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.
Selling Xeon short? They're selling Xeon as hard as they can, and it still falls short of Opteron.
More revisionism?
Intel relabels Itanic as "RISC"! LOL! The end is near.
(also confirms that Potomac misses the 3.5GHz launch speed goal, only managing 3.33GHz)
http://www.intel.com/products/roadmap/server.htm
Saul DeMone, I believe.
Serverworks and Sun are working together, so probably.
I think by "back asswards" he's referring to his many predictions over the years. :)
I was merely adding to Chris' list of your other pronouncements gone awry, which I believe he compiled in response to unwarranted snarkiness on your behalf. :)
Nobody claimed they EOL'd the x86 server line. It's clear from the record what Intel originally hoped the future of Itanic would be. And it made sense, given their exclusive rights to IPF. But their execution, and the market's reaction in general, dictated otherwise.
Oh, but we can. :) I found this:
http://www.investorshub.com/boards/replies.asp?msg=2163498
So chippy was calling for 10-20% IPC increase, and semicon was up at 15-20% IPC increase.
Well said! Wasn't he also one of the geniuses making wildly optimistic Prescott IPC predictions?
IMO they never existed
It's been established on this board, multiple times, that Intel originally intended Itanic to take over the entire server (and desktop) space.
Yes, all the Opteron 252 listings on the web have been for core rev encoded "BL" which we now know is Revision E with SSE3, 1GHz HT, etc.
The actual OPN core rev code for a 90nm FX is probably something like "BJ", as "BK" will be 1xx Rev E Opteron, and based on the AS vs AT thing, "BJ" should be the 939-equivalent FX.
With Intel launching 6xx P4s on Feb 21, I'm thinking that is the date for the 90nm FX-55.
BTW, there are also now many listings for the 852 on Froogle, with the expected "BM" core rev encoding:
http://froogle.google.com/froogle?q=852+Opteron&btnG=Search+Froogle
It's not that manipulation doesn't occur. It's that seeing a volatile stock drop 25% on a major warning isn't exactly evidence of manipulation.
Sun Readying Launch of 'Galaxy' Next-Gen Opteron Hardware
(URL: http://www.crn.com/sections/breakingnews/breakingnews.jhtml;jsessionid=ZIWFYS0EFT3LSQSNDBGCKH0CJUMEK...
)
By Elizabeth Montalbano,
7:50 PM EST Thu. Feb. 10, 2005
Sun Microsystems next quarter is expected to unveil its next-generation Opteron-based servers, the result of a project Sun co-founder Andy Bechtolsheim has been working on since he rejoined the company last February, CRN has learned.
The project, code-named Galaxy, will produce fully industry-standard, Sun-designed and fabricated servers on forthcoming dual-core Opteron technology from AMD, which the chip vendor is getting set to launch, said sources familiar with both company's plans. Like Sun's current line of Opteron-based servers, the new hardware will run various flavors of Linux, as well as Solaris or Windows.
AMD's dual-core processors double the performance of each server processor, in essence allowing an eight-way box to act as a 16-processor server, a four-way server to perform as an eight-processor box, and so on.
While Sun initially will launch an eight-way Galaxy server -- the vendor's first eight-way Opteron server -- Sun also is readying one-, two- and four-way Galaxy boxes, according to sources. Eventually, the Galaxy servers will replace the current line of Sun rack-optimized Opteron boxes, the Sun Fire v20z and v40z servers.
Sun also will leverage the Galaxy server designs for its next-generation blade servers, sources said. Last month Sun discontinued its current line of Sun Fire B100 and B200 blade servers in favor of a next-generation line currently in production, said Graham Lovell, a senior director in Sun's Network Systems Group. Sun will stop taking orders for the current blade family at the end of March, he said.
A spokeswoman from Sun, Santa Clara, Calif., said the vendor does not comment on unannounced products. However, Sun executives have already stated plans to launch an eight-way server based on AMD Opteron processors sometime in 2005.
