AMD and Intel General Discussion (not for getting help)

[xtc28]

Whats that, french for pee-on?

 

[JAYDEEJOHN]

I think is means, of course Ive already surrendered at the office, not sure tho

 

[xtc28]

Le bon temps roulet!!!!!!
Let the good times roll!!!

 

[JAYDEEJOHN]

Yeowsers!!!
http://www.fudzilla.com/content/view/15086/1/

SUNNYVALE, Calif.--(BUSINESS WIRE)--AMD (NYSE: AMD - News) today announced that according to the latest industry data, notebook manufacturers have chosen ATI Mobility Radeon™ Premium graphics processors to drive more than half of today’s discrete graphics notebooks. The data for mobile discrete graphics reveals a 36.5% market share gain, and 87.27% more units sold quarter-over-quarter through the first six months of 2009.1 AMD now holds the “Number 1” position with an overall 53% mobile discrete market share. The incredible momentum behind ATI Mobility Radeon Premium graphics processors is a result of:

http://finance.yahoo.com/news/AMD-Takes-Notebook-Discrete-bw-2174838590.html?x=0&.v=1

“As consumers increasingly demand greater visual computing capabilities in their notebooks, the market for discrete graphics is only expected to grow,” said Dr. Jon Peddie, founder and principal, Jon Peddie Research. “The increase in volume in the second quarter of this year, and in AMD’s mobile discrete market share, is a testament to consumers’ appetite for visual computing, and manufacturers’ readiness to satisfy it as we move into what has traditionally been the busiest season of the year.”

 

[fazers_on_stun]

^ Not too surprising - nVidia's problems with their notebook GPUs sorta caught up with them. However it's only one quarter, and NVDA did outperform the street in Q2, so they must be doing something less wrong 😀.

It'll be interesting to see how the notebook i5s with on-die IGP does in this market. If it's good enough, and cheap enough, it should put a dent in both ATI & NVDA IGPs and low-end discrete numbers.

 

[Reynod]

Now if only they could sell them at a profit ... there is the crunch.

 

[JAYDEEJOHN]

Well, its perfect strategy, pick up massive marketshare, lead your competition in a new arch deployment, gives better time on mobile latter release fates vs the competition, and the competitions old arch, which had a much higher tranny density to perf ratio that their current arch has, as it trends, nVidia has been using last gen in their lower end/mobile units, where the new 40nm node wont help in cost vs perf using G200 at 40nm, due to its lack of tranny gensity/perf.
It looks real good to me, looking down the road

 

[JAYDEEJOHN]

Remember, nVidias profits have fallen off dramatically, as well as their markershare.
The G200 has a very low size/perf ratio, and its costing them some now, but even moreso when they shrink it, as thatll be their mainstream moneymakers, and the shrink wont help in the size perf scenario
So, not only does ATI have a release lead, it will also have a costs lead as well, coming off huge market gains, not bad Id say

 

[xtc28]

Not bad at all!

 

[Reynod]

Except NVidia have argueably a better GPU and once they shrink the GT300 down and get it working then ATI are in the same boat as AMD is wrt Intel.

AMD should be putting more work into loweringthe power on their new dual core and upping the IPC by doing more work on the cache efficiency and latency.

Then they will have a decent platform to compete against Intel in the Notebook / netbook space.

A good CPU / GPU / Chipset solution will make them more competitive in the marketspace where the volume is greatest - that and the combo (cpu / chipset /IGP) area for mum n dad off the shelf stuff.

Just a few opbservations.

Eggs in more baskets.

 

[xtc28]

Good observations and you are right there too!

 

[JAYDEEJOHN]

Currently, there are NO G200s in the mobile section, none nada, zip.
With the shrink, we MAY see some G200s in mobile, but again, theyre more expensive to make, and their best has a 5% maybe margin in DT over ATIs best, and again, the size/perf ratio on the same node goes to ATI.
For cpus, I totally agree, a new stepping would help alot here, even with currently planned cpus.
If they can find a way to lower their cache latency, theyd have a very nice product indeed Thats been chasing them with problems sine 65nm, actually (havnt looked), but I bet K8 at 90nm was bad too, but it didnt rely as much upon it, and overall arch perf was enough to hide it.
If someone has info on K8s cache latency, itd be interesting to see if AMD has made any progress there since K8, I do know it wasnt as good at 65nm, thus we saw the 2.7Ghz doing the same perf as the 2.6

 

[xtc28]

Woah! GTX 280M is here or rather here in some places. Also, so is the mobile i7. we'll see what happens there. Really the 280M is an 8800 gt so i cant really say its a 200series but thats what Nvidia says anyway. Damn those rebadging pricks!!!

