K8L Architecture on the 65nm SOI Process Technology in Quad-Core Processors In Q2 2007

[TraumaRN]

http://www.digit-life.com/articles2/editorial/amd-guiseppe-amato-conf-part1.html

Definitely a good read. Some PR crap thrown around as well but all in all a good idea of what to expect from AMD on the server front.

Whatcha y'all think?

And can we attempt to keep the flames to a minimum.

 

[videoclone]

AMD wont be showing much performance where it counts with this offering.

I hope they bring out a new Core soon. K9, K10, K11 whatever its going to be named.
If they simply refresh the K8 on 65nm that?s going to suck.

 

[TraumaRN]

Originally posted by: videoclone
AMD wont be showing much performance where it counts with this offering.

I hope they bring out a new Core soon. K9, K10, K11 whatever its going to be named.
If they simply refresh the K8 on 64nm that?s going to suck.

Did you even read the article?!

 

[videoclone]

I had read it way before you posted this...

They say its the best thing to come out everrr.. but they would say that wouldn?t they

It's a server chip plain and simple for workstations, gamers and 99% of general PC users this chip will get slapped by a Core2 Duo at the right Mhz Level that intel can easily reach.

AMD will be number 1 in the 4P servers market with this K8L

2P servers, workstations & PC's. They have no chance.

I just ordered 3x 2P HP ML350 G5 Server with Xeon 5130's because it's faster then the opteron server i was looking at and it was cheaper.

 

[videoclone]

The next PC i buy will be a new core AMD CPU because i like AMD more then Intel and my X2 3800+ is more then enough for everything i do on a PC.

 

[TraumaRN]

Originally posted by: videoclone
I had read it way before you posted this...

They say its the best thing to come out everrr.. but they would say that wouldn?t they

It's a server chip plain and simple for workstations, gamers and 99% of general PC users this chip will get slapped by a Core2 Duo at the right Mhz Level that intel can easily reach.

AMD will be number 1 in the 4P servers market with this K8L

2P servers, workstations & PC's. They have no chance.

I just ordered 3x 2P HP ML350 G5 Server with Xeon 5130's because it's faster then the opteron server i was looking at and it was cheaper.

Pardon my sarcasm but can I have your magic crystal ball to see that it's gonna suck so much. I realize the article is mostly PR, but it's an interesting read and it's alot of info condensed into one page....that is why i posted so people can read and see what its about.

 

[Munky]

K8L is not 4 K8 cores slapped together... :roll:
The K8L is to the K8 as the C2D is to a P3. They going to beef up the decoding and the execution units, as well as improving branch prediction, and who knows what else. I would certainly expect a dual core K8L to be able to go toe to toe with an identically clocked C2D.

 

[videoclone]

From the Anandtech articles I?ve read they go on to mention that the K8L?s got next to no performance improvements and its all just minor automated step up?s for improved functionality.

That's very common for a new revision to an old core.

I think it will be a washout for single threaded apps and thats about 98% of the apps & games out there.

It does however look like a great server chip.... especially when you slap 2 of them in a box to make an 8P server going to 4 of them to get a 16P server and your flying.

 

[DrMrLordX]

What's interesting to me is the smaller L1 cache size per core vs every AMD chip released since the K7. AMD has been leaning on the large L1 cache for a long time.

 

[Viditor]

Originally posted by: videoclone
From the Anandtech articles I?ve read they go on to mention that the K8L?s got next to no performance improvements and its all just minor automated step up?s for improved functionality.

That's very common for a new revision to an old core.

I think it will be a washout for single threaded apps and thats about 98% of the apps & games out there.

It does however look like a great server chip.... especially when you slap 2 of them in a box to make an 8P server going to 4 of them to get a 16P server and your flying.

You are confusing K8L with AM2 or Socket F...this is assured as Anand hasn't POSTED any articles on K8L (it would be hard to as not even K8L engineering samples have been released yet...). You need to read the article a bit more carefully...

