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.
Joe, sgolds - Thanks! eom
64-bit Linux speeds IC design tool (from Pravin on SI)
http://www.eedesign.com/story/OEG20040115S0011
By Richard Goering
EE Times
January 15, 2004 (9:05 a.m. EST)
Latest Headlines
EEDesign
#
Synopsys lawsuit impacts Nassda quarterly results
#
'Obfuscators' render Verilog, VHDL unreadable
#
Monterey swings to profitability
#
EDA validation startup wins $10 million in funding
#
How EDA can improve '04 outlook
#
Tanner adds design tool
#
C-based co-development environment rolls for FPGAs
Archives
SANTA CRUZ, Calif. — ReShape Inc. is reporting significant speed and capacity increases for its PD Optimizer tool suite on 64-bit Opteron and Athlon processors running Linux — even compared to faster 32-bit processors. PD Optimizer helps manage IC placement and routing runs.
ReShape is also reporting eight new benchmarks that claim PD Optimizer reduced die sizes by an average of 15 percent, reduced inter-block interconnects by more than 25 percent, and reduced global buffer/repeaters by more than 40 percent, compared to designs done by hand.
Over a cross-section of benchmarks, ReShape's Design Aware optimizer ran an average of 1.7 times faster under the 1.8 GHz AMD64 architecture, compared to a Xenon 2.8 GHz 32-bit architecture, said Bob Dahlberg, vice president of business development at ReShape. Even though the AMD64 processor is slower, it has superior caching, Dahlberg observed.
In addition to speed, he noted, the 64-bit AMD architecture permits higher capacity. On one benchmark on a 4.6 million instance hierarchical system-on-chip (SoC) design, ReShape's tools, running on a 1.8 GHz Athlon machine, prepared 20 individual blocks for place and route from a new netlist in six hours.
Design Aware looks at the most recent chip build and makes changes to improve it on the next run. With each new netlist or design change, it accommodates the change and optimizes the result. After three or four builds, ReShape claims, the pins, repeater placements, and timing budgets converge, creating optimal results.
PD Optimizer for Red Hat Linux Version 3 with AMD64 processors is available this quarter starting at $350,000 per year. Macromedia's FlexLM for 64-bit Linux is required.
sgolds - The link worked fine for me. eom
Could someone explain "max pain" to me, please? Perhaps "define" would be a better word.
TIA,
Paul
yb - need a good laugh? Look at this!
http://www.siliconinvestor.com/stocktalk/msg.gsp?msgid=19694062
I believe DRBES has lost it! :D
Paul
sgolds - The edit window does truncate the long URLs, but you can re-paste the long URL into the edit window and it'll accept it. :)
Paul
Storage Computer Announces CyberNAS 64/Cluster Primed for the AMD Opteron Processor and AMD64 Architecture
(from DRBES on SI)
http://home.businesswire.com/portal/site/google/index.jsp?ndmViewId=news_view&newsId=20040115005...
LinuxWorld New York 2004
NASHUA, N.H.--(BUSINESS WIRE)--Jan. 15, 2004--
Storage Computer to Display Software at Linux World in New York
Storage Computer Corporation (AMEX: SOS) (www.storage.com), a provider of high-performance storage and data management software, announced today it has released CyberNAS 64/Cluster network attached storage software for the high performance computing market.
"The AMD Opteron(TM) processor with its advanced Hypertransport technology powers CyberNAS 64/Cluster software into ground breaking storage management terabit performance and petabyte usable online capacity," stated Sean Murphy, VP Product Development. "CyberNAS 64/Cluster software supports up to 300 storage nodes per cluster with projected aggregate bandwidth up to 10 terabits per second on multiple 10Gigabit Ethernet connections. Its capacity is scalable to 9 petabytes."
"With servers based on AMD Opteron processors and CyberNAS 64, customers have a powerful, flexible storage solutions for even the largest clusters. Especially for customers managing large datasets, such as digital video, audio or graphics, AMD Opteron and CyberNAS 64 provide a compelling solution," said Ben Williams, director, server/workstation segment, AMD's Microprocessor Business Unit.
"Breaking out of the limitations of 32 bit architectures allows our customers to move to the next level of storage performance and functionality," stated Ricardo Velez-McCaskey, CTO of Storage Computer. "It is imperative for HPC (high performance computing) environments to have strong storage options designed to accommodate their challenging needs. CyberNAS 64/Cluster is available now to compute cluster manufactures on an OEM basis."
