The AMD Execution Thread [2007 - 2017]

[2senile]

I thought we were supposed to be seeing HKMG 45nm Phenom IIs this year?
I've sort of been hanging out for that

As you see from my previous post; so was I.

There is a PDF, downloadable from AMD, which states it is an option for 45nm.
With the Phenom II 945 being released at 95W shortly I'm now wondering if they truly feel they need to bring it in just now, because of general process improvement & the added costs.

Would be nice to have now & with the spotlight on energy efficiency I think it would make their server class CPUs more attractive compared to the competition.

 

[dkanter]

As you see from my previous post; so was I.

There is a PDF, downloadable from AMD, which states it is an option for 45nm.
With the Phenom II 945 being released at 95W shortly I'm now wondering if they truly feel they need to bring it in just now, because of general process improvement & the added costs.

Would be nice to have now & with the spotlight on energy efficiency I think it would make their server class CPUs more attractive compared to the competition.

It's very unlikely that they would want to introduce HK/MG at 45nm. There's a lot of risk involved that AMD cannot afford to take, in terms of reducing yield, etc.

Also, the upside just isn't there. To really take advantage of HKMG you are going to design your CPUs differently. For instance, you will use more static logic and less dynamic logic. Since gate and subthreshold leakage go down, some circuit techniques you were using to keep those under control are no longer necessary. You would probably want to redo your layout since the HKMG should increase transistor switching speed (while keeping interconnect constant).

That's a lot of work to really improve your performance, and you only get to take advantage of that work for 1 year, instead of the normal 2 years.

I think the reality is that AMD and IBM won't use HKMG till 32nm, and that's the right decision for them.

DK

 

[trinibwoy]

Long live AMD

The old AMD is dead. Long live AMD. The Sunnyvale, Calif.-based chip maker once went toe-to-toe with the mightiest manufacturing company on the planet, Intel. No longer. Brains, not in-house manufacturing brawn, will have to be enough.

 

[nAo]

Oh..this is even better: http://bits.blogs.nytimes.com/2009/09/11/so-long-cpu-wars-it-was-fun/

 

[Dawgmatix]

Suspect timing of articles in the main stream media ?

Given that earnings release today

 

[trinibwoy]

nAo's link is pretty old

 

[v_rr]

http://finance.yahoo.com/news/AMD-Reports-Third-Quarter-bw-851313827.html?x=0&.v=1

 

[neliz]

Graphics up ~20% versus Q2 and actually turning in profit, all whilst at lower ASP's. Looks like they're doing something right!

 

[Silent_Buddha]

Interesting...

Margins are up for the company. That might mean the CPU division is starting to recover slightly as ASP for CPUs is up while ASP for GPUs is down. Things should improve further in this area also as AMD will be able to command higher margins on their GPU chips without serious competition from Nvidia. As well, it appears that currently Intel is happy with the CPU situation and isn't putting further pressure on AMD CPU chip prices.

Continued to reduce losses, and a good chance they'll show a profit in Q4 depending on how well the CPU division does. If the CPU division doesn't end up losing more money, just the GPU division might be able to pull them into profitability if AMD can increase supply of Rv8xx chips.

Q4 should shape to be interesting as we'll get to see what impact Rv8xx has. If they can supply enough chips to meet demand, this could be an interesting quarter considering they have Enthusiast, performance, and mainstream chips out now for the new line. Added to that, products were well timed to match release of Lynnfield and Win7.

Regards,
SB

 

[Geo]

That looks like with a little luck and no double-dip, a <gasp>, umm. . what's that word again they used to use? Oh, right, *profit*, might be possible in 4Q.

 

[v_rr]

http://phx.corporate-ir.net/phoenix...tId=2387871&WebCastId=915659&StreamId=1371282

 

[trinibwoy]

How long do they have to keep the foundry on their books. It looks like they could be sitting pretty once they shed all the manufacturing baggage. Of course, they'll still have the debt problem but cash flow should improve substantially.

 

[v_rr]

Advanced Micro Devices Inc. Q3 2009 Earnings Call Transcript
http://seekingalpha.com/article/166...-inc-q3-2009-earnings-call-transcript?page=-1

 

[DegustatoR]

http://www.brightsideofnews.com/news/2009/10/27/amd-set-to-increase-radeon-hd-5800-prices.aspx

 

[neliz]

http://www.brightsideofnews.com/news/2009/10/27/amd-set-to-increase-radeon-hd-5800-prices.aspx

Interesting, So TSMC's problems make AMD earn more money, and nVidia without a high-end part untill next year.. I wonder for who this is exactly bad news..

 

[DegustatoR]

Interesting, So TSMC's problems make AMD earn more money, and nVidia without a high-end part untill next year.. I wonder for who this is exactly bad news..

Earn more money by essentially missing this holyday season with 5800 line avialability? Interesting idea. Then I guess NV will earn even more money, yes?

 

[no-X]

That's relative. Do you know how many cards are sold per day/week/month? Do you know if these numbers are worse than numbers related to e.g. GTX280 or 8800U? HD5800 boards are cheap, so demands is higher than supply. That's the only arguable finding. Without knowing the exact numbers you can't say they are missing holiday season.

Anyway, selling 10 boards with $20 profit or 20 boards with $10 profit results in the same earnings.

 

[DegustatoR]

That's relative. Do you know how many cards are sold per day/week/month? Do you know if these numbers are worse than numbers related to e.g. GTX280 or 8800U? HD5800 boards are cheap, so demands is higher than supply. That's the only arguable finding. Without knowing the exact numbers you can't say they are missing holiday season.

When you can't provide enough cards for the demand in October you most probably won't be able to provide nearly enough cards in November and December.
And since when does $300/$400 became "cheap"? (5870s are selling for $500-550 here in Russia at the moment btw.)

Anyway, selling 10 boards with $20 profit or 20 boards with $10 profit results in the same earnings.

Selling 10 boards with $20 profit certainly results in less earnings than selling 100 boards with $10 profit. And that's exactly how it's starting to look like for 5800 series results for the upcoming holyday season.

 

[no-X]

And since when does $300/$400 became "cheap"?

Since the launch of $830 8800Ultra and $650 GTX280.

5870s are selling for $500-550 here in Russia at the moment btw.

GTX280 was twice as expensive as HD5870 here in Czech Republic btw.

When you can't provide enough cards for the demand in October you most probably won't be able to provide nearly enough cards in November and December.

Enough for customers or enough for making some profit? That's not directly related.

 

[DegustatoR]

Since the launch of $830 8800Ultra and $650 GTX280.

I've heard that Boing 787 is quite costly too but somehow I still don't think that $300/$400 for a videocard is "cheap".

GTX280 was twice as expensive as HD5870 here in Czech Republic btw.

So? It was top of the line and fastest single-GPU card on the market. 5870 won't be neither quite soon so I really don't see how anyone would pay $500 for it.

Enough for customers or enough for making some profit? That's not directly related.

Enough to make the most out of being the first to market with a new GPU. AMD will probably sell every 5800 they'll be able to ship but that's probably will be far from what they could've sell if there were no avialability issues.
Right now 5800 is looking more and more like a repeat of 4700 story happening during year's biggest buying spree.
At least 5700s seems to in good supply...