Monday, September 18, 2006

AMD's CTO built his own 4x4 box

It must have been fun. I bet he got all the parts for free -- except the case, maybe.

Remember the story that Intel CTO used a picture of an Opteron board during IDF?

22 Comments:

Anonymous Anonymous said...

Building a "4x4 box" is just building a 2P system.

Many people have been doing that for a long time.

I doubt "4x4" is really any cheaper than a low-end 2P system. The only difference will be less testing of the motherboard because it is for gamers and not for the data center.

It looks like K8L is going to have very aggressive power scheduling. If it doesn't work better than these 2 dual-core Opterons, it means the processors will never run full speed unless you disable "PowerNow" (if even then).

So unless software gets pervasively multithreaded, there will be no real speedups.
The other thing that stands out is how "4x4" is really just "SLI for processors". Other than for a small fringe of users, it is just a gimmick:

Phil Hester himself said that “The 4X4 initiative was driven by multithreaded games,” even though there are hardly any “true” multithreaded games on the market. Is AMD looking ahead in the gaming arena? Absolutely.

So maybe one day there will be some mainstream use for "4x4", but no one can say when that will be or what that will be.

Somewhere down the road, 9-12 months from now, we will either have amazing systems from AMD or we will see that K8L is another AM2 dud design.

Meanwhile, there will be hundreds of blog entries, just on this blog, prattling about it. Ah, mankind; so many people, so little talent.

10:50 AM, September 18, 2006  
Blogger core2dude said...


Wanting a more defined answer about IPC and where it will be in relation to Intel’s Core 2 Duo architecture, we went back to AMD and asked for exactly that. Gamers know this is going to be of the utmost importance. Phil Hester left us with this:



“With our next-generation architecture we expect to continue our track record of delivering industry-leading performance. When you combine this with 4X4 you change the rules of the game, and provide an enthusiast-class platform that is upgradeable, scaleable and unparalleled in terms of performance.”

In other words, he didn't answer the question. This is the most hilarious article I have ever come across. When you ask about IPC of K8L, the dude starts talking about K8L. This can mean only one of three things:
1. He is deaf.
2. He is an idiot.
3. He is not so sure about IPC of K8L.

Take your pick.

11:04 AM, September 18, 2006  
Anonymous Anonymous said...

RE: Core2dude

Phil Hester

Born:
April 30, 1955, Corpus Christi, Texas

Education:
University of Texas at Austin, BSEE, 1976; MS, engineering, 1981

Current position:
CTO, Advanced Micro Devices Inc.

Previous jobs:
# Newisys co-founder and CEO, 2000-05
# IBM Corp., 1977-2000: CTO, PC Division; general manager, Integrated Product Solutions Division; vice president of hardware development, RS/6000 Division

Awards and patents:
# Owner of 12 patents
# IBM Outstanding Technical Achievement Award for RS/6000 work


Core2dude, you make me laugh..AT you. LOL

11:52 AM, September 18, 2006  
Anonymous Anonymous said...

"...This can mean only one of three things:
1. He is deaf.
2. He is an idiot.
3. He is not so sure about IPC of K8L.

Take your pick."


You forgot the right answer:

4. He doesn't want Intel to know what they are up against any earlier than is necessary, and Core2Dud is an idiot.


Oh yes, and 4X4 is a gimmick. That's what innovative companies do when they are beat they come up with gimmicks to allow the enthusiast market to still thrive with their products. What was Intel's innovative gimmick the last 3 - 5 years?

12:30 PM, September 18, 2006  
Blogger core2dude said...

Well, that then rules out the first two. In that case it has to be the third one: they are not sure of K8L IPC.

P.S.: 12 patents is nothing. Winners of Max patent award at IBM routinely file (and get) 60+ patents a year and those at Intel get 40+ patents a year.

1:01 PM, September 18, 2006  
Blogger Kalle said...

Actually he just forgot the fourth point:

4) its a secret

1:44 PM, September 18, 2006  
Blogger Sharikou, Ph. D said...

When you ask about IPC of K8L, the dude starts talking about K8L

IPC of K8L is already discussed during the tech analyst day, 60% improvement on integer and 200% improvement on FP. Dirk Meyer told you AMD will have a faster core (compared to Intel).

