
 
500 k yeah that sounds good, so it ends up not being much at all. These will be the lowest margin units. Low end laptops GPU's
FWIW those aren't really "low end" notebook GPUs. If you want to see low-end notebook gpus, look at something like gm108 (or one of the three hundred variants using mars/sun/iceland on the amd side, albeit they are nowadays mostly exclusive on notebooks with amd apus). If I look at notebook offers, I certainly get the impression gm108 (useful or not as it's barely if at all faster than the IGPs it is sold with) outsells all other nvidia chips combined (by a huge margin at that)...
 
Total discrete sales are like what 15 million units per quarter? We are looking at 20% of that for all of AMD cards, that's 3 million. Out of that, Apple is getting around 500 thousand sales, for AMD that is a huge chuck, for nV not so much.
 
Last edited:
Back
Top