Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
It's rather the strategy for 2008 not the production problem anymore. AMD would like define Athlon has a good performing budget CPU, Phenom as superior, expensive brand.
For OEM market Athlon is ok - the price for 5x00 models are higher, 65nm production is cheaper than Phenom.
Still not as fast as Intel but getting better.
I don't think that 3.x GHz Phenom is a solution to fill the money hole.
Anand doesn't show any reality about quad-core performance.
Crysis is the first game optimized for multi-core system.
Phenom setup should have better:
- core as this is B2A revision
- chipset supporting AM2+
- HT3 up and running
- DDR2 800/1066 MHz
and I'm sure it would lead to even deeper losses.
There is no danger of Advanced Micro going out of business, Intel has said :)
http://www.bloomberg.com/apps/news?pid=conewsstory&refer=conews&tkr=INTC:US&sid=aJT60dh3...
HT3 + 3GHz Phenom + 3DMark06 + Crysis game benchmark will give an answer.
It may be a tough choice for those last remaining anything-but-Intel customers.
In that case AMD will be smarter choice. :)
Why?
- I can use my old motherboard with AM2 socket.
- I can save one day of my life not installing system and other
apps for a new Intel CPU/motherboard, another one in 2009.
- I'll be able to put faster AM2+ CPU in 2009 (45nm?) saving another 150$ for another motherboard.
I bought X2 for 89$, replacing my A3200+.
I'll take tri-core for 100-150$ next year.
I need one core more for VMWare hosted Linux running in background. Two is not enough, four will cost too much ($200+).
And I don't care about percentage difference in benchmarks.
I care about my money.
AMD has no real weapons to fight with in the mid to high
end desktop and mobile markets.
As reported AMD has more than 50% of retail in US and China.
Most of the revenues comes from OEM deals and distribution.
High end desktop is very profitable but this is 5-7% of the overall income. Intel has sold a million quad-core CPUs and it took 10 months! 1mln chips for 500-2000$ is mere 100mln USD per month - and most of them as servers chips not C2D.
First AMD is ready to sell a lot of K10 cause OEMs have proven socket F constructions. They continue to sell the same servers/rebranded boxes with new processors.
Second, AMD hopes to upgrade thousands servers during next 9 months. I see Hector saying every day that each dual-core socket F Opteron sold in 2006/2007 MUST BE replaced by quad-core Barca.
He's a victim of the 30% plan for sure.
Despite weak X2 position it was promised to reach 30% of the market till 2009. All this price cuts where focused to increase AMD's marketshare at any cost cause it was promised to institutional investors.
So he's 6 months late Barcelona victim, channel disaster in last Q4 - Turions/Athlons order mistakes.
Or maybe he was trying to get rid of Hector and take his chair?
45nm info
With regard to cost strategies , AMD management reiterated that its Fab 36 should get to full production of 65 wafer outs by September.
Management also maintains that Fab 36 should see the first 45nm production on schedule for late 2Q07 with a ramp over the following 9 months. AMD notes that it remains closer to Intel (if 45nm ramps on time) with around a 9 month lag than it has ever been on geometries.
We recognize investors may remain somewhat skeptical around the execution challenges for 45nm.
Full coverage here:
http://www.investorvillage.com/smbd.asp?mb=476&mn=53745&pt=msg&mid=2314636
POV Ray again.
The Inquirer says that the demo was based on 1.8GHz HE parts scoring 4000pixels per second. Intel can achieve 4677 with 3.0GHz clock (also 4x4cores). Assuming almost linear scaling with HT3, 2.5GHz should score about 6000 beating Intel by 29%.
Comparing test to this results:
http://www.extremetech.com/article2/0,1697,1976400,00.asp
it is difficult to believe that Barcelona is
5.6x faster than Athlon FX 2.8GHz and 4x better than C2D.
First (?) Barcelona benchmark
http://www.fudzilla.com/index.php?option=com_content&task=view&id=982&Itemid=1
At least he is delivering some numbers in opposite to Sharikou ;)
Fuad Abazovic is a guy from TheInquirer afair.
Regardless of Q1 loss AMD rose more than 10% in the last a few days. I really understand one thing that the Q1 results were known EARLIER and now someone knows what will happen in Tunis, Q2 or later buying shares now.
Can someone explain what kind of magic is that?
And what SEC is doing when company posts (600m) and roses 10%??!
Do you know where the summer ends this year? In second half.
So July/August is still on track.
In the meantime have you noticed that Barcelona+AgenaFX is going to be showed in two weeks in Tunisia near R600 premiere along with R650 announcement (65nm shrink).
I really don't understand this pessimism.
It is rather Intel PR effect than the reality.
Is Core2Duo so good that AMD has no chance to sell anything?
Or nVidia portfolio?
