Saturday, June 30, 2007

K8L , K10 , K10.5 or ... something else ?

But, do you have any proof that Barcelona and Budapest cores are really called so? No. !

K8L is Charlie Demerian scrap "invention". K10 not in 2007, nor 2008? Perhaps dead? Haha.
What he could hear about is only Griffin mobile processor.
K10 is AMD's marketing baloon
K10.5 is an scholar example of FUD.

So what is the truth?. We desperately need it. Well, all of it and nothing of it.

K8L name is correct in indication that it is not true K10 design. Everything else is wrong.
K10 name is correct on its future persepctives, but not on its current architecture state of the art.
K10.5 is actually what K10 is initially intended for. In 45nm technology. But,
AMD is late in technology, behind Intel, and suddenly crippled original design
became christened at AMD marketing as K10.
Alas, that design lacks integrated DDR3 memory controller, SSE4, and due to backward
comaptibility uses for now only 2GHT per second FSB. Why faster FSB without SSE4?. Why
SSE4 support without faster FSB?. Beside, old 65nm technology doesn't allow clock scaling
up too much, at least in 4 Core implementation. And, why it should? 45nm comes in 2H2008. Why to bother?
Lets announce it as low power design architecture and wait for 2H2008.
Revision C in 45nm will encounter all what is missing now.
Thus, K10.5 should be real
K10.
AND WHAT ABOUT CURRENT K10?
My suggestion is that it should be called (the most appropriately): K10_
;)
Beside, do you know any application package that supports SSE4 instruction set now?
But would be at disposal ina year, when 45nm will be ready at AMD. Does it mean AMD will always lack behind Intel in SSE support. Well , no. SSE4 seems is the last SSE sequel.
What will happen after? look at the accelerators boxes below. Definitely, AMD has a future.


Real K10 is a real monster and Intel killer. However,
Intel hopes Nehalem will save its future in 2008/9 .

EDIT: Someone from Russia posted this important link (with love) as a comment.
Oh, by the way, seems that Digitimes confirms my humble reasoning.

AMD intends to modify its process technology to increase the yield rate and frequencies of Phenom processors in order to improve the competitiveness of the CPUs against Intel's Penryn family, said the sources. AMD has already notified several motherboard makers that Phenom will now begin test-production in November or December, and will launch in first quarter of 2008, noted the sources.

Thursday, June 28, 2007

AMD Barcelona hits 2,5 Ghz on October 15th !

If TACC can get enough Barcelona chips from Advanced Microsystems by October 15, its system will land near the top of the next Top 500 Supercomputers list, Sun says.

The Inquirer reports the following Barcelona on October:
2360
2.5GHz
120W
TBA

The third part of the Sun riddle is this: No, Sun Microsystems' 33.25-inch-tall Sun Blade 8000 is not too big, company executives say. (big blade for high 120W TDP)

Thus, we have double size Sun Constellation blade (compared to IBM's and HP's) and if some 13 600 2,5 Ghz Barcelonas are provided by October 15th we will get some 420 Tflops at the University of Texas.

That will immediately drive B1 Barcelona supercomputer among the world's few top fastest systems. Is anything needed to say more on Barcelona performance? But let estimates its performance. 420 Tflops divided by some 13 600 Barcelona chips gives on average 30+ Gflops per chip. Yes we have 4 cores inside, but the problem at supercomputers is in communication efficiency (up to only 25%) not in processors. Thus, still nothing spectacular. But is Sun actually waiting for the first AM3 versions so called Budapest? That would push Constellation switching system performance from 2 to 5,2 GHT. Anyway. Current implementation should enable it in a year. But, revision B2 in 45nm should run faster than 2,5Ghz and dissipate lower than 120 W. 13600 times 120W gives 1,5MW. However, you have to add at least 1,5Mw for the other system parts, so you get some 3MW overall power consumption. The system is packed with 1.7PB of disk storage, and when finalized, will have 105TB of memory.
Each Constellation configuration is centered around a design that accomdates 48 blades per rack with 16 cores per blade. Hmm, pretty similar design with Cray's. Some 72 racks conected by Infiniband switch Magnum gives overall 420Tflops. Some 6 Tflops per rack. Pretty impressive.

