AMD's response

[darkdemyze]

In opposition to some claims I have been reading, seems that AMD already has a response to Intel's pre-lived "victory" if you will.

Part 1 of 5
AMD is keeping it a secret, but this just shows they may not be as far behind as some people think.

Part 2 of 5
Seems like just more speculation as the interview shifts toward the server market.

Part 3 of 5
AMD Live! coverage.

Part 4 of 5
Enterprise computing.

Part 5 of 5
Business relationships.

 

[BrownTown]

The problem is that they are jsut updating the old core next year whereas Intel has a compleyelt new core, so the improvements AMD gets won't be nearly as signifigant as Prescott to Conroe will be.

 

[Bobthelost]

Reads as: Nothing really stunning in the near future, incremental improvements but that's all. Without seeing how the retail AM2 performs saying that'll be enough or not to compete with retail Conroe would be foolish.

 

[peleejosh]

Originally posted by: BrownTown
The problem is that they are jsut updating the old core next year whereas Intel has a compleyelt new core, so the improvements AMD gets won't be nearly as signifigant as Prescott to Conroe will be.

But you are overlooking the fact that prescott sucked so bad, and thats the only reason that the improvement to conroe is so significant.

 

[Acanthus]

Originally posted by: Bobthelost
Reads as: Nothing really stunning in the near future, incremental improvements but that's all. Without seeing how the retail AM2 performs saying that'll be enough or not to compete with retail Conroe would be foolish.

Thats what i read, pretty much "OMG WTF DDRII... etc"

Intel showed their hand, just because AMD is quiet doesnt mean they have pocket bullets

 

[Jules]

Originally posted by: BrownTown
The problem is that they are jsut updating the old core next year whereas Intel has a compleyelt new core, so the improvements AMD gets won't be nearly as signifigant as Prescott to Conroe will be.

And the good thing about it is that you dont know jack what there doing. lol.

 

[Duvie]

A: Well, I acknowledge or I recognize that we are disclosing less of our future plans than people would like us to, and part of that is because of the competitive nature of the industry. It doesn't help us to give too much heads up to the competition. That said, we have a very solid architecture, one that's scaling well. It's scaling well in terms of multicores, and it's scaling even better in terms of multi-sockets. That's excellent because clearly the industry is going towards multicore designs, and although the software part of it will always trail the hardware, that move is resulting in greater parallelization of applications.

You've picked up on an 8KL, so you could say it seems as though AMD is tending to follow more of an evolutionary design path than a revolutionary one, and I think that's a correct view of where we're taking the technology in 2007. We have a good technology base, so it's easy for us to expand it, without having to re-invent the wheel, unlike Intel, which has to try to claim that the NGMA is something new.

We don't need something new, we just need to continue down the path of improving memory bandwidth through new technologies, improving speed, through faster clock rates, larger caches, and, in certain areas that are driven by customers who demand improvement, absolute performance. It's clear that we will improve both the integer and floating-point performance of our cores, independently of improvements in transistor technology, manufacturing technology and so forth.

I think this part gives me even more of an idea that in fact quad cores may come sooner and be the response...What is more evolutionary to the current A64/opteron design then scalability of cores...i think it may help further to sum up why ther has been so many rumors of increased memory controller speeds...DDR2-667 to DDR2-800 to DDR2 1066...especially when we all know the AMD is not bandwidth limited now...it only goes to assume they would only increase bandwidth if and when it is needed....

Now that being said it is not certain listening to hi response on performance crown and how that translates into all the market segement. They may concede a victory in desktop while they stay ahead on the server and enterprise market...so why I say quad cores, more cache, increases memory controllers...it may very well be an opteron only market...

 

[BrownTown]

Originally posted by: MyStupidMouth

Originally posted by: BrownTown
The problem is that they are jsut updating the old core next year whereas Intel has a compleyelt new core, so the improvements AMD gets won't be nearly as signifigant as Prescott to Conroe will be.

And the good thing about it is that you dont know jack what there doing. lol.

there have been several articles about AMDs core improvements on the Inq which have been at least partly verified by other sites, so it isn't true to say that we know jack. First off all the new stuff isn't till 2007, so its well after Conroe initially launches, secondly, everything so far points towards small core revisions, not towards anything signifigant in terms of performance boost.

 

[Duvie]

Originally posted by: BrownTown

Originally posted by: MyStupidMouth

Originally posted by: BrownTown
The problem is that they are jsut updating the old core next year whereas Intel has a compleyelt new core, so the improvements AMD gets won't be nearly as signifigant as Prescott to Conroe will be.