Storage Computer will be showcasing CyberNAS 3.0 at Linux World 2004 (http://www.linuxworldexpo.com) in NewYork, NY January 21st-23rd in the AMD Booth #413. The Storage Computer CyberNAS software will be demonstrated on Microway(R) Navion(TM) Servers (www.microway.com), which are built with AMD Opteron processors.
A high performance, feature limited, version of the CyberNAS software, CyberNAS RED, is available at no cost from Storage Computer at www.storage.com.
About Storage Computer Corporation
Storage Computer Corporation (AMEX:SOS - News), delivers high performance storage management solutions, develops and manufactures software-driven, multi-host storage solutions for powering advanced business applications. Based on open system commodity hardware components and architectures, Storage Computer's high-bandwidth storage technologies supports a great variety of applications, including advanced database activities, interactive multi-media, multi-cast video content, medical imaging and more. Company information may be found at http://www.storage.com.
This press release may contain forward-looking statements to future events or future financial performance that involves risks and uncertainties. These statements are made pursuant to the safe harbor provisions of Section 21E of the Securities Exchange Act of 1934. These statements are only predictions and actual results could differ materially from those anticipated in these statements based upon a number of factors including those uncertainties and risk factors detailed from time to time in reports filed by Storage Computer Corporation with the Securities and Exchange Commission, including our most recent reports on Form 10-K and 10-Q.
Contacts
Storage Computer Corporation
Todd Viegut, 603-880-3005 x183
tviegut@storage.com
Wasn't IBM supposed to release their results before market open today? Anyone heard?
Paul
Edit: In answer to my own question:
http://www.thestar.com/NASApp/cs/ContentServer?pagename=thestar/Layout/Article_Type1&c=Article&a...
IBM nearly triples Q4 profit at $2.7B
BRIAN BERGSTEIN
ASSOCIATED PRESS
NEW YORK — IBM Corp.'s fourth-quarter earnings more than doubled and surpassed Wall Street expectations today, continuing a streak of good financial news from technology companies.
In the last three months of 2003, IBM earned $2.7 billion (all figures U.S.), or $1.55 a share, on revenue of $25.9 billion.
Those figures all surged from the comparable period a year earlier, when IBM posted a net profit of $1 billion, or 59 cents per share, on revenue of $23.7 billion.
IBM's earnings from continuing operations were even better — $1.56 per share. Analysts surveyed by Thomson First Call were expecting $1.50 per share and revenue of $25.0 billion.
IBM's report came less than 24 hours after the technology sector saw strong earnings releases from Apple Computer Inc., Intel Corp. and Yahoo Inc., which all indicated that the long stretch of soft technology spending by businesses may finally be waning.
"We enter 2004 with good momentum," said Sam Palmisano, IBM's chairman and chief executive. "The client buying environment is steadily improving. We are enthusiastic about our prospects for this year and beyond."
IBM signed $17.3 billion in services contracts in the fourth quarter, a big increase over the third-quarter figure of $15 billion. On IBM's last earnings call, in October, the chief financial officer, John Joyce, had said services bookings in the fourth quarter could be around $14 billion, and some analysts even recently predicted that the company would fall short of that number.
"This was a very good quarter for IBM and an encouraging end to a year in which we steadily gained momentum and posted record revenues," Palmisano said.
"The company's performance would not be possible had we not made investments during the downturn and fundamentally repositioned IBM for leadership in the high-value enterprise space, which remains our sole focus."
However, while IBM's revenue rose nine per cent in the fourth quarter, much of that gain resulted from weakness in the U.S. dollar. Without currency fluctuations, the increase would have been just one per cent.
For all of 2003, IBM showed net income of $7.6 billion, or $4.32 per share, on revenue of $89.1 billion. In 2002, net income was $3.6 billion, or $2.06 per share, and revenue was $81.2 billion. That 10 per cent revenue growth would have been 3 per cent without currency fluctuations.
The earnings report from the technology bellwether was originally scheduled to be released next Tuesday afternoon. IBM spokesperson John Bukovinsky said the report was moved up simply because the company had finished preparing its numbers earlier than expected.
Keith - re: Toshiba. I can't swear to it, but I'm pretty sure they have not.
Paul
This should be an interesting article when it comes out at tecchannel.de.
"19.01.: AMDs mobiler Athlon im Detail
Unterwegs mit 64 Bit: AMDs mobiler Athlon bringt 64-Bit-Power in die Notebooks. Wir erläutern die Details der stromsparenden AMD64-CPU."