Oh, by the way, 2,5 Ghz Barcelona is known under the name Intel killer. What to say for B2 revision? Something deadly like B2 bomber.
Intel needs 3.2 Ghz parts or more to match Barcelona / Agena at 2.5 GHz, so 45nm manufacturing is needed to be in place or Intel's famous low power design should draw more than 120W TDP.
Sort od shoot in its own foot. Thus, lets go 45nm says Intel. We have 5 quarters advantage in 45nm. However, let calculate together, between 1Q 2008 when Intel will start to deliver 45nm and 2H2008 when AMD will, is 5 quarters? Intel is seems wrong in grammar, it should betetr use verb had advantage in 65nm. In 45nm it is less than 3 quarter, but not more than 3 quarter. Beside, Intel's investments fall year by year (just as its margins), so it is not clear how it will maintain its technology leadership in the future?

Wednesday, June 27, 2007

AMD hits stellar 2 Petaflops in Sun's Barcelona Constellation

If TACC can get enough Barcelona chips from Advanced Microsystems by October 15, its system will land near the top of the next Top 500 Supercomputers list, Sun says.

Sun says the Constellation features a total 131,072 processor cores, can operate at up to 1.08 petaflops, has 100T bytes of memory, and an I/O speed (the speed at which data can be moved in and out of the central processing unit) of 3T bytes per second. A terabyte is 1 trillion bytes.

Sun chose the AMD Opteron quad-core chip, code-named Barcelona, over rival Intel's quad-core Xeon 5300 chip, because it believes AMD's will be "the fastest chip on the market this year," said Andreas Bechtolsheim, chief architect and senior vice president of the systems group at Sun.
Constellation is targeting IBM, which held an industry-leading 47.8 percent market share on the earlier Top 500 list, because its Blue Gene line is "the fastest machine out there," Bechtolsheim acknowledged. But Sun's Constellation is built with open industry-standard components, such as x64 processors, Sun blade servers, Sun Fire x4500 storage, and open source Solaris or Linux operating system software, while IBM's Blue Gene is built from custom-made components.

Intel Core Duo in 64bit reliability troubles


This is a reliability update. Install this microcode update to improve the reliability of systems with Intel processors. After you install this item, you may have to restart your computer.


DON'T PATCH IT, AND YOU WILL GET A PERFECT HARDWARE STORM FROM INTEL, FREE.


A COUPLE OF WEEKS ago, we heard that Dell was dealing with a certain situation considering Intel dual-core MCW and quad-core KC marchitecture, and that the company was releasing urgent BIOS and microcode versions for its line up.

We learned that the affected CPUs are the Core 2 Duo E4000/E6000, Core 2 Quad Q6600, Core 2 Xtreme QX6800, QX6700 and QX6800.

In the mobile world, people with the Core 2 Duo T5000 and T7000 need to visit Microsoft's site, while the server guys will want to use motherboard BIOSes if they do not rely on Microsoft Windows operating systems.

The affected servers are Xeon 3000, 3200, 5100 and 5300s - or just about every model from the second generation of Core marchitecture.


IBM goes 3 Petaflops


IBM claims that its new Blue Gene/P supercomputer, operating at petaflop speed performs more operations than a 1.5-mile-high stack of laptops.

The 1-petaflop Blue Gene/P comes with 294,912 processors and takes up 72 racks in all. Hitting 3 petaflops takes an 884,736-processor, 216-rack cluster, according to IBM. The chips and other components are linked together in a high-speed optical network.

Monday, June 25, 2007

Man should go to suicide Mars mission !!!



Who says? Edwin "Buzz" Aldrin, the second man walked on the Moon.

Where he said that? In France.

Who said it earlier? Theoretical physicist and astrobiologist Paul Davies.

And even earlier? Science fiction writer Jules Verne in 19th century in his novel From the Earth to the Moon. As you know, Victorian age technology wasn't enough to conquire even the Moon.
So he used brave people to get the Moon and REMAIN there, because the only available technology wan't enough to return them back.
The names od those people: Michael ARDAN (pretty close french spelling of Aldrin), Nicholls (quite close permutation to Collins). Please note one more permutation of Ardan and Collins names !.

Where they were from?
From France too.

How they reached the Moon orbit?
By super gun.