1:51 PM, September 18, 2006  
Anonymous Anonymous said...

"4) its a secret"

Basically, anyone that is upgrading to AM2 is upgrading based on hope.

1:58 PM, September 18, 2006  
Anonymous Anonymous said...

IPC of K8L is already discussed during the tech analyst day, 60% improvement on integer and 200% improvement on FP. Dirk Meyer told you AMD will have a faster core (compared to Intel).

Show us the 60/200 links.

AMD said it'll have 'leading performance'. Don't they have an FX62 ad with 'leading performance' claims?

2:01 PM, September 18, 2006  
Anonymous Anonymous said...

Too lazy to look up the proof yourself man? Theres plunty of proof if you read it. But nooo people are too lazy and want to have the silver flying carpet and thrown rolled out before them.

Lazy bones, We aready know K8L is killer like he says and I know for a fact spec for spec Conroe is in no compitition for K8L. Conroe couldn't even wave a stick at even 1 of its cores. Its aready more then 50% faster then K8's alone. (K8L is I'm talking about) Conroe is only 10% faster then K8's.

But I guess some people will have t o wait for it to really come out to say we told you so. That would leave you speachless. Well guess you don't know about K8L well good for you its best you don't know its specs. To be blunt it uses similar data processing and works on faster cycles per clock. If you know what that mean then you know Duo2 will be no more...

3:18 PM, September 18, 2006  
Anonymous Anonymous said...

man...we need to have some crusade against those fake Intel Centrino ads..trying to spread FUD. Actually all Intel ads.. And all this in the middle of exploding laptops and hackable wireless chipset ..

I think Intel should be banned from having processor ads for laptops,desktops etc ...why ..if I buy a laptop ..I might leave it to the vendor's good judgement (HP or Toshiba or oven SONY) to decide what processor to use. I would fully trust the vendor to select a particular processor to meet the needs ...why should I care if it is an AMD or Intel. Cost is another factor that decides.

Ofcourse this is only true for the consumer market ...But if i am a gamer or planning to deploy enterprise level servers or high perf computing platforms I might get into the nitty gritty...

What is then the purpose of the Intel laptop ads ???

I remember reading like somewhere in Europe that banned an Intel ad that said something like "for better music use a Centrino " ..atleast those folks have some commonsense

3:45 PM, September 18, 2006  
Anonymous Anonymous said...

Both camps wants to be ahead of the ball game but you have to realize a new design like conroe is more advance than amd64 in many respects and that is why their benchmark are ahead to any amd 64. K8L is design to be ahead on conroe design such as double the FPU which translate improve performance. With both camps are improving their design, consumer benefits all of these aspects. We don't want both of them to loose because if this happens all consumers will face a nighmare because they can monopolize the market and they will not come up any idea to improve their design. There will always be intell fans and there will always be an AMD fans. You can never change their belief once they started liking the product. It's just a facts of life!

6:17 PM, September 18, 2006  
Anonymous Anonymous said...

Don't get so confident.

http://www.hkepc.com/bbs/itnews.php?tid=671374&starttime=0&endtime=0

Kentsfield is launching November at 2.67GHz Extreme Edition and by Q1 mainstream Kentsfields will be available at 2.4GHz.

http://www.hkepc.com/bbs/itnews.php?tid=671378&starttime=0&endtime=0

You can complain about TDP but they have Cloverton in Q4 too. There are several Clovertons with 1333MHz FSBs afterall and the most exciting is the 2.33GHz 1333MHz one with a 80W TDP. Not bad at all. This means that the mainstream Kentsfield's could easily have a 80W TDP as well. (The chart list them all as 1067MHz FSBs, but you can tell the 2.33GHz model has a 1333MHz FSB by it's clock speed. You can't get 2.33GHz with a 1067MHz FSB. Plus, the 2.33GHz and 2.67GHz Clovertons have their names ending in a 5 which denotes 1333MHz FSBs while regular 0s have 1067MHz FSB.)

Note that a 50W TDP Cloverton is arriving in Q1.

6:28 PM, September 18, 2006  
Blogger N4CR said...

