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.
Got a new job writing code?
Well good luck to yer, what 20 years now?
Oh I've just got in from work. I see you have some questions for me.
"Paging is a lower level feature than selectors. They really have very little to do with each other, which seems to be your confusion. Now, you mask off the two low bits of a selector and use the result as a byte offset into an LDT or GDT. From there you can get the base address, limit and associated information about the behavior of the memory range. The processor does this automatically. Thus, for example, if a ring 0 selector base address is 0x200000 and its offset into the table is 0xC0 then the address 0xC0:00000000 points to the physical address 0x00200000 by means of this processor translation."
No, this is wrong - you're thinking about the old segmentation model and ignoring paging. What you have described in the generation of the *linear address* not the physical address. What you say is only true on old non-paging systems.
In the segmentation model the address is formed from a selector and an effective address to generate a linear address. If, and only if, paging is disabled the linear address is the physical address.
If however segmentation is EFFECTIVELY disabled by setting the selectors to base 0x0, limit 4GB then the effective address is the linear address. NT kernel does just this - its the flat model.
To help everyone the flow is:
selector+effective (via GDT or LDT) -> linear -> paging -> physical
"Segments are an artifact of 8086 addressing. They can be directly shifted and added to an offset to get a flat address. Selectors are used in a similar manner in segment registers (CD, DS, ES, etc.) but are used (as described above) by the processor as offsets into a GDT or LDT to generate a flat address by the OS. You can not directly translate a selector into an address, you have to do a table lookup. This lookup is only available from ring zero instructions (available only to the OS and ring 0 drivers)."
No the segment registers contain the selectors which index into the GDT or LDT. They are not "similar" or "related to" to segment registers, they are the content of the segment registers. Perhaps this is why I wonder. . . .
And onward:
"1. There certainly are selectors in Windows. You can not execute a single x86 instruction without them - your assertion is absurd."
Yes segmentation cannot be completely disabled but the init phase creates selectors that all have base 0x0 and limit 4GB so the effective address *IS* the linear address and effectively they are irrelevant. That's the flat memory model. So any linear address can be accessed as code or data by context (either a data reference or a code reference). But its not absurd, just that CS, DS, ES etc are irrelevant in this model and we only have to worry about the effective address which becomes the linear address. You don't have to create a selector to access memory, its all mapped for you (but paging provides process separation). See, programmers don't have to worry about segment registers in W32. If the designers could have dumped them they would have. In AMD64 they have been dumped, except for some minor useful legacy bits.
And onward:
"2. Selectors do not address all of memory, they address a virtual 4GB address space. You are having a hard time with this concept! Different applications are given different selectors and can not physically read or write into another application's space without creating a shared selector to point to a shared area of memory (generally data area). A long time ago Microsoft said this is bad programming practice in any case, and developed Named Pipes as an alternative (which work whether the two processes are on the same machine or across a network)."
No this is wrong again. Segmentation was the original way to protect user spaces but given descriptors in the tables that span the whole 4GB they all have access to *ALL" of the linear space. Its paging that provides separation of process spaces. That's the advantage of paging. No banging on descriptors to allow naughty reads and writes between processes. That's why they *had* to use pipes. And they used to bang, like XORing bits in the descriptors to allow writes, that was the system call to make a segment writeable I remember.
And onward:
"This is the core of your misunderstanding. It does not work that way, different processes have different base addresses and only a virtual address of 0x00000000. Sorry about that, but you are just plain wrong on the facts."
Doesn't matter what linear space each process has, the linear address under paging is *local* to each process, paging maps it to physical.
And onward:
"Grow up! Come back in a few years when you've had a chance to understand your mistakes and we can chat again."
Whoops did my little comment upset you? Really I hope you learn something from this.
And lastly:
"P.S.: Did you really think that all applications overlay each other at address 0x00000000? Don't have to answer that here, this is a question for thought."