Who created the best gun at the moment of Verne Moon mission concieved? Englishman Armstrong.

Where Moon shot was made?
From the place only 300 kilometers from Cape Canaveral. Well, that was called good prediction. Some 100 years earlier.

Is today exists any plan to shoot supplies by gun out of the Earth? Yes, there is. By two stage shells.

Moon people with the main Marsian, during the 35th Moon landing anniversary.

The only really unsolvable problem, as I see?
The men you initiated current NASA Mars mission initiated the tide of terror suicide missions world wide. Would be there some patent right infringements on suicide mission?
How the patent disputes would be solved. What if one of sides wouldn't be satisified with the court descision. How to solve that conflict peacefully?

DAAMIT: Phenom troubles for Intel's management

More troubles on Intel. For Intel's current "winning" strategy essential ingredient was assumption on AMD's bad performance. Seems that system performance now on are controlled by I/O speed,
not primarily by clock speed, as in the PR age was the case. From current 2GTps HT, future Phenoms goes double to 4,2G HT3 in 1Q08. Opterons will go up to 5,2G. Quite enough to keep Phenom overall performance rise of the complete AMD line . And AMD's shares too, by the end of this year.



Sunday, June 24, 2007

Cutting all around, Intel got a wound from its own sword

While Intel Chief Executive Paul Otellini has grabbed market share, he's still fighting AMD--and sacrificing profits--to get it. After a string of gains for AMD, Intel took 80.2% of the worldwide microprocessor market in the first quarter of this year, up 4.5% from 75.7% during the prior quarter, according to market researcher iSuppli. However, Intel's operating margin for the quarter ending March 31 fell to 18.9% from 19.2% during the year-ago period. That's well off Intel's historic range of between 30% and 40%.

What will happen with margins now when all Intel's 16 000 manufacturing people are out and some fabs closed? What will maintain high margins in the future? Monopoly? No.
Technology advances? No,
45nm will last only 6 months. Architecture advances? Intel's 32 bit optimized architecture compared to Barcelona? Volume advantages? No. AMD now has two fabs, second source in the Far East.
An old Bosnian tale says: Who kills with sword, dies from it in the end.

Moreover, AMD is keeping the pressure on, promising it will win back some of that lost ground this quarter.Scrappy AMD has a history of surprising Intel by defying the numbers and cranking out a better chip, with fewer resources. If AMD's Barcelona is a hit, Intel's price cuts will look prescient. I mean if AMD surivive (and it will) it will be impossible for Intel to rise again its margins to the golden monopolistic age heights. I am curious, if Intel's margins diminish who would finance Intel's ambitious new fabs? Israel, like in latest Kiryat Gat fab?
BUT BE AWARE, in that scenario, only this will left to Intel's visionary management board:


The following map when to be realized, would drive them mad too.


Chapter 7 + Chapter 11 gives immediate 50 Tflops


Seems that University of Bergen acadamia stuff doesn't read The Inquirer and Ass hardware, very "confident" circumized hardware sources. They recently "found" Cray is dead after Barcelona delay and AMD filed some chapters 7 and 11, after being inherently unable to make Barcelona, so Intel will produce their 64bit processors in the end.

"We are honored that the University of Bergen chose Cray as their partner to provide very high-end computing capability to support the university's extended research community," said Peter Ungaro, Cray's president and CEO. "This decision strengthens our international presence and fortifies the Cray XT4 supercomputer's leadership in meeting the demanding needs of leading-edge scientific researchers for unrivaled performance and scalability."

Those people had a chance to chose, and still made an unbelievable error like that? Or, those two "sources" are simply bunch of the circumized lyars?

More on Bergen computing community here.
Petter E. Bjorstad +47 55 58 4171 petter@ii.uib.no


Saturday, June 23, 2007

AMD Barcelona benchmarks in July, LIVE !

AMD QUAD core Opteron roadmap

On July 1st (after 5 years of development), for $259-$399 / Month you can hire the following 8 cores for testing and work (based on quad core K10 Barcelona):

Opteron 2258HE 1,9 Ghz 68 Watts
Opteron 2260HE 2,0 Ghz 68 Watts
Opteron 2262 2,1 Ghz 95 Watts
Opteron 2264 2,2 Ghz 95 Watts
Opteron 2266 2,3 Ghz 95 Watts
Opteron 2268 2,4 Ghz 95 Watts

If so, you could buy Barcelona, perhaps even on August 1st. Wait a week to check.