Sun, which has traditionally depended on hardware sales to drive its revenue, has undergone a makeover in the past several years to reposition itself as a company providing business solutions comprised of hardware, software and services. But Sun has struggled to convince the market of its new role, and has seen revenue fall off sharply and its stock price languish in the $3 to $5 range for the last 2 12 years.
In the past year alone, Sun also has drastically revamped its OS and server strategy to stymie the loss of revenue to competitors IBM, Hewlett-Packard and Dell, all of which began selling commodity servers running Linux long before Sun got on the bus. Last year, Sun cozied up to AMD as its primary chip partner, and officially discontinued its Intel-based server line last month in favor of exclusively selling AMD- and Sparc-based systems. Some familiar with the company's plans said Sun may eventually phase out its proprietary Sparc servers in favor of an entire AMD-based line, though Sun executives have never officially stated this intent.
Sun's move to providing more industry-standard server technology, as well as plans to open-source its traditionally proprietary Solaris OS sometime this year, show the vendor is returning to the original strategy of building products on standards-based technology -- the one that made Sun a hardware powerhouse in the first place, solution providers said.
"It's back to the future," said one VAR familiar with Sun's plans. The solution provider, who asked not to be named, said the fact that Bechtolsheim is behind Sun's new standards-based server strategy is a boon for the vendor. and will give customers it may have lost to competitors an extremely compelling reason to return to buying Sun boxes in favor of Dell, HP or IBM.
"Here you have Andy Bechtolsheim creating industry-standard boxes that run any OS and have a kick ass-architecture, [as well as] are higher performance and [provide] lower cost of acquisition," said the VAR. "And with Andy doing his magic, you'll have boxes that architecturally are probably a lot more intelligent than most of the other volume boxes out there. How can people buying x86 not only take a hard look at Sun's Opteron systems, but take a hard look at Opteron running Solaris?"
http://www.ctlcorp.com/10,000.pdf
Both of these productions were graciously provided by Intel Corporation.
Stephen Moll(steve@ctlcorp.com)
What a surprise... not.
How much did Intel pay for that article? LOL. Notice the little white box man is from Portland, Oregon. Hmmm, who else is in Portland?
It may change, with HP rolling out many new Opteron products on Monday, including 2 blade servers, an SLI dual Opteron workstation, and the high-volume HP DL 385.
Also, the new Rev E Opterons will see an impressive boost over Rev CG, mostly in 2, 4, and 8P servers, with HT freq going up 25%. Also, they may manage a ~5% IPC boost over Rev CG, thanks to other enhancements. And of course, they just bumped up another 200MHz to 2.6GHz.
Let me rephrase that then, as "be aware that the data is somewhat misleading (perhaps unintentionally) unless you look closely."
Revision E Opteron 252 launching Monday with 4 new HP products:
Workstation, DL 385, and two blade servers.
Here's the article:
http://news.com.com/HP+to+debut+new+Opteron+servers+Monday/2100-1006_3-5571645.html?tag=nefd.top
Hewlett-Packard plans to unveil on Monday its first blade servers using Advanced Micro Devices' Opteron processor, as well as a more conventional server and a workstation that use the new chip.
HP last year let Opteron into its ProLiant server family, which until then only used Intel chips. AMD in 2003 pioneered the extension of such x86 chips with 64-bit memory, which permits easy access to more than 4GB of memory. But Intel followed suit in 2004 with its "Nocona" version of the Xeon processor.
The ProLiant expansion was significant in light of HP's commitment to the higher-end Itanium processor line, a rival 64-bit family from Intel that HP helped develop. However, the vast majority of x86 software, including Microsoft Windows until later this year, is still 32-bit.
On the same day and also at the LinuxWorld Conference and Expo in Boston, Intel and AMD are expected to unveil new higher-end chips for dual-processor servers, said sources familiar with the companies' plans. The Opteron 252 will run at 2.6GHz and boost the speed of the HyperTransport communications links from 800MHz to 1GHz, while Intel will release the new "Irwindale" version of Intel's Xeon, which doubles the amount of high-speed cache memory to 2MB compared with existing "Nocona" Xeon models, the sources said.