The whole 'quad threading argument' is tired. The future is clear - more and more cores/threads, so programmers need to quit whining and get off their butts. No different to everyone saying '64 bit is so far ahead/waste of time etc no use blah blah'. Look at where we are up to now... All it takes is a bit of PERSUASION!

(and pervasive blogging ;) )

RE: IPC links. Read the article posted....

When talking to both Phil and Pat, we got some interesting answers about advancements in Instructions Per Clock with the Barcelona Pat Conway said that we will see a “performance improvement of 150% over the next two years that will be within the same power envelope.”
150%.. but over two years. Probably waiting for HT3 enabled and revised k8l versions to come later? Enough to make conroe look like a glorified celeron on release. (Looks over at the little 'opteron 1.4ghz that could' back in 03~).

Same thing here again I think....

6:42 PM, September 18, 2006  
Anonymous Anonymous said...

The Kentsfield used by Toms was just a little over PE965 in terms of wattage, which is in turn, just a little over FX62.

7:26 PM, September 18, 2006  
Anonymous Anonymous said...

“performance improvement of 150% over the next two years that will be within the same power envelope.”

Core 2 arrived with less wattage, more performance. Comparing PE965 with Kentsfield, they'll stay within the same power envelope, but Kentsfield already has the lead in gaming with the Core 2 goodness, and the ~80%(I'm going to assume that 150% means 50% more than current offerings) gains in threaded apps.

7:44 PM, September 18, 2006  
Anonymous Anonymous said...

Remember the story that Intel CTO used a picture of an Opteron board during IDF?

This can mean only one of four things:
1. He is blind.
2. He is an idiot.
3. He has no clue.
4. He copies so much of AMD stuff, he's confused.

8:01 PM, September 18, 2006  
Blogger Kalle said...

"IPC of K8L is already discussed during the tech analyst day, 60% improvement on integer and 200% improvement on FP."

Ah, those theoretical numbers. I know that Core2 had 200% or more FP performance improvement too but I'm not 100% sure about its integer performance. I think it was somewhere around 40-60%.

Someone should run superpi under 64bit windows to see how much better it is compared to netbursts or K8 or whatverer else. 64bit because then the macro-op fusion wouldn't mess up the results. Superpi should show integer performance rather nicely.

12:23 AM, September 19, 2006  
Blogger N4CR said...

I might leave it to the vendor's good judgement (HP or Toshiba or oven SONY)

HP/Compaq makes some turion notebooks, some of the first I can remember were HP/Compaq that we had through work.

As far as I can remember.. Sony uses no AMD mainstream desktop cpus. Toshiba does not currently either.... for a long time. Last model I can remember was the sattelite 1550CDS.. k6-2 380 Mhz. CXT core. Got one right next to me :)
Good little machine if not a little toasty. Few mods fixed the air circ though.

3:49 AM, September 19, 2006  
Anonymous Tritosh said...

Idont understand intel fanboys...

sharikou atacks the kentsfield because only gets better performance when you use super heavy threaded applications, and they all defend so..

sonow sharikou posts something related to a quad core k8, or 4 x 4 system wich does very fine in multithreaded just like kentsfield, but here they come the fanboys of intel, now atacking the k8 because "its depends too much on multithreaded applications to deliver good performance"

Iswear these kiddies only try to find whatever excuse to bash AMD

7:59 AM, September 19, 2006  
Anonymous Anonymous said...

On a side note, it seems like Intel is gonna launch a bunch of single core Conroes:
http://www.dailytech.com/article.aspx?newsid=4252

I guess they have A LOT of Core2Duo coasters to cut up?

12:42 AM, September 21, 2006  
Anonymous Anonymous said...

Just like how amd fan boys are soo quick to believe the k8l hype with ABSOLUTELY NO BENCHMARKS but when conroe was being benchmarked everyone didnt believe it even though the numbers were staring them rite in the face. They came up with all the stupid excuses which eventually got owned once conroe was released.

X6800 beats the fx-62 chip on the average (across all benchmarks) by atleast ~20% so where people get this information about conroe being only 10% faster than k8 is beyond me. Probably the same people that are waiting for their reverse hyperthreading bios update.

5:29 PM, September 26, 2006  

Post a Comment

Links to this post:

Create a Link

<< Home