Friday, June 22, 2007

Defense Secretary Robert Gates has no email !?

Er, "I don't do e-mail. I'm a very low-tech person."

BAD HARDWARE would like to suggest that the problem is likely more in another issue.
Lack of trust is on display in NSA organizational structures [and] behaviors across the Enterprise.

Let me put things this way: Former CIA director don't do email.
Al Qaida low ranked soldiers drive Boeings like a byke.
Who will win the war on terror? The most likely, technically superior side.

Briefly Trailblaser project failed after 6 years, and now in plase is Trailblaser 2.0
What the hell is Trailblaser? Project for your personal email and conversations recon.

Thursday, June 21, 2007

DAAMIT: AMD Phenom roadmap !

AMD's shares are on the abrupt rise.

  • AMD targets Barcelona to launch in Summer 2007 and expects ramp in frequency and volume subsequently. We can expect low clocked Barcelona in August timeframe
  • Phenom series to be launched in Nov-Dec with DVT samples in Sep/Oct timeframe
  • Phenom FX AM2+ part called FX-80 in Q4 timeframe and TDP for the FX series are yet to be determined
  • 45nm Deneb, Propus, Regor and Sargas STARS cores in H2 2008
  • Single-core Lima has been removed from the roadmap so single-core Orleans will be replaced by dual-core Rana in Q1 next year. The mainstream segment will shift from single core to dual core by early next year
  • AMD expects to provide next schedule update in July

Wednesday, June 20, 2007

DAAMIT: AMD Barcelona on mid August, up to 2,4 Ghz

That will be enough to show Intel its 4 teeth.


Models will start at 1.9Ghz that was demoed at Computex.
Top Barcelona would reach 2,4Ghz at launch. FSB is 1Ghz.
And Bad Hardware predicted its top clock 2,5 Ghz to completely defeat Intel.
If launched in mid August, that means Barcelona production QUIETLY started, because for each chip production fab manufacturing process needs at least 8-9 weeks. Production quantities by mid August.
We had slight time frame slip, but isn't that quite normal for complex products?. Where else you can find 4 cores in a single die? Intel will get some, but only by brute force in 45nm, half a year later. But after another 6 months, Barcelona in 45nm would follow. So? Who is the king of the X86-64 hill?

We are eagerly waiting for Phenoms too, some three months later. That should be in the second half of November. Just in time for Christmas rush hours.

DAAMIT: AMD invests in a new fab


DAAMIT. Intel should make AMD designed processors there? Idea is risible, as Mike should say.
All those laughing are of course nutz.

Intel rises as The Inquirer falls

The Citigroup analysis also notes that AMD plans on selling one of its two microchip plants in Dresden, Germany.

Some stupid analysts seems don't know that AMD's old 130nm fab goes to Russia, with love. Cache goes in the opposite direction towards NY, with love too. Strange behaviour for The Inq reportedly fabless company.
Let me notice that Intel's cache goes in the opposite direction to the somewhat noisy Middle East, precisely Kiryat Gat, Israel.

Sunday, June 17, 2007

Intel rises as The Inquirer falls

It is unbelievable what a number of lies launch The Inquirer, formerly distinquished hardware online journal. The latest is on that Intel will manufacture AMD's chips. AMD has its own fab in Dresden for top designs, and Chartered second source for the lower ones. AMD has ready 45nm production for the 2H08. Thus, why should anyone else manufacture AMD's designs?. Especially Intel, whose manufacturing process is absolutely incompatible with AMD's. Well, AMD something shares, The Inquirer reports. So what , it shared ie. its advanced virtualization to the others, and that doesn't mean AMD will leave virtualization to Intel, as The Inquirer stuff would lie. Goldman Sachs protege Intel, and by the way, they know the best? Well, some analysts during the worst Intel's fall have given recommendation to buy Intel, so we stupids have to listen them again? Oh come on, please. Those people are seems the same who bought The Inquirer, during the business visit of Magee to Israel. That explains The Inquirers' fall. Man, The Inquirer was the only hardware institution we had in this lying world. What money could buy it? Now, no one seems left. Lyars heaven remained. World of sex, lies and benchmarks.
Personally, Mike thinks the idea that Intel would fab AMD chips is risible and there is presented as a joke.