Linear or physical? Why not linear? Just some address that the descriptor tables map to some arbitrary physical page. You have to be more precise in your questions. Each process gets its own linear in NT.
There I've answered your angry points.
Apology accepted.
You see I wasn't being obnoxioux, just accurate.
Writing an OS for a processor w/o an MMU is actually quite interesting. Even primitives like semaphores are challenging. The Z80 was the first mass-market device that made it possible. Win3.0 was an amazing exercize in making software do the work of an MMU.
Well believe what you want. I tried to point out where he was fundamentally wrong and I was met with flat out rudeness. Read my original post.
I am senior to him and his experience and am published in the x86 world. I am certainly not an application writer though I have written quite a bit of code though mostly at the system level. I have written an OS from scratch (how many can say that!) albeit on a z80. It was used for real-time data ack and embedded but later used for a controller-based GPIB i/f as well. That was in the days when z80 was king.
While there are sometimes grey areas in interpretation I do not expect misquotes and condescention. His analysis of the CS & NX scenario was flat-out wrong. When I pointed out the role of selectors, or lack of, in NT kernel there was a whole lot of scrambling going on in case you didn't see it.
I am grateful that he has decided to stop putting out misinformation.
Your moniker may well be right.
"I generally try not to give advice to those who will not learn."
We should all be so lucky, finally you understood that you have nothing to offer. Frame your prognostications about CS & NX, they're good for a hoot.
Good, then its time you learned to RTFM instead of spouting nonsense.
I just can't stand arrogant little punks who try to cover their mistakes. You're ten-a-penny in the business.
Better still why don't you give us all a laugh and explain just how GDT, LDT fit into the paging model. Let's be precise how exactly do your selectors that you crow about actually accomplish a virtual->physical maping in a paging based system? What are these selectors you are talking about actually DOING in your picture of the VMM?
Don't tell me they are essential. Tell me why the OS designers initialize them as they do. Answer the question why they are not actually disabled (I know the answer).
Perhaps you could give us an insight into how you believe selectors and segmentation are just "related". Are, somehow, selectors part of your view of paging?? Is the concept of flat memory model even dimly in your understanding?
Come on, give us all a review of page translation tables and where the bits actually are. What bits have relevance in 64-bit mode?
Perhaps you could enlighten me where I was seemingly confused between physical & virtual memory? As I read me post I made no such reference. Now you claimed a load of crap about NX and the use of CS selectors. What, pray tell, happens to your world view when all selectors have the same base and a segment limit (32-bit) of 4GB. So any code can be executed in any space?
I think you're full of it.
There are no selectors in Windows anymore. Segmentation was abandoned in favor of paging in the NT kernel. All of the selectors address the full 4GB in Win32 and overlap.
The 32-bit mode of the Athlon64 does support legacy use of the selectors in both legacy and compatibility.
AMD64 "long-mode" largely disables segmentation, except for specialized fs & gs usage.
NX protection mechanism
You said:
"So, decaw, as I understand it NX does apply to code written to a data buffer. A lot of the media buzz talks about overwriting into stack areas, but that is a subset of what NX does. It protects pages irrespective of whether they are stack entities or data buffers (or even code segments)."
Well implied in this statement is whether stacks and buffers are *always* protected. Microsoft elects to establish stacks with NX protection. It doesn't have to, that's how disable DEP for named apps works. Similarly a buffer can be allocated (on a page basis) that is either NX protected or not. I think I recall that there is a new DOCUMENTED system call for this purpose. Things like resources (icons, version details, strings) may also be NX protected if required.
Of course you are in error in saying CODE segments can be NX protected. That would defeat the whole point I think. Code segments must be execute-enabled else what would they be good for?
I cannot see any basis to claim that a reboot is an efficient way to recover from an NX violation. It must be a bug because the #PF handler is wrong (probably because its not yet initialized in this case). In driver code the #PF NX handler should simply force the driver to return an error. There are many ways for drivers to return errors, in this case failed to initialize.