Dell confirmed it planned to launch new Irwindale servers Monday, and IBM and HP are expected to announce their own with the chip the same day. AMD, Intel, HP and IBM declined to comment for this story.
The faster HyperTransport link is expected across all AMD Opteron processors as the company moves the chip family to a new manufacturing process with 90-nanometer (billionths of a meter) features, a source familiar with the plans said. Also arriving
with the new Opteron generation will be support for the NX, or "no execute" bit that helps hinder worms and viruses and the addition of the SSE3 instructions Intel added to its chips last year to speed multimedia operations such as decoding video.
Dell doesn't sell Opteron servers, though the company is warming to the idea. IBM sells an Opteron model only for high-performance technical computing customers. The most eager partner is Sun Microsystems, which has passed HP as AMD's top Opteron customer and which plans an eight-processor server and other new systems in coming months.
Intel and AMD are planning new dual-core processors that combine two processing engines on the same slice of silicon. AMD's are scheduled to arrive first, midway through this year, while Intel's won't reach high-volume production until 2006.
However, some customers, computer makers and software companies won't have to wait until 2006 for dual-core Xeons. Intel is expected to announce Monday that it will send thousands of systems with the new chips to those business partners and premium customers in the second half of 2005.
[ Please, a few *thousand* systems a couple months early? A PR stunt. ]
HP is expected to offer two dual-Opteron blades, the BL25p with the 2.6GHz Opteron 252 and the BL35p with the 2.4GHz Opteron 250. The BL25p is larger than the BL35p, but offers more storage capacity.
The BL25p is expected to cost $3,499 with one processor and 1GB of memory and $5,209 with two processors and 2GB memory. The corresponding BL35p prices are $2,899 and $4,209.
According to an HP Web site, the workstation can accommodate dual Opteron processors and up to 16GB of memory. It uses dual nVidia graphics cards employing the graphics chipmaker's Scalable Link Interface, or SLI, technology, which connects the two into a single, more powerful graphics subsystem.
Note that they cheat in P4's favor, comparing a 3.2GHz P4 to a 3800+ A64, the latter still using significantly less power. They do note in a footnote that a P4 3.8 would use 50-70W more.
I also wonder if they have Cool'N'Quiet enabled for the idle measurement of the 3800+?
And, of course, Winchester has much better power char. Too bad they didn't throw in a 3500+ Winchester.
Also, the "high P4" (actually mid-range) has only 1/2 the memory of the A64 system. 512MB vs. 1024MB.
LOL. "Powerful players". No doubt the same ones beaming mind-control rays into your head from above.
AT = CG core, but 940 pinout. You'll see "AT" is the code for 1xx rev CG Opterons, as well.
AMD's capacity increases substantially as 130nm converts to 90nm. So there's your capacity expansion by mid-year.
(Consider 193 / 114 = 1.69 and 144 / 84 = 1.71 times as many die per 200mm wafer for K8 parts. For 130nm K7 to Sempron K8, it is 101 / 84 = 1.20 times for Barton and 84 / 84 = 1.0 times for Thoroughbred.)
But what makes you think it was capacity that limited K8 in 2004, and not demand?
AMD stock is not generally a volatile stock...
Um, the stock was at $3 in Spring 2003.
but what was all the hoopla in november driving the stock well into the twenties and then wham back to $15
The stock briefly approached $25, but had a period in December in the low 20's. With tax-selling in the new year, it dipped below $20 in early January.
Then AMD had a big warning, there was something of an overreaction to ~$15 for a few days, and now the stock is back around $18.
The "hoopla" is volatility. I'm sorry you lost money on your options. But that doesn't mean there's some giant conspiracy afoot.
It's not odd. It's called "volatility".
OT: Yeah but Q4 GDP kinda sucked compared to expectations. And of course, when you simultaneously cut taxes and spend like a wild-man, it does tend to boost the economy. Too bad the Republicans have become the party of big spending, big deficits, and theocracy.
