AMD K10 Inside Article.

[radnor]

For a long time i didnt read a good article like this one.
If your pacient, and you got a good few minutes to burn, This article is excellent.


http://www.insidehw.com/Editorials/Columns/AMD-K10-Architecture.html

Happy Readings

 

[TechnologyCoordinator]

Thanks for posting that.

That goes really deep, beyond my understanding!


I find it funny for K10 to be called "bleeding edge", but I guess under the hood it may be (according to some parts of the article). However, the translation into real performance isn't always there.

 

[radnor]

Thanks for posting that.

That goes really deep, beyond my understanding!


I find it funny for K10 to be called "bleeding edge", but I guess under the hood it may be (according to some parts of the article). However, the translation into real performance isn't always there.

Yup, that happens alot in CPU world. Anyway, ive read it here at work, but im gonna take a deeper look when at home to fully understand it. Its really complete about K10 Arquitecture. Thanks for the Bump TC, i think its a great article for a nice techie forum like one.

 

[spud]

Thanks for posting that.

That goes really deep, beyond my understanding!


I find it funny for K10 to be called "bleeding edge", but I guess under the hood it may be (according to some parts of the article). However, the translation into real performance isn't always there.

Ya bleeding edge if the Core 2 didn't exist.

Word, Playa.

 

[hannibal]

Well there is allso very good article at Anantech:
http://www.anandtech.com/weblog/showpost.aspx?i=443

Summa summarum it says that Barselona has no any chance in single threaded games, but it's good in multi threaded tasks, due it's architectural design.

Actually very well writen article that makes sence, but does not give confort to gamers...
Hard to say if real multi threaded games chance the thing, but it seem that even with them slow L3 cache will cause headache AMD users.

 

[justjc]

Thank you radnor for posting the article.
I had thought AMD had lost it when they announced that K10 would also be the basis for their 6 and 12 core processors, was hoping for Bulldozer, but after reading the article it seems to me that the Barcelona architecture might just be good enough. If nothing else it's built for the future and have potential for improvement.

 

[Mathos]

Yeah, I read the anand article too. Nice to see one of the site back up what I've been saying all along with the L3 speed slowing it down. Just makes me wonder why they aren't increasing L2 cache with Shanghai instead of the L3. Only other thing I can think of is that they're speeding up the IMC/L3 too.

By the way, just received my 9850 today, and 4gb of Mushkin DDR2 1066 memory. Not sure whether I wanna try and put it in tonight. Need to get a secondary HD back in the system to run a driver back up just in case. Not to mention the Barracuda drive I have in there acts like it may be failing. Either that or the sata ports on my board are fubbing up.

 

[doomsdaydave11]

Sigh, fanboy bashings are constant...:

2003: ATI fanboys constantly bash NVIDIA fanboys for their crappy FX cards,
whilst the ATI 9700's dominate them...

2004-2005: ATI and AMD dominating the enthusiast market. Little AMD's AMD 64's laugh at the trashy P4 Prescotts.

2006-2008: Core 2 Duo's revolutionize modern computing, while NVIDIA takes back the graphics card market.

Soon to be: AMD's Raedon 4xxx series trashes NVIDIA's current stuff. Price drops FTW

I've just decided to stop reading this fanboy vs. fanboy crap. For the last 10 years, the computer world has been a fricken rollercoaster.

I'm confidant AMD won't blow this one as bad as they did the Phenom X4's and especially X3's. Every company has its ups and downs. I would have thought that people on these forums should know that by now.

 

[TechnologyCoordinator]

I'm confidant AMD won't blow this one as bad as they did the Phenom X4's and especially X3's. Every company has its ups and downs. I would have thought that people on these forums should know that by now.

Please define "this one". The article is about K10, which in your words is already blown.

Radeon 3XXX series competes very well with nVidia, but fell short of the crown.