I accept it as a joke of course,I got it first as a joke, but not in the context of the Goldman Sachs claims. Those Goldman people earn enoromous, exponential money on their claims, so, The Inq writings shouldn't be considered as a joke for us, poor people, in any context. Simply, small hardware man currently sucs too much to Goldman Sachs and similar, only to make some rich people even richer. I still remember well Wall Street analyst's recommendation buy Intel during Intel's the worst fall, not so long time ago.

And finally, my comment on Mike's comment: Intel launched not so long time ago rising idea on 20Ghz Pentium by 2005/6. But, I didn't suck it (at least up to the projected moment). Would you suck something so fast rising? Old Bosnian tale says: Nothing flies so fast as the lie, but nothing falls as fast as the lies fall, too.

EDIT: Idea on Mike's visit to Israel is not proved from two independent sources, so please take it as a joke for now. I don't want to be a lyar too. I would really hate myself than.

Friday, June 15, 2007

AMD new architecture quietly launched !


The launch event in New York comes after almost a year's worth of hype and promise from AMD.
Something will have to lift AMD to assure the company's future. It has cut costs and jobs in an attempt to become profitable by the end of this year, but analysts have warned about the company's cash consumption rate.

Intel CEO compares AMD’s success to a “comet passing in the night
Night, that is right word for the Pentium reign. However, instead of shortly lived comet reflecting Intel wishes in the darkness more appropriate word for AMD would be fireworks given above,

Please note that event above happened on April 22nd 2003. All similarities with current affiars are of course intentional. Thanks Intel for idea to suggest Barcelona announcement with fireworks, like here:


Thursday, June 14, 2007

International Space Station computers have failed !

Finally, a truth on bad hardware. Even in the space.

COMPUTERS that control key life-support systems aboard the International Space Station have failed, and in a worst-case scenario could force all crew members to leave.

The failure of the computers, which also control navigation, is a major problem for engineers, called in to determine what caused the problem.
ISS program manager Mike Suffredini tonight admitted the computer failures could, in the worst case, force the crew the leave. But that was not something he was really worried about at this stage, and expected the problems to be fixed within the next couple of days.

"We always have an option to depart," he said.

We are eagerly interested who is the manufacturer of those bad computers?

Russian computers took control, but they failed too, so the ISS could float to open space. Hail to bad hardware.

Russian space officials were "cautiously optimistic" the problems in the German-built computers that run Russian software could be solved, a Roskosmos official was quoted as saying.

The six computers were restarted on Thursday and worked for three hours before breaking down again, Russian media reported.

No cause for the problem has been established but a leading theory cited by Russian officials has been electronic "noise" caused by magnetic fields from nearby cables that may have increased once the new solar panel was attached.


Wednesday, June 13, 2007

President Bush watch has been stolen during the friendly visit to Albania !

Our guardian angel is being robbed ! Watch out !.

Look at the President's left arm at that historical moment ! Some dirty Albanians used the moment of handshaking, when the greatest of all the Presidents has to use BOTH of his hands simulateneously. Alas, his former right hand is in jail and the left one left without the watch. Who wants to see the whole sincere video, please enjoy here.

That man delivers us supreme security. But, sleep without worries, his watch still shows 19th century time frame, as each NEOCON watch does.
What will now happen with KOSOVO independence? Nothing of it before the watch return?
KOSOVO WAS NEVER ALBANIAN. TRY FINDING ANY PROOF OF THEIR EXISTANCE IN KOSOVO OLDER THAN 100 YEARS.ALL THE SERBIAN CHURCHES IN KOSOVO ARE AT LEAST 600 YEARS OLD . That's quite correct, but in Bosnia there was almost no any Serb church some hundred years ago, however there is Republika Srpska today. Right? So the Serbs deny Albanians the same what they are actually doing in the Bosnia. Right?

Intel cuts its quad core chip price for whopping 50% !

