When will AMD implement multi-treading?

OICAspork

Newcomer
As I understand it, AMD has patents for multi-threading. At this point they are taking the performance crown from Intel in all but hyper-threading enabled applications... Does anyone know AMD's plans for implementing multiple-thread processing ability? Is it on any of their roadmaps yet?
 
OICAspork said:
As I understand it, AMD has patents for multi-threading. At this point they are taking the performance crown from Intel in all but hyper-threading enabled applications... Does anyone know AMD's plans for implementing multiple-thread processing ability? Is it on any of their roadmaps yet?

i saw 2 reviews about the new AMD FX-53 .. it's a monster ! the P4 3.4EE has some problems to compete with it :oops:

RainZ
 
rainz said:
OICAspork said:
As I understand it, AMD has patents for multi-threading. At this point they are taking the performance crown from Intel in all but hyper-threading enabled applications... Does anyone know AMD's plans for implementing multiple-thread processing ability? Is it on any of their roadmaps yet?

i saw 2 reviews about the new AMD FX-53 .. it's a monster ! the P4 3.4EE has some problems to compete with it :oops:

RainZ
Thats, true. but 2 reviews i saw had also asked for hyperthreading from amd. They were doing encoding and using the computer for other purposes, in this type of test amd lost.

I hope amd adds this feature soon. ;)

later,
epic
 
well i hope intel adds x86-64 bit .


Hopefully soon :)

Intel has its buzz word and amd has its buzz word. Both cross liscense so they can both use the tech. But they are both happy when each other are out of the red so intel can't be a monoply and amd wont be crushed.

When new buzz words come out like sse 3 then I'm sure amd will get ht in the core following intel getting sse 3
 
Intel has already stated that they're going to extend their x86 to 64bits and be compatibile with x86-64. I don't have a link to the article but I think it was as Ace's.
 
Saem said:
Intel has already stated that they're going to extend their x86 to 64bits and be compatibile with x86-64. I don't have a link to the article but I think it was as Ace's.

yes we all knew it would happen at some time. What i'm saying is at everypoint since the k6-2 amd would have a fancy new thing and intel would have a fancy new thing. Then they would both ge tnew fancy things and they would take each others old fancy thing .

Intel had mmx , amd 3d now. intel sse which had some of 3d now in it. Amd 3dnow + and mmx , intel p4 with sse 2 , amd athlon xp with sse .
amd hammer with sse 2 and x86-64 , intel p4 with ht. P4 with x86-64 , amd ht .

Works
 
Tahir said:
At the moment AMD see HT as a disadvantage. Heh.
Yes but for how much longer. Id love to do a couple of cpu intensive tasks in the future. ;) Encoding porn takes too much time from surfing for more porn. ;)

later,
epic
 
1. What are their plans for moving beyond single-core/single-threaded designs? CMP? CMT? SMT?


AMD told me there are no "real" ("looking into it") plans for SMT or Hyperthreading, but it was clear to me (although the spokesman didn't confirm) a dual core CPU is coming in the future. The most important platform improvement for Quad/Dual Opteron systems will be a faster HT link between CPUs, which will produce tangible performance increases. Dave Everitt confirmed that the performance increase due to this faster Hypertransport link will be better than just a few percentages.
Other sources say that a "Coherent" and faster clocked Hypertransport should be ready in the beginning of Q3.

Source
 
epicstruggle said:
Thats, true. but 2 reviews i saw had also asked for hyperthreading from amd. They were doing encoding and using the computer for other purposes, in this type of test amd lost.

I hope amd adds this feature soon. ;)

later,
epic

On the other hand, going to a 64 bit OS will help the Athlon 64 a lot when it comes to encoding/decoding.
http://www.anandtech.com/systems/showdoc.html?i=1961&p=4

It gains 15.5 %. And then the software wasn't even written for 64 bit operation. At least not according to Anands.
The gaming performance in that preview is old news though. I've seen newer ones where the difference was minimal for 32 bit apps.

Div X encoding:
http://www.aceshardware.com/read.jsp?id=60000256

Also gains around 15%.
 
Peronally I would prefer AMD to create a scalar x86 core (apart from the FPU/SIMD pipeline, which can remain tripple issue, just keep it in-order) and put as many of those on a chip as they can manage ... add the ability to have lots more outstanding prefetches too and I wouldnt mind if it didnt have MT.
 
Not sure how good the idea is; way too much software still relies too heavily on single-thread performance, and to achieve good performance, you still need rather deep pipelines or *very* many cores. A 486-class core (1 instruction per clock as theoretical maximum) @ 1.2M transistors, IIRC, maxed out at ~133 MHz at 0.5 micron, which would scale to only about 500 MHz at 0.13, and given the design of the 486, I would estimate it to have ~1/3 of the IPC of an AthlonFX, so you would need 15+ cores to match a 2.4Ghz AthlonFX - I am not convinced that such a setup would actually require less logic than the AthlonFX.
 
You can make em run at the same clock as superscalar cores, they are just a lot smaller and stall a little more ... the non SIMD part is just there for housekeeping and to run windows&office anyway.

I dont expect them to do it :) I would just like them to do it.
 
sorry for the ignorance, but isnt part of the delay in multi-core cpu's, that microsoft, might tell you that you need 2 licenses to use their os?

later,
epic
 
MfA said:
You can make em run at the same clock as superscalar cores, they are just a lot smaller and stall a little more ... the non SIMD part is just there for housekeeping and to run windows&office anyway.
Yeah, but making them run at higher clock speeds means inserting extra pipeline steps, costing lots of extra transistors; a scalar core capable of >2GHz at 0.13u will be a lot larger than the 486-class cores of my example, and stall a lot more. It might be easy to introduce switch-on-event mulththreading in such a core, though.
I dont expect them to do it :) I would just like them to do it.
It would be interesting; if the software support was there, it would be a matter of finding the best tradeoff of core complexity versus number of cores, which may or may not give useful results.
 
epicstruggle said:
sorry for the ignorance, but isnt part of the delay in multi-core cpu's, that microsoft, might tell you that you need 2 licenses to use their os?

later,
epic

well Win 2000 Pro and Win XP pro both have support for 2 CPU's so I guess from that standpoint only the "home"version users would suffer.
 
Druga Runda said:
epicstruggle said:
sorry for the ignorance, but isnt part of the delay in multi-core cpu's, that microsoft, might tell you that you need 2 licenses to use their os?

later,
epic

well Win 2000 Pro and Win XP pro both have support for 2 CPU's so I guess from that standpoint only the "home"version users would suffer.

then again HT works in XP Home right?
I'm sure MS would play nice and provide an upgrade patch to let it happen
 
The plan is to implement it in Xeons first, and then bring it to the desktop.

AMD's not interested in hyperthreading because its effectiveness is directly proportional to the inefficiency of the processor. It only works if your processor can't keep the pipes full.
 
Back
Top