I doubt there will be 130nm X52 Opterons. And of course we DO have both 130nm FX-55 (now) and apparently 90nm FX-55 in the near future. The point was merely that the OPN Mike got from the disti listing cannot be correct for a 90nm part, as 90nm is Rev D and later, not rev CG.
kpf, it is rather improbable that the 4 x52 listings with identical OPNs are incorrect. And it is not possible for a 90nm FX-55 to be "Rev CG". I think it is much more likely that they would copy over the old OPN than that they would magically invent a "90nm FX-55" entry when they already have an FX-55 entry.
In short, no need to get all Keithish on us.
Yeah, okay. BTW I edited the post to add info about what the correct OPN should look like for a 90nm FX-55...
That's not a 90nm part. "AS" is code for Rev "CG" core, which is by definition a 130nm part.
Anything else listed for those 90nm FX-55 entries? I suppose they might be using the old code if they don't have the new one yet...
http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/30430.pdf
The 90nm OPN should be something like:
ADAFX55D__5B_ where the "B_" will indicate which 90nm core rev.
Alternatively, the other format would be:
ADAFX55B_BOX where again the "B_" will indicate which 90nm core rev.
Well that analyst isn't thinking very long-term, is he? If flash continues to have troubles it ensures that AMD will get rid of it.
If Intel is really enabling 64-bit on Celeron, I predict there will soon be a boxed, retail version of XP Pro x64.
Windows XP Pro x64 is available for public download in RC1 form, and RC2 just went out today to beta testers, and supposedly will be available to the public soon (but that is according to THG.)
There is no "home" version (yet).
No gamer/enthusiast is going to want Smithfield/Smithfield-EE after they read the reviews and see the benchmarks. The software just isn't there yet to take advantage of the second core in the gaming area.
I am just "noodlin" here too, but I guess that I am probably correct in saying that you are probably incorrect.
In 2006, perhaps Ixse is only willing to consider 64-bit capable parts. In that case, Yonah is written off.
Interesting? I guess if you find Smithfield being crushed by DC Opteron interesting...
DC 2.4GHz Opteron = 2 x 250 = 2 x "4000+"
Top Smithfield is 3.2GHz x 2.
Both have 1MB L2 per core.
No, AMD never said they had an inability to supply parts. They commented on unexpected PIB demand, generally. And the part was widely available in Q4. It's only recently that they've dried up. You were making a big deal out of this last week, and now we know the reason: they were getting rid of 130nm parts in preparation for the 90nm launch.
The first proper review.
It doesn't matter. It's about perception, and futures. Geeks about to spend ~$1000 on a new CPU won't want the one lacking the new instructions. All I'm saying is that AMD should be more concerned with getting rid of CG FX-55s prior to a Rev E FX-55 intro than they should be prior to a Rev D intro. I'm not saying the FX-55 dry-up is conclusive, just that it might be a piece of evidence in favor of E stepping.
To summarize the evidence:
- First we know that 90nm must be D or E stepping. So the FX-55 90nm must be one or the other.
- We also know that the Opteron 252, at the same clock (2.6GHz), is also 90nm, and a later core rev than current D4 90nm cores, due to its "BL" core encoding in the OPN.
- Arguing in favor of E vs. D stepping for both of these parts:
-- Why release a NEW D stepping (252 "BL" different than rev D4) a couple months before E stepping needs to be launched?
-- D stepping parts have only been seen up to 2.2GHz so far.
-- D stepping was indicated to go no further than 2.2GHz by the recently leaked slides. (Although they would be called into question by an early E stepping FX-55 OR a 2.6GHz D stepping, so the slides are in question.)
-- FX-55 CG dry-up favors E stepping
-- Public instruction docs updated on 2/4/05 to include SSE3 instructions, part of Rev E parts.
- Arguing against:
-- Leaked slides seemed to indicate a later San Diego launch, (but perhaps this was only the FX-57).
-- Leaked slides showed Venice (Rev E A64) launch in late April, (but Opteron and FX could move to Rev E before the A64).
-- While instruction list has been publicly updated, the 128-bit media instruction document is stale, lacking SSE3 details at this point.