AMD's last two releases failed to re-capture the crown. AMD has broken the "I'm faster, no I'm faster, no I'm faster, etc..." cycle both in the CPU and GPU market. On the video card side AMD is still doing good and has competitive products. However, on the desktop CPU side AMD is failing miserably.


The article was very well written and had tons of information. I just wonder why they had to ruin their credibility by calling the Phenom "bleeding edge" when in fact the results it produces falls far short of "bleeding edge" and would be better described as "mid-range".

 

[radnor]

Please define "this one". The article is about K10, which in your words is already blown.

The article was very well written and had tons of information. I just wonder why they had to ruin their credibility by calling the Phenom "bleeding edge" when in fact the results it produces falls far short of "bleeding edge" and would be better described as "mid-range".

In that case, i think its bleeding edge in a arquitectural conotation. I think we all seen too many "Bleeding edge" products that floped miserably, not because they were perfect on paper (they were), but because on pratice, they floped.

3dfx was a nice example of it.
The Kyro II Chip was a excelent example of it
Netburst and G5 CPUs Arq. are also a great example.

History just tends to repeat itself. They were all bleeding edge in terms of design and arquitecture, many of them advanced for their time. But, they floped anyway.

 

[keithlm]

History just tends to repeat itself. They were all bleeding edge in terms of design and arquitecture, many of them advanced for their time. But, they floped anyway.

What are you talking about?

You VHS fanbois... dang you big tape fanatics. We all know that Beta RULZ.

SONY4LIFE. Long live BETAMAX. You just wait until we have LAZERDISC!!!!

THe VHS consortium is THE DEVIL.


(Sorry had to be done.)
========================

 

[TechnologyCoordinator]

What are you talking about?

You VHS fanbois... dang you big tape fanatics. We all know that Beta RULZ.

SONY4LIFE. Long live BETAMAX. You just wait until we have LAZERDISC!!!!

THe VHS consortium is THE DEVIL.


(Sorry had to be done.)
========================

LOL, good memory Keith! Beta is a little before my time, so I got to go to Wikipedia for that one!

 

[JAYDEEJOHN]

Whattaya sayin? I shouldnt be shopping at Big Bobs Betas and Bell bottoms no more?

 

[exit2dos]

Good link radnor, thanks!


Now if I could only get Microsoft Bob to run on my PCjr.

 

[endyen]

For a long time i didnt read a good article like this one.
If your pacient, and you got a good few minutes to burn, This article is excellent.


http://www.insidehw.com/Editorials/Columns/AMD-K10-Architecture.html

Happy Readings

It is a good piece though it does have a distinct AMD bias.
It accuses the core2s of cache thrashing, without taking the software advantages of smart cache into account. Then it turns around and says

Much more could have been gained with optimizations and compilers written specifically for the K10 architecture,

I agree that barcelona is bleeding edge, may even be a half step closer to real AI. That does it no good though since it has to operate in a "real software" world.

 

[endyen]

Well there is allso very good article at Anantech:
http://www.anandtech.com/weblog/showpost.aspx?i=443

Summa summarum it says that Barselona has no any chance in single threaded games, but it's good in multi threaded tasks, due it's architectural design.

Actually very well writen article that makes sence, but does not give confort to gamers...
Hard to say if real multi threaded games chance the thing, but it seem that even with them slow L3 cache will cause headache AMD users.

I tried to read that anand bit but only got to

The fact that games prefer a 4 MB L2 is not going to change when you run it on a AMD CPU with integrated memory controller. A L2 can deliver the necessary data in 12-20 cycles, an IMC needs about 100 cycles.

before it really pi$$ed me off.
First off, even the core2 can handle memory @<90 cycles, the phenoms are doing it in the 40s.
Ever checked to see your memory usage in game? 4megs is sweet didly. Oh, if you are running on the rails, it's lots, but run into a wack of enenmy AI and what happens? That's where the low latency memory comes in handy, when you need it.