Rebooting is simply a bug, nothing more.
NX kernel issues & reboots
Well hypothetically then where is the issue? Obviously any memory violation during the driver initialization phase is going to be a problem ifthere is no trap set. NX falls into the general group of page-protection checks. An NX violation raises a #PF (page-fault) exception, just like a write to a read-only space - except the NX is tested when the instruction TLB is loaded rather than when an instruction is executed as is the case of a write fault.
So the page-fault exception handler is evidentally not loaded during the driver-init phase but the PAE paging mode has been enabled. That's a very narrow window when things can go wrong
I want to emphasize there is no magic in the NX bit. It is a very clean extension that uses existing hardware in a conforming way.
Well let's make that clear. Its not applications that pass NX-protected addresses into the kernel, we're only talikng about drivers that execute from an NX-protected buffer or stack. Thus there's only ONE driver that does this?? Mpegport.sys. That's amazing. All of the Win3.1 graphics drivers wrote code on the stack, it was part of the reference DDK. Plus it seems that the fault lies in the initialization code, only called at startup. I would think that run-time code would have the illegal access trapped.
This whole thing is a NON-ISSUE. It applies to just ONE driver. If an app passes an NX callback to the kernel it will NOT prompt for a reboot any more than if an illegal address were passed
NX doesn't mean that you can't run code from a buffer.
I'm surprised at you for saying this if you have any assembler experience.
NX can be applied where you want it. In the case of WinXP64 the stack is NX-protected (optionally you can use NX anywhere you want). Thus you can't just use the stack if you want to write executable code from your program. If you want to do that you will have to allocate some space elsewhere and mark it as Exec, Write explicitly. I'm sure a lot of programs do allocate a buffer explicitly and not just rely on the stack being available.
What's this nonsense about a reboot being needed if NX is violated. BS. That's not at all how it works. You just get a violation like any other badly behaved program trying to access memory illegally, it uses the same trap. You only get a reboot with megaport.sys because there is an unhandled exception in the driver routine at startup. That's a bug by Microsoft, pure and simple. If there was an NX exception in an application like Outlook Express there's no way a reboot is reqd or triggered.
I wish people here would read the freely available manuals on AMD64
and where did you get these stupid numbers from? Your canary pad at the last intel marketing meeting? Keep this feeble-minded rubbish where it belongs, on your pad. Don't insult the readership with this drivel.
intel is spinning all it can against NX and SP2. They see they have a disadvantage and are working feverishly to underplay its importance. We have intel marketing reps working here. The more they expose themselves, the better.
If its so unimportant let's hope that intel drops its plans to implement an NX bit.
Sounds to me like their engineers are having some trouble.
4GT technology for 32bit applications must be seen as a 64-bit benefit too. intel's repeated claims that there are no "64-bit" apps are nothing but fud, however the ability to simply give 32-bits apps 4GB to play in does give rise to "super-32bit" applications that are also part of the equation.
So we have, under WinXP64:
32-bit
super 32-bit
64-bit
Now it would be interesting to run some benches on big memory user apps, like Maya and games, set for 4GT.
BTW there is a whitepaper on 4GT at amd.com
http://www.amd.com/us-en/assets/content_type/DownloadableAssets/Expand_Memory_of_32-bit_App_-_Micros...
go to http://crystalmark.info/?lang=en and download CrystalCPUID
Changed AMD mobo display at Frys
Went to Frys yesterday. They have moved around their motherboard display. Used to be that the motherboards immediately to the right of the chip display cabinets were all Intel. Then much further along were the AMD ones. When Athlon64 came out those motherboards were relegated to the extreme right of the display.
It was like this:
INTEL INTEL INTEL ....... AMD AMD AMD AMD64
INTEL INTEL INTEL ....... AMD AMD AMD AMD64
Well all has changed. The AMD motherboards are now where Intel used to be and occupy a full 2/3 of the display area. Intel motherboards are now where AMD used to be only there are far fewer of them.