 

[soydios]

How about we wait until some engineering samples start hitting the benchmarking labs instead of making random-ass uninformed biased speculations? Hopefully at AnandTech first (wink to AMD).

 

[ahock]

One thing I noticed is that AMD gets the power numbers by adding the TDP of each components. Yes Intel numbers will be bigger but I think the real score would be numbers coming from the wall which Intel shows the strength now. This is also pointed out by the writer of the article

 

[myocardia]

Originally posted by: soydios
How about we wait until some engineering samples start hitting the benchmarking labs instead of making random-ass uninformed biased speculations? Hopefully at AnandTech first (wink to AMD).

Well, some of us want to read about any new technology that might affect our future computing. BTW, to the uninformed, the K8L isn't a version of the K8, like the AM2 chips are. AMD decided to not use the K9 name for marketing reasons (do you blame them?). They decided to call their next chip the K8.50, which is what K8L stands for, since L=50 in Roman numerals.

 

[Regs]

Most of the assumptions made on Daily Tech look to be coming true. Even though both articles are exactly the same except for one crucial statement that they now plan to transistion by Q2 of 2007 - great news.

Improvements to both the front and back end of the K8. I don't remember the K7 to K8 transition too well but I'd imagine before the K8's release they didn't share very much information then either. Other than Direct Connect and the onboard memory controller. Of course it wasn't until later we found out how they improved the actual core with better IPC for x86. This includes beefed up ALUs and FPU's .

I'm still confused between the terms "K8L" and revision H. Maybe because they still intend to keep the name Opteron as it seems the name has become quite the trademark.

My hopes have risen for AMDs "counter" for Intels trio in the past few weeks. I just wish they can make a engineering sample for demonstration before 06' is out.

 

[kenji4life]

While we are talking about x86 future, itanium etc etc I'll bring up a couple of points
First off, let's not forget this article:
http://www.dailytech.com/article.aspx?newsid=1521
and while intellectual property may be owned by intel, it hasn't stopped amd in the past..

on a side note, a coworker of mine whose father worked at intel here in hillsboro (look it up if you don't know) was going to leave intel as well but they paid him off to 'retire' edit: btw he was also a top engineer at that facility

 

[nyker96]

good article I like it. But of course it's still PR from AMD not independent results, should be taking with some reservation.

 

[Hard Ball]

Originally posted by: DrMrLordX
What's interesting to me is the smaller L1 cache size per core vs every AMD chip released since the K7. AMD has been leaning on the large L1 cache for a long time.

The L1 cache is actually the same as K8 64+64, and has been confirmed several times, such as here:
http://www.realworldtech.com/page.cfm?ArticleID=RWT060206035626
The original report from dailytech was actually a misunderstanding.

 

[swtethan]

Q2 2007 is when intel 45nm is out

 

[DrMrLordX]

Originally posted by: Hard Ball


The L1 cache is actually the same as K8 64+64, and has been confirmed several times, such as here:
http://www.realworldtech.com/page.cfm?ArticleID=RWT060206035626
The original report from dailytech was actually a misunderstanding.

Ah, I see. Thanks for the correction.

 

[Viditor]

Originally posted by: swtethan
Q2 2007 is when intel 45nm is out

Ummm...H2 2007 (near the end according to most reports). That's Q4.

 

[XBoxLPU]

Originally posted by: Viditor

Originally posted by: videoclone
From the Anandtech articles I?ve read they go on to mention that the K8L?s got next to no performance improvements and its all just minor automated step up?s for improved functionality.

That's very common for a new revision to an old core.

I think it will be a washout for single threaded apps and thats about 98% of the apps & games out there.

It does however look like a great server chip.... especially when you slap 2 of them in a box to make an 8P server going to 4 of them to get a 16P server and your flying.

You are confusing K8L with AM2 or Socket F...this is assured as Anand hasn't POSTED any articles on K8L (it would be hard to as not even K8L engineering samples have been released yet...). You need to read the article a bit more carefully...

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2768&p=3