 

[amdfangirl]

Currently, there are NO G200s in the mobile section, none nada, zip.
With the shrink, we MAY see some G200s in mobile, but again, theyre more expensive to make, and their best has a 5% maybe margin in DT over ATIs best, and again, the size/perf ratio on the same node goes to ATI.
For cpus, I totally agree, a new stepping would help alot here, even with currently planned cpus.
If they can find a way to lower their cache latency, theyd have a very nice product indeed Thats been chasing them with problems sine 65nm, actually (havnt looked), but I bet K8 at 90nm was bad too, but it didnt rely as much upon it, and overall arch perf was enough to hide it.
If someone has info on K8s cache latency, itd be interesting to see if AMD has made any progress there since K8, I do know it wasnt as good at 65nm, thus we saw the 2.7Ghz doing the same perf as the 2.6

Well that was just a shrink. They also reduced the cache from 2mb to 1mb.

 

[JAYDEEJOHN]

True, I never owned a 1mb model, but it only gave about a 2-4% boost anyways
Back then, having more cache on a AMD really meant having less cash heheh

 

[Reynod]

If someone has info on K8s cache latency, itd be interesting to see if AMD has made any progress there since K8, I do know it wasnt as good at 65nm, thus we saw the 2.7Ghz doing the same perf as the 2.6

http://techreport.com/articles.x/16147/3

The above has a link to Phenom I which includes an examination of the memory subsystem for K8.

Scott is an awesome writer - probably more interesting than Anand, and just as savvy.

They should hire him here ...

 

[JAYDEEJOHN]

I like the pod casts, reading now, ty rey

 

[JAYDEEJOHN]

So, it appears i7 is 20%+ faster, tho its access to main is a small bit faster, interesting
I wish theyd have done a better breakdown on the L2 for the K8, just cant conceptulize this in my head right now, and the top graph showing just 2D and total BW somehow doesnt break it down for me
Question
Once the 512 is used up on say the old K8 smaller 512 vs the 1mb models, before it get to 1mb, howd that break down to latencies?
How many cycles would be lost between 512 and 1mb? I know it amounts to 2-4% overall gains, just wondering tho

 

[JAYDEEJOHN]

Roh Roh!!!!
"Since Larrabee 3 is planned for middle of 2010, in case it doesn’t slips again, Larrabee 4 is being mentioned as a potential launch candidate. If Intel skips Larrabee 3 and decides to go for Larrabee 4, this will cause further delays and might push the release of Intel’s GPU by at least late 2010 early 2011."

http://www.fudzilla.com/content/view/15108/38/

Dontcha just love FUaD?

 

[archibael]

"Can Intel ship bulk 32nm by the middle of 2010?"

Um... yeah. Clarkdale is 32nm. What is he smoking?

 

[Reynod]

For a comparison of the cache structural differences jd look at the older Barcelona Techreport article.

Those little coloured 3D graphs tell you a lot.

Thats why I often point out just how slick and sophisticated the Intel L1 / L2 Cache is, and the prefetch, compared to AMD.

Intel make very good cache arrays ... thats why even the 65nm / 45nm Core2 line we so good despite having a FSB system.

 

[JAYDEEJOHN]

I think people are starting to wonder about LRB, hes in the business of getting hits on his site, so a good "no proof" story
TY for the TR links rey, love the 3D graphs, easier on the brain

 

[roofus]

I will believe Intel has a good GPU when I see it. I have not seen it and the rumors are easy for me to dismiss considering most pre-release rumors are dripping with lies and bias.

 

[JAYDEEJOHN]

I wouldnt say lies, as much as hoped for abilities, which are solely dependent on others, especaily in gaming and apps
Hell, theres a hex core coming soon, with 12 threads available, you excited yet?

 

[roofus]

not really. we don't have much that really makes a quad break a sweat in the DT world so i guess i am one in the minority that thinks their current quad will be plenty for a while. the leap from a good dual core to a quad was not significant enough for me to celebrate the arrival of even more unneeded cores. i am happy with my purchases but none leave what i got rid of in the dust that's for sure.