Its now
AMD AMD AMD AMD AMD64 AMD64 ... INTEL INTEL
AMD AMD AMD AMD AMD64 AMD64 ... INTEL INTEL
Frys also is offering a full range of AMD64 chips.
I guess there's one reason for the change eh?
Inventory Valuation
Some real misunderstanding of Accounting 101 going on w.r.t what physically is in AMD's inventory assets.
Notice that FASB rules mandate that inventories are carried on the books at the LOWER of cost or market value.
Thus, if you have a pile of 2001 playboy calendars sitting on the shelves its likely that their inventory value is about 10c even if your cost was $1 UNLESS they include Dick Cheney's bit-on-the-side.
For high markup items like chips nearly all of the inventory will have lower cost than market value, then the pieces in inventory will (99% probably) be at the cost of wafer+fab costs+packaging (if finished parts). For Athlon64 parts this may be $50ea. Then a $50m inventory will represent 1m parts.
AMD had a stated inventory of $726m last quarter, compared to $692 in the 1st qtr and $697m at the end of 2003. A lot of that must be wafers and unpackaged die. The rise of $34m in inventory in the last qtr I take as very good news and an indication that yields were substantially up.
Solaris on itanic
Begs the question - how much is Intel paying for this port?
I think rather a lot as they must be trying to curry favor with Sun, to keep their Xeons in the x86 line and stymie AMD64
I really like the mobile Barton in the system. Previously I had an 1800+ in there and it was hot, here in California that can make things pretty uncomfortable.
The performance is pretty amazing - I do some video editing on that machine. It was a toss up whether I went Athlon64 or Barton but I just didn't have time to do a complete motherboard swap.
Some of the big manufacturers should look into using mobile chips in cool desktops maybe. Kind of a "thin & light desktop"?
Microsoft's settlement with the state gives up to $1.1bn of vouchers to be spent on any computer hardware & software. I think the vouchers will be sent out before Xmas. Should be a nice boost to the industry. I'm getting $250 back and that will be going to a 939 Athlon 64. Details of the settlement can be found by googling Microsoft California Settlement.
I believe desktop Bartons are now multiplier locked. Also the old GA-7DX mobo is a 133MHz bus so I needed a mobile. I bought a Barton 2600 for $99.
"decaw,
Athlon Mobile Barton PowerNow on Desktop
So this works only on Mobile Bartons, not Desktop Bartons?
Joe"
Athlon Mobile Barton PowerNow on Desktop
Hmm, got this working well and processor runs at just 10C above system temp on a really old Gigabyte GA-7DX.
Here's how to do it (on W2k):
Install the Powernow driver from the AMD site.
Get CrystalCPUID from crystalmark.com, and install - easy.
Enable multiplier management and put the command line into startup. Set up the thresholds for upping multiplier and voltage.
It runs in the systray and gives multiplier (and voltage) increases as processor demand increases.
Thanks for the info mmoy. Now I know that I don't have to throw away my VS6 ENTERPRISE I'm a lot happier going to WinXP64
Visual Studio 6 generates AMD64 code?
Did I miss this? I thought a new compiler was needed.
According to this article you can dload a new version of the SDK from MSoft with a 64-bit toolchain and set it up with an existing VS6 compiler.
http://www.devx.com/amd/Article/21313
Am I stupid or was this not made public before?
Intergraph Payment of $10m is ONE-TIME
Therefore it must not be included in earnings/share.
Then earnings beat by almost 3c.
Why is AMD sandbagging????
Compare Sandisk and Apple today.
Similar companies except SNDK is a one-trick pony and AAPL is riding on retail sales of one item, the iPod.
Hardly recipes for long term growth - and what do we see? Both are strongly up after their earnings.
AAPL has revs, ttm, of 7bn. AMD will have revs at least 5bn this year. Market cap of AMD is 4bn, AAPL 12bn. AAPL has 5bn cash tho' while AMD has 1bn so call it a 2:1 ratio, still that would put AMD at $25.