Translation;
19.01(19 Jan) AMD's mobile Athlon in detail.
Underway with 64bit : AMD's mobile Athlon brings 64-bit power to the notebook. We cover the details of the power saving AMD64 CPU.
Paul
Welcome, Amkeer! Haven't seen you since I stopped visiting RB, before it went insane over there.
Paul
Keith - maybe so, or maybe AMD picked up what Intel left.
sgolds - re: "All will soon be known."
Yes, indeed! ;)
Paul
Keith, sgolds - It will be interesting to see how the two FX53s, 939 & 940, are different. I can't imagine 939 won't have a measurable effect; so do they want two FX53s performing differently?
Paul
CJ - Thanks. So what kind of software might be able to use an even higher bandwidth with low latency to main memory? Something that hops around a lot with big chunks of data so that the caches would have to be refilled? That is the kind of situation I meant to refer to when I referred to "bandwidth starvation".
Is there any important software with a big or strategic market where Opteron could capitalize on the mem controller-to-core bandwidth advantage, if they could increase the bandwidth to main memory?
Of course, I guess they've already done it in a way with multiple CPU systems, but I was thinking more in the single CPU arena. It might become even more important when they come out with the dual-core chips; 2 mem controllers per chip would be a lot of pins!
Paul
Mysef - Yes, the e325 is a dual Opteron server.
Paul
IBM - Bristol Meyer Squibb Opteron sale. Don't know why they call it "massive", it's 64 of the e325 servers. Ashlee can't seem to get her prefixes straight, she uses both x325 and e325, but in each case they specify Opteron.
http://www.theregister.co.uk/content/61/34859.html
Paul
kpf - OT
you sure you're not just having a bad day? The thread didn't seem that bad to me. I'd give the guy another chance. :) And I know I've missed replies/posts before, and it kind of alters the whole conversation.
Paul
Doug - Ed probably got peeved that AMD locked the multiplier for K8; after all, he is an overclocker! ;)
Paul
PS. And now they've done the same for XP!! That must really tick him off!
CJ - You're getting close to what I'm looking for here. So what is the bandwidth to the caches of(for the sake of nailing down 1 figure) a 2.2GHz K8? What's the width of that bus? Did they increase it to 128bits from the 64bits(?) of the XP?
In looking at the 64FX data sheet, it says the L1 is capable of 2 simultaneous 64bit loads/stores, etc., so I'm assuming this means they did change to 128bit bus. So that would mean 2.2*128= 281.6Gb/s or 35.2GB/s, right?
So what are the other internal factors/housekeeping/overhead that would limit the theoretical in practice? Assuming unlimited DDR bandwidth for the sake of discussion.
Just how "memory starved" is K8? And could the core keep up with what it could theoretically be fed?
Paul
PS. Feel free to deflate/reveal any ignorance on my part here.
mas et al - That part I knew. What I don't recall is the width of the memory controller-to-crossbar bus, or for that matter the throughput/configuration of the crossbar to the rest of the core. I guess what I'm wondering is what the max theoretical throughput is, although that's kind of hard to nail down with changing frequencies, isn't it?
Paul
jjayxxxx - Sorry, guess I should have read the rules better. I sent my entry by e-mail. Well, here it is again:
Name:j3pflynn
Team:iHub
AMD CONTEST
(a) Total Revenues = $1150
(b) CPU Revenues = $645
(c) Flash Revenues = $505
(d) EPS (diluted) = $0.290
(e) Closing Stock Price = $18.20
INTEL CONTEST
(a) Total Revenues = $8710
(b) Intel Arch Group Rev = $7680
(c) Intel Commun. Group Rev = $1030
(d) EPS (diluted) = $0.300
(e) Closing Stock Price = $34.00
Anyone know offhand what the theoretical peak bandwidth memory controller to core is in K8? Is it more than dual channel PC3200 DDR(1) can provide?
If so, I expect that every bit of bandwidth and latency advantage that 939 can offer will be used; unlike nForce2 where the EV6 "bus" was already saturated. Dual channel PC3200 matched P4 closely, though, but didn't seem to pay off like one might expect.
Time to start digging, I believe.
