Or in general not to cede that part of the server market to ARM without a fight. This apparently draws a lot more power than the A1100 will, but I get the strong impression that for many workloads Intel still beats ARM in performance/watt
I'll note these are on Intel's latest and greatest 14 nm process, vs. the 28 nm Global Foundries process for AMD; don't know how they directly compare, especially since Global Foundries is now at arms's length from AMD so to speak, which makes it less likely they're going to be able to push the margins of the process, but that further suggests Intel will benefit from performance/watt. Add in the maturity of x86-64 toolchains, ironically gifted to Intel by AMD (sure wish AMD had kept their eye on the ball there), vs. the brand new 64 bit ARM architecture, and AMD and company have a difficult task ahead of themselves when power isn't an issue for batteries but for machine rooms.
AMD should be running with Samsung's/GF's 14nm FinFET process. Even if it's not quite as good as Intel's, at least it puts them in the general area of competitiveness. At 28nm, AMD has no chance, no matter what market we're talking about. This is why I'm a strong believer in AMD getting acquired by a rich company that is committed to take on Intel in the chip business. AMD will never get anywhere on its own. It's just experiencing a slow-motion death. It's getting less and less money, which means it will have less and less money for research and high-volume production, and it's going to fall further and further behind Intel to the point where nobody can save it anymore.
I'd rather AMD was acquired by someone like Samsung or Qualcomm (which already bought AMD's mobile GPU unit) and "lose its independence" or whatever nostalgic issue we're talking about, than see it die in slow-motion over the next 5-10 years with its "independence" intact.
Outside of an asteroid falling from the sky and destroying Intel's foundries, I just see no way for AMD to become competitive against Intel again without spending billions of dollars in research, marketing and partner deals over the next 3 years. If they get those billions now, they might actually have a chance in 3+ years from now.
There's something in the AMD culture, for sure. A long time ago, like +- their very fast 486s (https://en.wikipedia.org/wiki/Am486), someone noted that over it's history AMD had never made stockholders money. Significant successes were not followed up, in a pattern that we see with x86-64.
One possible fly in that ointment: how are AMD's ... arrangements with Intel structured? AMD is playing some of these games because Intel chose them to second source the 8086, 8088, and 286 and that resulted in their getting various rights to future chips (https://en.wikipedia.org/wiki/Advanced_Micro_Devices#IBM_PC_...). If these don't hold if AMD is acquired, it wouldn't likely be attractive. Or companies that only want to play the ARM game, might not be interested in joining all the drama of AMD, including the remaining ATI graphics stuff, which are irrelevant to server chips.
Are you sure ARM at 28nm is that losing of a proposition for servers? Don't ARM designs have quite a bit of power margin to waste? Although that of course doesn't help the performance/power ratio.
Actually, we're at a point where graphics stuff is becoming more relevant to server chips. Especially with more parallelizable workloads being pushed to GPUs for processing, it's interesting to see where the distribution of CPU/GPU focused tasks goes.
Of particular note in the world of data analytics, there are some exciting things on the horizon for using GPU cores to accelerate database workloads[0][1]. Of particular interest here is removing the communication overhead between CPU and GPU and avoiding the memory translations necessary to move between GPU VRAM to system RAM.
AMD's success in the iGPU market and their HSA design (along with supporting APIs) could be enough to win AMD some market share.
In reply to those pointing out non-graphics use of GPUs, ah, yeah. I don't follow that, so I didn't know if AMD/ATI was still/ever? a player, nVidia gets what little press I see of that.
Aren't the economics of this are based on gamers buying large quantities of the chips, thereby spreading out the non-recurring engineering (NRE) costs of making the chips? Is AMD/ATI doing well in this separate from CPUs market? All I can remember is people still complaining about poor device drivers.
AMD is still quite competitive in the GPU market, but their major progress lately is in their HSA[0]. A big problem with GPU computing is that you have two distinct memory spaces, so whatever is being done on the GPU needs to be loaded into VRAM, worked upon, and then returned to system RAM.
The goal of HSA is to break this down. If you have GPU and CPU on one chip with a single pool of RAM that both can access a lot of communications complexity and overhead goes away.
All of this is still very new tech, so what I am saying is conjecture and possibility, but there is a lot of potential in this area.
I am not able to say whether the magnitude of this is similar, but you might think of it analogously to AMD64 architecture. AMD was first to the table with 64 bit processors. They are also first to the table with this. Intel may catch up and blow them away, as has now happened with 64 bit processors, but right now AMDs integrated graphics far beyond Intel's offerings.