SNDK has revs 1.3bn, ttm, and a cap of 4.1bn, cash 1bn
When there's blood in the streets and people are crying the sky is falling - that's the time to buy.
Volume record
I do believe intc will set a volume record today, certainly a $ trading record.
Anyone know better?
Those buybacks of intc stock look awfully expensive right now, that should further dent their intel capital profits which were estimated at $60m for last qtr and only came in at $39m.
Dell also has inventory problems??
According to this Dell's inventory of notebooks is too high. Guess they're not selling either.
http://www.digitimes.com/news/a20040714B2053.html
Seems that the Intel disease is spreading. Watch this space.
Will sales be up 100% YOY?
Q2'03 sales were $645m
$1290m will be double last years qtrly sales, a nice statistic.
In Q1'04 AMD forecast Q2 sales to be approx flat with Q1'04, at $1236m
At least we're dumping the -40c loss of Q2'03, now we'll have a P/E to play with. Today's earnings + 15c from trailing 9months
Personally I think $1.3bn is in the bag, ~ evenly split between memory & CPU. And an eps north of 22c/share.
Nocojones looks like a very early alpha part.
That the CPUID function disables its use on XP64 tells me that intel really didn't want any reviewers trying it and afaik the Linux port of iAMD64 is just in development.
Right now its a 32-bit chip that may run most AMD64 instructions properly. Enough to send out a few samples to developers to test CPUID code and do the switch from legacy to long mode. Who knows maybe they've just shipped the developers a simulator, like AMD did some years ago.
It looks a long, long way from shipping. Years even. Intel must be screaming, yelling and begging MSoft not to lauch XP64.
The bug list must have lots of so-far blank pages at the back in a 3-ring binder.
Its a channel check.
You said:
"Dell IS testing opteron, but not just for speed, but for demand, ease of manufacture, ease of part availability, LIKE they always do. But they will not hop on the bandwagon until they see mass acceptance, which they have not seen, I am guessing."
SIG is reporting based on their checks in **channel**. That's talking to suppliers, not to Dell. Testing would be for 10s of units that would never show up in and investigation. However if the channel is saying that it means that Dell is buying xolumes of product and the asscoiated parts.
Dell would never buy an existing design like the Newisys. They want functional and cheap. The Newisys design had bells and whistles, including a PowerPC mgmt monitor on board. That's not Dell's way, they want reliable, low cost for their mainstream, not like the big Compaq boxes.
Microcode Patch Mechanism
AMD has had a processor microcode patch facility in their processors for some time, its in the XP.
You can see here:
http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/27532.pdf
That the model 10 AthlonXP revision A2 actually had a bug in the mechanism that made it unreliable (errata 22).
That there are no listed errata in the Athlon64 microcode patch category suggests to me that it works and that's how a BIOS change will fix the problem. Cheap and easy.
BTW reverse string movs are much rarer than forward string moves and even then they are normally prefaced by a mov ecx, somevalue instruction as the count of the repetition, which is not a microcoded instruction. Thus it is unlikely that the bug would cause real harm compared to software bugs that happen all the time.
SAMSUNG!
This is the important part of the PR release on Mirrorbit. Samsung is #1 in NAND flash and now the biggest (I think) seller of cellphones.
For Samsung to get behind this new NOR flash is HUGE. Shows that it meets their needs better than their in-house made flash.
"dual core test wafers went through the fab already several weeks ago"
Well, said with a confident air, like someone who was there?
So you're saying - correct me if I'm wrong - that the tapeout was completed and the test batch passed muster? That was the reason for the PR piece?
So the next step would be twiddling the masks to optimize speed & yield?
Sorry for so many questions. Your post was short enough to be enigmatic. Just a generic overview of the necessary steps - and where AMD could fall behind if something goes wrong - would be more than welcome.