Paul
another Xeon-Opteron workstation article coming up on www.tecchannel.de (16Jan04)
"16.01.: Xeon vs. Opteron in Workstations
AMD hebt die Taktfrequenz der kleinen Opterons auf 2,2 GHz an. Dadurch hat AMDs Opteron 248 selbst Intels Xeon mit 3,2 GHz eindeutig im Griff."
AMD raises the clock frequency of the little Opterons to 2.2GHz. Thus AMD's Opteron 248 clearly has even Intel's Xeon 3.2GHz in its grasp.
Doug - While I often disagree with much of what Ed writes, I do find he's willing to criticize equally, though it's been directed more toward AMD lately. I find him useful to engender some critical lines of thought that may not have occurred to me, taking it all with a grain of salt.
Paul
Keith - No, but my mother was, so I grew up learning German once I was about 5 or so. Never having had formal training, my grammar and my technical vocabulary are quite weak, but I'm pretty comfortable speaking it. Without too much opportunity to speak it, though, I get pretty rusty; so I'm tickled that my son has really taken to it in school, so I can work on it with him.
Paul
PS. I still have a lot of family over there in Bayern, as well as in Zuerich and Bern, Switzerland. Can't afford to get over there much, though. :(
AMD vs. Intel 2003 patents 904 - 1595.
from c't
"... Bei der interessanten (und - wie erfahrene Geflüsterleser wissen - weinschweren) Frage Intel gegen AMD braucht man indes Intels zahlreiche Firmenaufkäufe der letzten Jahre gar nicht mit hinzuzurechnen - allein mit den unter „Intel Corp“ verzeichneten US-Patenten liegt der Marktführer im Jahr 2003 wieder klar mit 1595 zu 904 in Front, nachdem sich der Konkurrent in den vorausgehenden drei Jahren vor Intel platzieren konnte."
rough translation of bolded section:
the marketleader(Intel), with 1596 to 904, is clearly in front again in 2003, after the competition having been able to place ahead of Intel in the preceding 3 years.
Paul
IBM will use Newisys 4way box.
http://www.theinquirer.net/?article=13586
"SOURCES thousands of miles from East Fishkill tell me that IBM is set to follow Sun and will resell Newisys four-way Opterons systems in the future."
Paul
Keith, it sounds like a good fit to me; I suppose there could be at least 2 basic reasons for not doing it:
1) IBM already has other customers lined up(perhaps not publicized yet), and therefore won't have the capacity.
2) They can't come to a price agreement that is beneficial to both.
Paul
sgolds - Yes, you have a point there. An AGP - PCI Express conversion on the card, like nVidia is doing, is a lot easier than having a converter card with the extra connection and height,especially with the "big dog" cards that might make someone want to stretch out their "investment".
Paul
yb, I imagine latency is more of an issue when you have multiple "devices" contending for memory/CPU access, whereas with video cards, they're all by themselves, and throughput is more of an issue.
Paul
PS. BTW, for any interested here's an PCI Express overview pdf from Intel.
http://www.intel.com/technology/pciexpress/devnet/docs/PCI-Express-Overview-Oct2003.pdf
Doug - receive, yes; send, no. eom
kpf - IIRC, PCI Express supports a greater per-pin bandwidth than aHT, and is supposed to be better suited for backplane/interconnect applications than aHT, but I'm no expert on the matter. I think aHT may have the advantage in latency, but that's not so much the issue with graphics as with CPUs.
Paul
yb, I'd be surprised if someone doesn't make themselves a good business out of either making adapter cards to allow use of AGP cards in PCI Xpress slots, or making motherboards that support both an AGP and PCI Express slot. There's alway someone ready to capitalize on the slice of the market that just upgraded and wants to use their equipment longer.
Paul
yb, that does sound promising. eom
Paul
Andy, Doug - I suspect Keith's point is that, even if someone claims to be as Intel employee(of whatever type), that doesn't mean they really are. How someone represents themselves on the internet is not necessarily reality, which is why he encouraged focussing on the content.
Paul
Keith - out of curiosity, since it's been a while since I've perused an official AMD pricelist, was there any change in teh 8xxx pricing?
Paul
chipguy, my mistake. Thanks for the correction.
Paul
dougSF30, certainly a poster's bias/viewpoint matters regarding how you interpret it, but anyone who's been around here for a little while know already. Anyone who's new(even free accounts) can certainly look at a poster's historical posts quite easily and get a very good idea of where they're coming from, without tying up bandwidth or constantly bringing on these exchanges that turn into name-calling matches.
Paul
dougSF30 - re: PMs - free account.