We must not forget AMD isn't the only ARM64 player... HP and IBM are rumored to be working on chips as well as all the normal ARM players (Qualcomm, Broadcom, Samsung, etc...).
This one particular chip may fair better than the very first to market ARM64 server chip, but future iterations made by 5-7+ companies in competition with one company are likely to succeed.
It has seemed for a while Intel is in a dire competition for the low power market; something they consistently fall short of regardless of usually having better performance marks (Atom).
Yes, there are many arm licenses out there. But relatively few fabs to make them. Even less that are interested in relatively low volume markets like competing for servers, er, lowend servers, er, lowend non-intel non x86 servers.
Not exactly a huge market compared to smartphones, tablets, and related mobile devices.
The core count is good. My suspicions is that this could greatly reduce compute intensive costs in data centres. Thus bringing down cost per hour. Sweet.
Tested HLE quickly once and didn't get speed-ups. Maybe I was using it incorrectly or not in the right situation.
I'd be interested if anyone succeeded with getting speed-ups from HLE or TSX.
15 comments
[ 3.9 ms ] story [ 44.7 ms ] threadI'll note these are on Intel's latest and greatest 14 nm process, vs. the 28 nm Global Foundries process for AMD; don't know how they directly compare, especially since Global Foundries is now at arms's length from AMD so to speak, which makes it less likely they're going to be able to push the margins of the process, but that further suggests Intel will benefit from performance/watt. Add in the maturity of x86-64 toolchains, ironically gifted to Intel by AMD (sure wish AMD had kept their eye on the ball there), vs. the brand new 64 bit ARM architecture, and AMD and company have a difficult task ahead of themselves when power isn't an issue for batteries but for machine rooms.
I'd rather AMD was acquired by someone like Samsung or Qualcomm (which already bought AMD's mobile GPU unit) and "lose its independence" or whatever nostalgic issue we're talking about, than see it die in slow-motion over the next 5-10 years with its "independence" intact.
Outside of an asteroid falling from the sky and destroying Intel's foundries, I just see no way for AMD to become competitive against Intel again without spending billions of dollars in research, marketing and partner deals over the next 3 years. If they get those billions now, they might actually have a chance in 3+ years from now.
One possible fly in that ointment: how are AMD's ... arrangements with Intel structured? AMD is playing some of these games because Intel chose them to second source the 8086, 8088, and 286 and that resulted in their getting various rights to future chips (https://en.wikipedia.org/wiki/Advanced_Micro_Devices#IBM_PC_...). If these don't hold if AMD is acquired, it wouldn't likely be attractive. Or companies that only want to play the ARM game, might not be interested in joining all the drama of AMD, including the remaining ATI graphics stuff, which are irrelevant to server chips.
Are you sure ARM at 28nm is that losing of a proposition for servers? Don't ARM designs have quite a bit of power margin to waste? Although that of course doesn't help the performance/power ratio.
Of particular note in the world of data analytics, there are some exciting things on the horizon for using GPU cores to accelerate database workloads[0][1]. Of particular interest here is removing the communication overhead between CPU and GPU and avoiding the memory translations necessary to move between GPU VRAM to system RAM.
AMD's success in the iGPU market and their HSA design (along with supporting APIs) could be enough to win AMD some market share.
[0]https://news.ycombinator.com/item?id=5595738
[1]https://news.ycombinator.com/item?id=8787414
It might have been irrelevant a few years ago, but OpenCL and GPU computing is showing up in more and more server applications.
Aren't the economics of this are based on gamers buying large quantities of the chips, thereby spreading out the non-recurring engineering (NRE) costs of making the chips? Is AMD/ATI doing well in this separate from CPUs market? All I can remember is people still complaining about poor device drivers.
The goal of HSA is to break this down. If you have GPU and CPU on one chip with a single pool of RAM that both can access a lot of communications complexity and overhead goes away.
All of this is still very new tech, so what I am saying is conjecture and possibility, but there is a lot of potential in this area.
I am not able to say whether the magnitude of this is similar, but you might think of it analogously to AMD64 architecture. AMD was first to the table with 64 bit processors. They are also first to the table with this. Intel may catch up and blow them away, as has now happened with 64 bit processors, but right now AMDs integrated graphics far beyond Intel's offerings.
[0]http://developer.amd.com/resources/heterogeneous-computing/w...
This one particular chip may fair better than the very first to market ARM64 server chip, but future iterations made by 5-7+ companies in competition with one company are likely to succeed.
It has seemed for a while Intel is in a dire competition for the low power market; something they consistently fall short of regardless of usually having better performance marks (Atom).
Not exactly a huge market compared to smartphones, tablets, and related mobile devices.
I am curious to see if this gets adopted into mainstream.