Thanks DDB
AMD COMPLETES design on dual-Opteron
The PR piece says that design has been completed. Only Anandtech equates this with the start of the tape out process.
Does anyone know if these are necessarily sequential steps? Wouldn't tape out perhaps have already started? Isn't tapeout part of the design process? I mean Anand is showing pictures of what look like test parts already.
To my mind the design is the tough part, tapeout is surely a quick step once the design is complete. Testing and parameterization will be the lengthy jobs yet to be done.
Some informed clarification would be welcome.
If design is complete is the 2H05 timeframe paced by development or available fab capacity?
Second of 2 Nvidia PR pieces.
Again not linked to AMD.
*************************************************
NVIDIA and AMD Deliver Improved Security to Protect the PC Desktop
Wednesday June 2, 9:01 am ET
NVIDIA Hardware-Optimized Firewall Solution and AMD Enhanced Virus Protection technology enabled by the upcoming Windows(R) XP Service Pack 2 Can Provide PC Users with Added Level of Protection Against Common Attacks
# COMPUTEX - TAIPEI, Taiwan, June 2 /PRNewswire-FirstCall/ -- Recognizing the need to thwart network level hacking attacks and other security risks, NVIDIA Corporation (Nasdaq: NVDA - News) today announced collaborative efforts with AMD and leading Taiwanese motherboard companies to help promote enhanced secure PC platform technology solutions and deliver innovative features and increased levels of protection for desktop PCs, workstations, and servers based on the NVIDIA nForce(TM) and AMD64 computing platform.(Logo: http://www.newscom.com/cgi-bin/prnh/20020613/NVDALOGO )
"Recognizing the downtime and costs associated with cleaning up after network infiltrations and virus attacks, it made perfect sense to both us and AMD to build security features directly into the silicon level," said Drew Henry, general manager of platform business at NVIDIA. "The combination of the NVIDIA nForce MCPs with the AMD Athlon FX and AMD Opteron processors provides end users with the best possible platform for performance, features, and protection."
Already, NVIDIA and AMD have each incorporated cutting-edge security features at the silicon level to help combat the direct threat of networking- level hacking attacks and other common security risks. NVIDIA's solution -- NVIDIA Firewall -- is integrated directly into the Company's newest NVIDIA nForce media and communication processors (MCPs) for AMD Athlon(TM) and AMD Opteron(TM) processor-based motherboards. It provides driver-based protection the moment the PC is connected to the Internet or internal network, as well as a graphical user interface allowing customers to configure the Firewall, easily modify settings, and manage the security settings remotely. AMD64 technology processors, which include the AMD Athlon 64 FX and AMD Opteron processor families, incorporate Enhanced Virus Protection (EVP) technology that will be enabled by the upcoming Windows® XP Service Pack 2 to help eliminate certain security risks and viruses by prohibiting common "buffer overflow" attacks. The combination of NVIDIA nForce MCPs and AMD Athlon 64 and AMD Opteron processors provides users with a great baseline level of protection. NVIDIA and AMD are both committed to providing customers next- generation security technologies in order to make the PC platform as secure as possible.
"PC security is one of the most pressing and frustrating issues facing computer users today," said Marty Seyer, vice president and general manager of the Microprocessor Business Unit, AMD. "Viruses impact performance. Without hardware to address viruses, even great performing systems are potentially relegated to performing like 20-year old PCs. AMD and NVIDIA recognize the frustration and have done something about it. AMD's Enhanced Virus Protection technology, which will be enabled by the upcoming Windows XP Service Pack 2, is an ideal complement to NVIDIA's Firewall. Together they strengthen the PC platform to help withstand the ever-increasing number of computer attacks."
NVIDIA is also working with the top Taiwanese motherboard manufacturers, including market leaders ASUSTeK and MSI, to help market their line of nForce- based motherboards featuring integrated security features for the AMD64 market. Together, the companies will help educate consumers on the importance of PC security and begin promoting built-in security as a "must-have" feature for today's PC enthusiasts.