And the good thing about it is that you dont know jack what there doing. lol.

there have been several articles about AMDs core improvements on the Inq which have been at least partly verified by other sites, so it isn't true to say that we know jack. First off all the new stuff isn't till 2007, so its well after Conroe initially launches, secondly, everything so far points towards small core revisions, not towards anything signifigant in terms of performance boost.

I would say adding cores is not a small revision and he definietly hints that is in the future....Added cache in "the right markets" in a small tweak....added fpu units may seem liek a small tweak because as was suggested the K8 architecture is so good and its evopltuionary path has been known since conception....

I would say a quad core does nothing in the desktop market as parallelization means little to most around here only concerned in fps....But for me a CAD designer it is huge....

I am dumping my X2 now to build myself a dual core dual opteron 270 setup in the next month.....

 

[mamisano]

Duvie, you should check out the new SuperMicro boards. They are very sweet, especially with the Broadcom chipset. I just upgraded my Server to an H8SSL-i with an X2-3800 and 2GB Ram. The system is rock solid.

 

[Duvie]

Originally posted by: mamisano
Duvie, you should check out the new SuperMicro boards. They are very sweet, especially with the Broadcom chipset. I just upgraded my Server to an H8SSL-i with an X2-3800 and 2GB Ram. The system is rock solid.

YHPM

 

[Artanis]

But for me a CAD designer it is huge...

Yet most of CAD/CAM/CAE software is singlethreaded, so best option for running this kind of software is high speed single core CPU.

 

[Link]

Originally posted by: BrownTown
The problem is that they are jsut updating the old core next year whereas Intel has a compleyelt new core, so the improvements AMD gets won't be nearly as signifigant as Prescott to Conroe will be.

Well, I don't think Conroe isn't exactly a new core, either. Correct me if I'm wrong, but didn't Conroe/Yonah evolve (updated) from P-M, and P-M is an offspring of the tweaked P3 core?

 

[Artanis]

Conroe has a different history than Yonah, it's not coming from a updated PIII.

 

[BrownTown]

in a way you can say that every CPU is jsut a revision of a previous design, but lets jsut say that the difference between Yonah and Conroe is much more signifigant than the updates from Banais to Dothan to Yonah. And the update from K8 to K8L is likely to be much less signifigant based on all current information. Just looking at the fact that Conroe has a wider pipe, and a deeper pipe should indicate that it is a signifigant change. K8 to K8L is expected to widen the execution units, but not the front end, and not deepen the pipe.

 

[Vegitto]

Dude, if we go down that road, you could say that an advanced processor that carries an Intel/AMD core is just a whole bunch of transistors, and that a single transistor is just a step away from a processor. <--- Sarcasm, not true.

 

[SlitheryDee]

Originally posted by: BrownTown
in a way you can say that every CPU is jsut a revision of a previous design, but lets jsut say that the difference between Yonah and Conroe is much more signifigant than the updates from Banais to Dothan to Yonah. And the update from K8 to K8L is likely to be much less signifigant based on all current information. Just looking at the fact that Conroe has a wider pipe, and a deeper pipe should indicate that it is a signifigant change. K8 to K8L is expected to widen the execution units, but not the front end, and not deepen the pipe.

In the transition from Northwood to Prescott Intel lengthened the pipeline of the P4 from 20 stages to 31 stages. Yet that wasn't significant enough to label prescott P5. I think the very fact that prescott performed roughly the same as Northwood on a clock for clock basis with a datapipe that's over 50% longer speaks volumes for what can be done with a simple core revamp.

 

[Duvie]

Originally posted by: Artanis

But for me a CAD designer it is huge...

Yet most of CAD/CAM/CAE software is singlethreaded, so best option for running this kind of software is high speed single core CPU.

Maybe the old shite!!! Nothing I run...

AutoCAD since 2000 is multithreaded for pan, zoom and rendering...

ADT2004 (Architectural desktop) is highly multithreaded for rendering which I use for animating walk-thrus...scales incredibly with each core

3dsmax6 and 7 are both extremely multithreaded...same as above

POVray is multithreaded and has been now for 1 year...Cinebench...

Thelist goes on, but like I said the big 3d players are all multithreaded that I use...The faster I can animate and render my project the faster I can get done and make changes. Older systems would take so long that making slight corrections was a real pain...

Antiquated 2d design shite maybe!!!



The only apps I am concerned about are some of the multimedia ones...Soe of them scale well with dual cores but take very little advanatge of more cores after that...If you go back and look at the Techreport review done back in May of 2005 you will see that all the CAD apps he tested used 4 cores very efficiently in the dual 275's test.

 

[Artanis]

OK, if you call CAD software like Catia V5, Unigraphics NX, SolidWorks, SolidEdge, etc. "old shite"...