Yeah, bad products were always low priced. Someone feels Barcelona real 4 way breeze over The Mediterranean?.
50 % price cut actually means buy our dual core - get four.
But, what for?
According to Intel partners, price reduction for Core 2 Duo processors is planned in 2 , September from 24% to 32%. Man, where are the Core margins?

DRAM goes out at IBM


IBM is pitching the invention as a technology that could provide a cheaper and more flexible way to temporarily upgrade the available memory in computing systems ranging from PDAs to servers. According to the description of the patent, the technology includes three separate parts - a connector, a container to hold RAM as well as a cable that couples the connector to the container.

IBM says that the connector can be inserted into a DRAM slot just like a common memory module, allowing the memory installed in the external box to be addressed by the motherboard as another bank or DRAM. When there is no external RAM module plugged in, no memory is recognized by the motherboard's memory addressing unit, the patent claims.

IBM does not describe the invention as a replacement for typical memory upgrades, but as an option to increase the available memory temporarily and allow users to share RAM upgrades, simply by switching the module from one computer to another – especially in devices where RAM is not easily accessible, such as notebooks. According to the patent, the RAM box can provide up to four memory slots for every DRAM slot within a PC case.

It was unclear which type of cabling IBM intends to use for connecting the connector with the external box.

IBM has not announced an actual product that makes use of this technology.

Tuesday, June 12, 2007

AMD Barcelona enters B2 revision


What the hell happens? Bad Hardware revealed Barcelona B0 revision at Cray 8 months ago. At Computex, AMD demoed only that Cray's development B0 version at non respectable 1,6 Ghz. We can't be satisffied with a 1,6Ghz chip, made some 9 months ago, right? Now, what is AMD's next move?. AMD is currently in B1 stepping (provided that the starting one wasn't B-1), and AMD internal evaluation showed that it is still not quite competitive with Intel. So, Barcelona B2 respin will be needed, and finally , but only in August or September, AMD would have at disposal an phenom competitive answer. Anyway, 3-4 months between the revisions is a phenom, furious production pace at AMD. And regarding illiterate computer hardware journos, let me to mention that B2 chip revision usually means pretty stable , mature product. Now when AMD's second , Far East fab just started first production ever, of low priced shrinked K8 chips, no need for launch of not quite finished superb products that could cut prices of the first really high volume production at AMD. People who don't understand that are inevitably Intel's arse holes or ordinary morons. Who wants glued chips in his mission critical server pool? Above is of course genuine 4 core Barca pic , seems the first ever publicly seen, and below is quite ready, 4 way AMD Barcelona board. Sorry, no revision data.




Monday, June 11, 2007

Intel double crashed in a full swing, AMD disqualified

Smaller core gives 4 times more Gflops at the same area or power.
Back to the future in 2012: Please note that half of small core is essentially 486 with 1,2M transistors in 65nm, though 150 times higher clocked then in original 1000nm implementation !. Simply, at lower geomeries bigger core shrinks become more and more inefficient in all terms, in each reincarnation . Naturally, smaller cores shrink better in the lower geometries. Will VIA be a competitor in some 5 years?



Finally, a truth on bad hardware.
Seems that Intel's Montreal high gear F1 drive caused terrible self destruction. AMD was disqualified in this race, but stayed complete. Obviously smarter choice, for now.

Personally, I believe that Intel will brake when switch into 45nm. Its low leakage technology is so fragile, just like Intel's racing car.AMD delayed its efforts to see what will happen. I think AMD will only optimize its technology in 45nm. There is no sense to do it in 9 months, provided Barcelona will start only in Q4.

Only Intel drives you without the problems !

With a little financial help of Credit Suisse



Saturday, June 09, 2007

AMD plans in Europe could complicate Iranian nuclear issue


MOSCOW. June 9 (Interfax) - The U.S. plans to deploy elements of its missile defense system in the Czech republic and Poland will seriously hamper efforts to solve the Iranian nuclear problem, Russian Foreign Minister Sergei Lavrov said.

"The deployment of an AMD system in Europe can seriously impede these efforts [on a peaceful settlement of the problem of Iran's nuclear program]," he warned journalists in Moscow on Saturday.Experts are actively working to clarify all aspects of Iran's nuclear program, including through the UN and the IAEA."Nobody has provided proof that the Iranian nuclear program has a military dimension," he said.