"ASUSTeK is dedicated to providing users with the best technology possible, we are as concerned about security as we are about performance," said Joe Hsieh, Director of ASUS Motherboard Business. "Together, NVIDIA and AMD are helping to bring innovative security technologies to the market at the silicon level, providing end users with a certain level of comfort and protection right out of the box."
"The fact that we are able to provide these innovative security features at no extra cost to our customers is truly revolutionary," said Mr. Rick Wu, Motherboard Product Marketing Manager at MSI. "By providing a combined, protected platform, NVIDIA and AMD are truly developing the next-generation in secure networking and computing."
Major motherboards utilizing the security features offered by NVIDIA and
AMD include, but are not limited to:
-- ASUS K8N-E Deluxe
-- Epox EP-8KDA3+
-- Gigabyte K8NSNXP-939
-- iWill DK8N
-- MSI MSI7025
-- MSI MSI7030
*********************************************
http://biz.yahoo.com/prnews/040602/sfw006_1.html
First of 2 Nvidia PR about AMD
Strangely not linked to AMD by stock symbol on Yahoo.
**************************************************************
NVIDIA nForce Professional Solutions Adopted By Leading Server and Workstation Motherboard Manufacturers
Wednesday June 2, 9:01 am ET
IWILL and Tyan Utilizing NVIDIA nForce Professional MCPs for AMD Opteron-Based Motherboard Solutions
# TAIPEI, Taiwan, June 2 /PRNewswire-FirstCall/ -- COMPUTEX -- NVIDIA Corporation (Nasdaq: NVDA - News) today announced that the Company's NVIDIA nForce(TM)3 Professional media and communications processors (MCPs) have been selected by IWILL and Tyan Computer Corporation -- two of the industry's leading professional motherboard manufacturers -- for their new high- performance server and workstation motherboard solutions.(Logo: http://www.newscom.com/cgi-bin/prnh/20020613/NVDALOGO )
Targeting the single and multi-processor AMD Opteron(TM)-based professional computing markets, IWILL and Tyan have selected NVIDIA nForce3 Professional MCPs for their highly integrated, single-chip design, and advanced features and functionality, including hardware-optimized security solutions, system stability, and IT manageability.
"Our NVIDIA nForce3 Professional 250-based DK8N workstation motherboard with support for dual AMD Opteron processors is fast becoming one of our most successful products ever," said Sunny S. Chang, Sales and Marketing Director at IWILL. "Working with industry core-logic leader NVIDIA enables us to develop cutting-edge products for the professional user and provide them with innovative features, performance, and stability unmatched by other professional solutions."
"With their renowned success with their NVIDIA Quadro line of professional graphics solutions, and now NVIDIA nForce3 Professional MCPs, it is clear that NVIDIA is on the forefront of technological innovation and we are excited to be working with them to develop products based on their next-generation platform technology," said Danny Hsu, Senior Director of Marketing at Tyan. "With NVIDIA, Tyan will be bringing professional users the compatible, stable, and reliable 32-bit and 64-bit computing platform they have been asking for."
Designed for professional computing environments, the NVIDIA nForce3 Professional MCPs incorporate advanced technology solutions, including enterprise-class networking, fault tolerant storage designs, and optimizations for NVIDIA Quadro workstation graphics solutions -- which are currently supported by an industry-leading selection of certified software applications. The NVIDIA commitment to product longevity further ensures IT investment protection.
"We are thrilled to have partners of such high caliber as IWILL and Tyan -- two leading suppliers of motherboard solutions to the retail channel and OEM professional markets," said Drew Henry, general manager of platform business at NVIDIA. "Professional users demand performance, stability and reliability from their workstation and server platforms, and we're excited about the possibilities that IWILL and Tyan will be enabling in the professional computing market."