 

[Fox5]

Originally posted by: SlitheryDee

Originally posted by: BrownTown
in a way you can say that every CPU is jsut a revision of a previous design, but lets jsut say that the difference between Yonah and Conroe is much more signifigant than the updates from Banais to Dothan to Yonah. And the update from K8 to K8L is likely to be much less signifigant based on all current information. Just looking at the fact that Conroe has a wider pipe, and a deeper pipe should indicate that it is a signifigant change. K8 to K8L is expected to widen the execution units, but not the front end, and not deepen the pipe.

In the transition from Northwood to Prescott Intel lengthened the pipeline of the P4 from 20 stages to 31 stages. Yet that wasn't significant enough to label prescott P5. I think the very fact that prescott performed roughly the same as Northwood on a clock for clock basis with a datapipe that's over 50% longer speaks volumes for what can be done with a simple core revamp.

Yet Pentium Pro to Pentium 2 to Pentium 3 had almost nothing changed and yet got new model numbers.

 

[SlitheryDee]

Originally posted by: Fox5


Yet Pentium Pro to Pentium 2 to Pentium 3 had almost nothing changed and yet got new model numbers.

Ah that was back when intel had it easy. Did you notice that since the original Athlon was launched no-one declared any "new architectures" unless the performance increase over the previous generation warranted it? Thats one thing AMD has done for us.

We also know that the A64 is highly derivative of the original Athlon. Of course there have been lots of changes since then (couple pipeline stages, branch predictor, IMC, etc.), but the increases in performance over that previous generations warranted a new brand name and that's how it should be.

 

[Duvie]

Originally posted by: Artanis
OK, if you call CAD software like Catia V5, Unigraphics NX, SolidWorks, SolidEdge, etc. "old shite"...

wRONG!!!

http://www.lycos.com/info/catia--catia-v5.html?page=2

catia 5 IS NOW MULTITHREADED...

look there and read the pdf file here...

http://72.14.203.104/search?q=cache:4Tx...en&gl=us&ct=clnk&cd=1&client=firefox-a

and some of that is old considering they are listing older cpus in the file...

While many processes within SolidWorks are multithreaded, there is limited gain in a multiple CPU system. Typically, where the multiple CPUs make the most sense is where other computer-stressful applications (e.g., FEA) are run at the same time. Again, a good benchmark or performance test will help determine the value.

from here... http://www.solidworks.com/swexpress/mar04/200403_techtip_02.cfm

This is from March 2004

SolidWorks is multi-threaded. Many of the user interface activities such as redraws and dialog
box interaction, etc., take advantage of this technology. However, the solving process used for
parametric modeling is by nature very linear and cannot take full advantage of parallel
processors. No benchmark tests have been done to determine the increased speed associated
with running on a multi-processor machine but there should be a slight performance gain.

So yes but very limited in this case..... http://www.javelin-tech.com/main/support/pdf/hardware.pdf


SolidEdge

Where hidden lines are required, their processing is multithreaded.


Unigraphics NX

http://www.tenlinks.com/NEWS/PR/ugs/091205_increase.htm



So basically all you need to do is google and see you dont know much.....


I will grant you in most activites dual core does not help me in CAD...It as more a function of my graphics card and amount of ram....however in rendering where it matters more for me then 3d modeling of mechanical and machine parts it is HUGE!!!!

 

[chilled]

I'm just relieved AMD are doing something...

 

[Artanis]

Comme on Duvie as you can see there Catia is multithreaded only in DMU and Analysis modules, UG NX only in some functions like boolean and offseting operations and so on. All I want to say is that this kind of software takes more benefits of a single high speed core than of a dual-core CPU. At least for now. To have an image of performance scaling with CPU speed, google-up more for some SpecViewPerf results

 

[Duvie]

Originally posted by: Artanis
Comme on Duvie as you can see there Catia is multithreaded only in DMU and Analysis modules, UG NX only in some functions like boolean and offseting operations and so on. All I want to say is that this kind of software takes more benefits of a single high speed core than of a dual-core CPU. At least for now. To have an image of performance scaling with CPU speed, google-up more for some SpecViewPerf results

Exactly....I stated that but most of the 3d apps that use similar ray tracing rendering all are multithreaded or they are extremely antiquated....

For me and 3d users that animate they are likely using the software I am and therefore big advantages...I see little advanatges working in the 2d environment with my work CAD...most of the times I have seen companies using the Soilidworks have been very simple design of parts for machinery and what not they use to drive their machines. Hardly stressing current configs....However again I suggest you look at 3d rendering and then doing a 2 minute clip at 15-29fps....You will see there is no substitute for more cores....