No. It's just the time where AMD is working hard to deliver better products. Below 120$ AMD has very strong offer, better than Intel. Smaller profits - yes but still standing and preparing for next battle.
3.0GHz Opteron is available and 95W version will come soon The Register says here:
http://www.theregister.co.uk/2007/04/04/amd_opteron_2222se/
These parts will be 90nm I assume. AMD wants to increase clock or lower the power envelope to 95W.
Well, 6200+ is coming? It reminds last child of K7 - Barton 3200+ available with 2.2 Clock...
Re: Barcelona Demo not seen on CeBIT
Question: have you seen anything more than "Task Manager demo" for first Opterons? Athlons64? QuadFX platform? Brisbane (DDR2)? Turions? As I remember such wide official and public demo happen only once during x86 history, done by Intel Corp when they desperately needed such move (Conroe).
I'm pretty sure that some K10 samples are there. I would be glad to see that Dresden fabs can produce 2,000,000 of these per week more than satisfy own curiosity.
There are two things:
- acceptable price
- acceptable unit request
When Intel is more competitive than AMD in terms of performance AMD must sacrifice the margin to sell own chips.
OEM is calculating very simple - I want to have the same margin e.g. 5% for 1000$ PC and for 500$ so if you (Hector) want to sell me more CPUs I need lower price and more units.
If Hector is not able to deliver at promised price or (worse) deliver enough units the whole contract is doomed. Simple.
When AMD will be able to deliver enough chips, they will be able to get more contracts/designs with OEMs.
Second, OEM agreements cover future delivery so 65nm slow ramp causes more trouble.
The channel is doomed because OEMs must get chips first.
From my point of view, the production is a problem.
This server test is betweeen Socket 940 Opteron (DDR1) and latest Intel. Next one should happen between Clovertown and Athlon MP.
But AMD fails in production:
- Hector said yesterday that they cannot supply partners because they don't have enough chips!
- No 65nm in Europe and Asia
- Where is 65nm Turion and Sempron???
With this kind of pace we will see first Barcelona on Christmas.
Based on this benchmark I made some calculations assuming that this unknown Opteron is somewhere between 2.3GHz-2.8GHz.
It seems that K10 can be faster 37-46% faster clock to clock if we compare results to 2.66GHz Conroe listed on the same table.
It sounds incredible...but if it's true...
Wbmw I wasn't be so sure about that, it depends:
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2889&p=6
and here
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2933&p=9
Second, if you look at configuration systems how you explain this "miracle" that CPU+chipset+nV8000GTX draws about ~180/250W under load? I don't discuss that C2D is more energy efficient - it is better solution for now - but it doesn't mean that Barcelona cannot be faster and cooler.
BTW:
http://www.electronicsweekly.com/Articles/2007/02/23/40820/Intel+makes+date+for+45nm+Xeon+server+chi...
"Our basic message in yesterday's call with the financial community is that Intel is back in servers in a big way" - so 65nm Xeon 53xx were not enough?
5600+ 65nm simply doesn't exist. But soon...maybe...
You compare 90nm AMD to 65nm Intel. :)
Wall wattage
I mean the real power consumption for the whole system.
You put some workloads on two machines (DELL, HP) with similar configurations (disks, controllers) and you measure the power draw for each of them within the same time.
Regarding HPC the heterogeneous approach is also better cause using GPU drawing 240W (R600) against CPU 125W you are able to complete calculation 5-10x faster and draw less power.
It is obvious I think.
Barcelona power lead
Performance per watt considers both.
You forget that Intel numbers are the average numbers not maximum.
When Intel says 50W it doesn't 50 is the max.
Comparing Opteron now vs. C2D - it looks that power consumption and power management features are similar - just add memory controller and FBDIMM to the equation. The wall wattage matters - not the stickers on CPU.
Barcelona power consumption under idle or average workload is going to shine and we still talk about first bunch of 65nm process. I'm optimistic due to these lows (2.1GHz, HT1) because Hector is not giving all he has in cards.
It might seem small, but it may be enough to retain leadership in those benchmarks
AMD is trying to maintain power consumption on lowest possible level. Keeping power envelope is more important than performance lead. Barcelona at this speed can be 1% faster with 40% lower power reqs. This is the main goal and this is what customers are expecting. With 65/95W quad-core we should expect passive cooling - another reason to buy Barcelona.
Barcelona derivatives are going to keep this envelope in 2008 with larger L3 cache sizes thanks to 45nm probably and HT3 which obviously consumes more power.
I think that there is not indication that AMD needs more cash now. As Mayer said on the last conference the company has the plan. 65nm/larger wafers offering will cut production costs during next two quaters, during next 5 months we will see a lot of new products (R600 and cheaper cards, chipsets, Barcelona)
Capital expanditures should be much smaller during next 2-3 quarters cause 65nm toolsets are bought for FAB36 and FAB38 is 2008 expanditure.