AMD shows off the guts of its first ARM server chip

The most interesting tidbit I got out of that is that AMD has a memory controller capable of using both DDR3 and DDR4.
 
these are supposed to be socket compatible with x86 processors, right? but i'm guessing this is bga mounted.

is there any talk of what the new sockets will look like?

also--- regarding the ddr4 controllers. might amd actually have a ddr3,4,5 controller, considering this one and the one on the Xbox/PS4 processors?
 
No such thing as DDR5 yet. The PS4 uses GDDR5, which is derived from DDR3
 
Can't wait to build a box with one - hoping the price point comes down considerably.
 
Seattle-Reference-System-Board.jpg


shutupandtakemymoney.jpg
 
Does anyone know why AMD is not trying to tackle the consumer arm chip market?


qualcomm, samsung, nvidia... and that'd kind of it for arm right now. Are there too many large players for them to make a dent? Do they have some sort of constraints keeping them from producing their own consumer arm chips? (like some prohibition from qualcomm since amd RADEON graphics were sold to qualcomm and renamed ADRENO)

This seems like a way to sell a lot more chips, so why are they not trying to compete in that market? Not enough cash?
 
Opteron series always comes out a year before the FX desktop chips. Be patient little grasshopper.

:p
 
So how will it perform if it is pitted against intel processors?

It's at least as strong as any ARM. And an ARM can move a Sandy Bridge with a Bulldozer. I've seen an ARM fight at Gulftown and win. And an ARM scored higher than some Ivy league processors. An ARM waved at Lewis and Clark across a Field and lived to tell the tale. A single pen stroke from an ARM could shut down Roswe.. err Haswell.

The only thing an ARM might have a problem with would be the Devil. He lives down in a Canyon and it would take awhile to climb back out. Hell even Nvidia thinks ARMs are so awesome they are attaching some to their new GPUs.
 
Does anyone know why AMD is not trying to tackle the consumer arm chip market?


qualcomm, samsung, nvidia... and that'd kind of it for arm right now. Are there too many large players for them to make a dent? Do they have some sort of constraints keeping them from producing their own consumer arm chips? (like some prohibition from qualcomm since amd RADEON graphics were sold to qualcomm and renamed ADRENO)

This seems like a way to sell a lot more chips, so why are they not trying to compete in that market? Not enough cash?

If AMD customers aren't in that segment then why should they it has been dominated by designs which tend to be well ahead and the server market is a business segment AMD knows.

Where ARM has a good lead in flexibility and that is the reason that AMD went for ARM.

.
 
Does anyone know why AMD is not trying to tackle the consumer arm chip market?
The biggest ARM chip makers are vertically integrated: Samsung makes and uses its own chips, Apple does likewise, and there are a couple other large, established makers which not only modify the ARM core, but also include their own IP (Qualcomm, for example). Then there are dozens of smaller players (like AllWinner) who work on very thin margins, selling SoCs generally for under $15, sometimes much less.

AMD is using a stock licensed A57 core (and embedded A5 core) in its server chip, customizing the other parts of the SoC for server use. Apple is on its own, but it would be hard to challenge Qualcomm or Samsung, both of which make higher performance chips than the stock designs they are based on, with a stock licensed core.
 
The biggest ARM chip makers are vertically integrated: Samsung makes and uses its own chips, Apple does likewise, and there are a couple other large, established makers which not only modify the ARM core, but also include their own IP (Qualcomm, for example). Then there are dozens of smaller players (like AllWinner) who work on very thin margins, selling SoCs generally for under $15, sometimes much less.

AMD is using a stock licensed A57 core (and embedded A5 core) in its server chip, customizing the other parts of the SoC for server use. Apple is on its own, but it would be hard to challenge Qualcomm or Samsung, both of which make higher performance chips than the stock designs they are based on, with a stock licensed core.

The obvious counter example is nvidia. They don't have vertically integrated fabs, I think they use TSMC for their k1 chips. But they still seem to be able to come out with competitive parts.
 
So does this possibly mean there is a future of AMD-based fanless desktop computers that don't run too hot?
 
I recall some comparison of earlier ARM server processors,meant to "take over" the web server role from anyone else in clusters in data centeres, but in the end a single Xeon basically kicked the crap out of a load of ARM CPU's once you actually put any real load on the websites....
 
I recall some comparison of earlier ARM server processors,meant to "take over" the web server role from anyone else in clusters in data centeres, but in the end a single Xeon basically kicked the crap out of a load of ARM CPU's once you actually put any real load on the websites....

Exactly, it just wasn't feasible, which is why I really believe the Calxeda servers went down the tube.
ARM makes sense for mobile and ultra low-power, but when it comes to price/performance/watt, x86 is king.
 
When i had first read about ARM server farms 1-2 years ago? and the low power and heat it really had me interested but the performance just wasn't there for anything beyond a hobby website.
 
The obvious counter example is nvidia. They don't have vertically integrated fabs, I think they use TSMC for their k1 chips. But they still seem to be able to come out with competitive parts.

Actually they are not competitive at all. They allowed Anandtech to post selective results which never gave out power and performance either one or the other.

In reality you won't see them replacing snapdragon, they do end up in designs maybe because they paid for it.

There are only a few really great ARM chips for mobile out and those tend to be Qualcomm made.
 
Actually they are not competitive at all. They allowed Anandtech to post selective results which never gave out power and performance either one or the other.
Anand is free to purchase a Jetson TK1 board and run whatever battery of tests he wishes. As are you.
 
Of course, the Shield Tablet is also designed for gaming. Unfortunately, the Tegra K1 introduces quite a massive amount of dynamic range. While it’s fully possible for the Shield tablet to last 10 hours of continuous use on a single charge, running the GPU at full blast gives battery life similar to a gaming laptop. Realistically, if a game is made for Tegra K1 and truly stretches the GPU to the limit, battery life is only around two and a half hours, assuming display brightness is kept down to 200 nits.

very competitive :)
 
Back
Top