You are currently viewing our boards as a guest, which gives you limited access to view most discussions and access our other features. By joining our free community, you will have access to post topics, communicate privately with other members (PM), respond to polls, upload content, and access many other special features. Registration is fast, simple, and absolutely free, so please, <a href="/profile.php?mode=register">join our community today</a>!
Post subject: Re: What are the chances that we'll see a brand new AMD Bulldoze
Posted: Tue Oct 27, 2009 12:24 pm
Joined: Thu Jul 26, 2007 9:09 pm Posts: 18 Location: Arnhem, Gelderland, The Netherlands, Europe
inf64 wrote:
I'm really interested in seeing how much efficiency they managed to gain in singlethreaded code with their cluster approach(CMT),should be quite a big step up.
I really wonder how this clustered approach can lead to better efficiency in singlethreaded code. I do not see a higher IPC by limiting a cluster to two ALUs without being able to move instructions to the other cluster. To me Hyperthreading with 4 ALUs looks more powerful than CMT with 2x2.
It seems to me that for AMD it would be better to concentrate on better branch predication, schedular and Back-end than on the ALU-threading.
Post subject: Re: What are the chances that we'll see a brand new AMD Bulldoze
Posted: Tue Oct 27, 2009 12:41 pm
Joined: Sat Jul 25, 2009 2:08 pm Posts: 48
jamannetje wrote:
inf64 wrote:
I'm really interested in seeing how much efficiency they managed to gain in singlethreaded code with their cluster approach(CMT),should be quite a big step up.
I really wonder how this clustered approach can lead to better efficiency in singlethreaded code. I do not see a higher IPC by limiting a cluster to two ALUs without being able to move instructions to the other cluster.
Agreed. CMT or SMT can't improve singlethreaded code, v.v. it deteriorates singlethreaded performace. But i'm very sceptical that AMD is capable (or wishing) to implement any kind of SMT/CMT in Bulldozer due to the face of very complex tasks standing before them in 2010 and the lack of resources. http://www.fudzilla.com/content/view/16146/1/
Post subject: Re: What are the chances that we'll see a brand new AMD Bulldoze
Posted: Thu Oct 29, 2009 8:16 am
Joined: Tue Aug 07, 2007 2:04 pm Posts: 33 Location: Tampere, Finland
yfe wrote:
jamannetje wrote:
inf64 wrote:
I'm really interested in seeing how much efficiency they managed to gain in singlethreaded code with their cluster approach(CMT),should be quite a big step up.
I really wonder how this clustered approach can lead to better efficiency in singlethreaded code. I do not see a higher IPC by limiting a cluster to two ALUs without being able to move instructions to the other cluster.
Agreed. CMT or SMT can't improve singlethreaded code, v.v. it deteriorates singlethreaded performace. But i'm very sceptical that AMD is capable (or wishing) to implement any kind of SMT/CMT in Bulldozer due to the face of very complex tasks standing before them in 2010 and the lack of resources. http://www.fudzilla.com/content/view/16146/1/
AMD has had quite a lot time to create Bulldozer;
K10 was just improved K8, so it did not require the work of all AMD cpu design engineers for al long time, and it was released already 2 years ago. And AMD has been talking about bulldozer's multithreading already at 2005, which means they will have more than 5 years time time implement it (because they definetely have started the design before showing those slides)
Bulldozer has already had its delays, and I've understood that the situation is now that Bulldozer architecture will definetely be ready at 2010, but they decided to skip 45nm(which would make if quite expensive to manufacture, and waste resources by designing it for a process which would only be used a short time) and manufacture it at 32nm, (but not as their first 32nm chip, so that they can separate process problems from cpu architecture problems, which means only few months after the 32nm transition, going from end of 2010 to start of 2011.
2010 will be tough year for AMD's cpu side, but hopefully ATI will keep them floating until bulldozer arrives.
Post subject: Re: What are the chances that we'll see a brand new AMD Bulldoze
Posted: Thu Oct 29, 2009 10:38 am
Joined: Sat Jul 25, 2009 2:08 pm Posts: 48
ATI can't keep AMD afloat just because a whole graphics market is about 1/10 of cpu market. The best what ATI of AMD can do for AMD is to not pull AMD down.
Post subject: Re: What are the chances that we'll see a brand new AMD Bulldoze
Posted: Fri Oct 30, 2009 3:24 am
Joined: Tue Jul 24, 2007 4:01 pm Posts: 215 Location: Perth, Western Australia
yfe wrote:
ATI can't keep AMD afloat just because a whole graphics market is about 1/10 of cpu market. The best what ATI of AMD can do for AMD is to not pull AMD down.
It may (or may not) be 1/10 of the CPU market but it's about 1/4 of AMD's revenue, and with NVIDIA putting mid- to high-end chips on hold and Fermi not being available in quantity for a while yet (and being much larger and therefore more expensive to produce than AMD's Evergreen series) it looks like ATI could be an important source of income for AMD in the short term.
Post subject: Re: What are the chances that we'll see a brand new AMD Bulldoze
Posted: Fri Oct 30, 2009 10:26 am
Joined: Sat Jul 25, 2009 2:08 pm Posts: 48
Revenue doesn't have much sense, profit does have a sense. AMD doesn't provide us the split of cpu-gpu production costs, lets take for NVidia it's about 25%. Even IF AMD ate the whole gpu market (1b$) it would bring into AMD just about 200m$. It's about the same money that AMD is forced to payback as percents of 5b$ debts (many thanks to AMD/ATI deal).
Post subject: Re: What are the chances that we'll see a brand new AMD Bulldoze
Posted: Fri Oct 30, 2009 12:09 pm
Joined: Tue Jun 26, 2007 8:55 pm Posts: 720
yfe wrote:
Revenue doesn't have much sense, profit does have a sense. AMD doesn't provide us the split of cpu-gpu production costs, lets take for NVidia it's about 25%. Even IF AMD ate the whole gpu market (1b$) it would bring into AMD just about 200m$. .
AFAIK in 2008 year nVidia reported $800m net profits with far less then 100% marketshare, so I think you're numbers are slightly off
Post subject: Re: What are the chances that we'll see a brand new AMD Bulldoze
Posted: Fri Oct 30, 2009 3:02 pm
Joined: Fri Aug 17, 2007 2:55 pm Posts: 357
P4man wrote:
yfe wrote:
Revenue doesn't have much sense, profit does have a sense. AMD doesn't provide us the split of cpu-gpu production costs, lets take for NVidia it's about 25%. Even IF AMD ate the whole gpu market (1b$) it would bring into AMD just about 200m$. .
AFAIK in 2008 year nVidia reported $800m net profits with far less then 100% marketshare, so I think you're numbers are slightly off
While their 2007 was good (the $800 is for ops ending Jan 2008), Nvidia's last 5 quarterly profits, in millions of $.
Post subject: Re: What are the chances that we'll see a brand new AMD Bulldoze
Posted: Fri Oct 30, 2009 3:19 pm
Joined: Sat Jul 25, 2009 2:08 pm Posts: 48
P4man wrote:
yfe wrote:
Revenue doesn't have much sense, profit does have a sense. AMD doesn't provide us the split of cpu-gpu production costs, lets take for NVidia it's about 25%. Even IF AMD ate the whole gpu market (1b$) it would bring into AMD just about 200m$. .
AFAIK in 2008 year nVidia reported $800m net profits with far less then 100% marketshare, so I think you're numbers are slightly off
It's obvious that I calculated per quater. What numbers you see wrong here?
Users browsing this forum: No registered users and 2 guests
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot post attachments in this forum