IWILL is currently shipping products based on the NVIDIA nForce3 Professional 250 MCP, and NVIDIA is working closely both IWILL and Tyan during their development of new products based on NVIDIA's next-generation silicon, which will include additional support for PCI-Express and other new features.
************************************
http://biz.yahoo.com/prnews/040602/dew002_1.html
BROADCOM
Just want to point out the BRCM has a $13.3bn stock valuation and less than 1/2 the annual sales (ttm) of AMD.
Meanwhile AMD has a valuation of $5.5bn.
Hmm suggests a proper valuation would be more like $25bn, even discounting the PHENOMENAL GROWTH AND POTENTIAL OF AMD.
Sorry, that was shouting.
BROADCOM
Can I have a bigger font. Make it more bold.
What else do you want.
Its a strategic alliance with first product this fall.
Spin this intel.
Re: AMD to Present at Smith Barney Citigroup Semiconductor Conference
This is the same day and a few hours before the Intel mid qtr update, June 3.
That'll do a lot to stop any downward momemntum on AMD due to either a lowered guidance from Intel or an upward guidance from Intel.
Now what us investors would like to see is obvious <g>.
Geode NX processors.
Well I expect to see these in thin'n'light laptops very soon. I'm sure that Gateway/E-machines will use them in very short order as they said that they would be attacking the low end of the market, price-wise, with coming models.
This will collapse the pricing for ULV Pentiums in small laptops, like the Dell Latitude C400s.
Good move in adding these to the Geode brand in my opinion. Though Geode is currently thought of as embedded it was previously the bargain range for PCs. This keeps the Athlon moniker for premium range products and segments the market neatly without creating yet another brand that has to be supported. This way the Geode NX supports the old Geode name, pulling it up from near obscurity, and positions the range firmly in the value segment.
Wonder if Geode has a separate sales force?
They should also invigorate the set-top box/shuttle/home entertainment biz. I think this will do very, very well both as a revenue generator and to push Intel out of a profitable segment.
2600+ Replacement for 1800+
Well I finished installing a Barton-core 2600+ XP mobile in our old desktop which has a Gigabyte GA-7DX motherboard and had an AthlonXP 1800+ before.
I was going to rip out the innards and install a '64 and probably should have done so but time was a factor and I wanted it as easy as possible.
Found the GA-7DX had a new bios that supported up to 2600+ so flashed that first. Had to use XP-M because the old motherboard is only 133FSB and desktop Bartons use 166 (333 DDR) and up.
Bought the OEM processor at Newegg, amazing delivery - free overnight Fedex - for $99. Then had to go to CompUSA for those heatpads. $3.99 for 3!
Removed old heatsink and scraped off old pad, installed 2600+ w/ new pad, no problems.
Booted at 800MHz. No way in bios to change multiplier or voltage. Finally found CrystalCPUID which is amazing and allows everything to be changed including multiplier & voltage, even from a command line. Works a treat in the startup folder. So runs at 15x and 1.45v after boot. An alternative multiplier change utility, CPUMSR, just froze the machine.
Its much nicer than the old 1800+, runs a lot cooler. For video editing its that much faster that you can get a whole lot more done thanks to larger cache & multiplier.
For $110 incl tax and heatsink pad its def. a worthwhile upgrade. Its a tossup whether it would have been better to go Athlon64. I just didn't want to spend more than a couple of evenings getting everything running and there's a lot of stuff in & attached to this box. My nightmare would have been putting together a '64 and having some kind of instability pop up that could have prevented work next week.
So I think I did the right thing.
I've used the Hitachi (IBM) Travelstars and they have been fine.
Buy a little IDE-> laptop adapter, $5.99 I think at Frys, and use a desktop to xfer the data. Really easy. No gotchas. Remember to do the UDMA66 fix for Windows 2000 if that's what you're using.
Oh and I found that having an 8MB disk buffer on the drive made a lot of difference in a desktop system - not sure about laptops. I certainly would not go for a 2MB buffer on a drive ever again.