Eh, what about Iranian nuclear program and its first revenge target: Intel's 45nm fab in Kiryat Gat, Israel?
(With some exceptions; a 1997 U.S. law prohibits the collection and release of satellite imagery of Israel with a resolution better than two meters, for example.)
By the way, will wide deployment of AMD systems in US and Canada make North Korea nervous, as it yesterday made Intel's share holders a bit hysterical? That company AMD doesn't seem to me like a dead horse.

Friday, June 08, 2007

AMD Barcelona delayed to Q4 !?


Here is why, perhaps? But dual core K10 should be phenom. Right?
In the meantime, Barcelona bugs should be extreminated at Operating System Research Center (OSRC) in Dresden, Germany.

Additional research apparently also has been done in support of AMD's upcoming, though now potentially delayed, Barcelona core. For example, the OSRC has been involved in resolving any bugs in Barcelona's quad-core OS management before its launch.

AMD Phenom SHOULD NOT BE COMPARED WITH CONROE?

That's why it got its name / Phenom.

F1 Phenom, Montreal June 10th ?
Canadians and Italians will hail its victory dancing ( download here)?
AMD's venerable 130nm process, now not so phenomenal, goes to Russia, with love? Eh, that would fulfill AMD's need for healty cache.

With AMD revealing information about its upcoming desktop processor (Phenom), the company has stated that it delivers around 15-40% better performance than Intel's Core 2 Duo processor (Conroe). This comment has set off Matt Dunford, worldwide chief benchmarking manager of Intel, who said Phenom processors should not be compared with Conroe, they should be compared with Penryn.

Oh, poor, poor Conroe.

Intel's turbo stereoid desktops without turbo memory !

Do you know the first processor law: I am as fast as my memory is?

Why so? Surely a Robson module on desktops will not only make use of PCI-Express, it'll bring Intel more money and it'll also improve performance for everyone.
No one has yet given a concrete reason, only speculation and also stating that it may be as late as next year when we get one.

Or, Robson technology is only a tweak for bad performance of Intel's nootebooks compared to Phenoms phenom?

Wednesday, June 06, 2007

AMD Barcelona first Cinebench test !


Barcelona had to be announced on June 6th.
But, just as bad hardware found earlier:
AMD partners confirmed the highest running, POST and OS capable, Barcelona processor is 2.0 GHz.

That is why Barcelona was postponed the whole two or three months backward. It is necessary to tweak 65nm process and chip yields to at least 2.1+ Ghz clock to be cost/performance effective against the Intel. I admit that Intel is pretty well optimized towards two cores jobs.And AMD of course, needs some time to respin its design. The problem is that any additional delay requires change of corporate price policy and future strategy.

The guys from DailyTech had the opportunity to run a quick test on a quad-core AMD Barcelona PC at Computex:
We only had a few minutes to test the chip, but testers were able to run a quick 64bit Cinebench 9.5 before they were instructed to leave.


The AMD benchmark ran on a single-socket, K10 CPU running at 1.6 GHz on NVIDIA’s nForce Professional 3400 chipset. According to the system properties, the AMD system used 4GB of DDR2-667.
The most similar Intel system we could muster up on such short notice was an Intel Xeon 3220. The Xeon X3220 is clocked at 2.4 GHz, and ran on Intel’s Garlow workstation chipset (Intel X38). This system property profile stated the system utilized 4GB of DDR2-800.
Cinebench completed the default benchmark in 27 seconds for the 1.6 GHz K10; 17 seconds for the Intel Xeon X3220. The Kentsfield Xeon was 58% faster with […]

Thus, Barcelona has to be clocked up to 2,5 Ghz to run faster than Kentsfield competitor. We hope it will be in a month or two. After that, think about forthcoming 8 core dual die Barcelona. Intel will suck even in 45nm !. No NEED FOR BUDAPEST.

Speeds and Feeds

Core 2 Extreme QX6800

Penryn Quad

Penryn Duo

Codename Kentsfield Yorkfield Wolfdale
Clock Speed 2.93GHz 3.33GHz 3.33GHz
Process 65nm 45nm 45nm
Cores
4 4 2
Total L2 Cache 8MB 12MB 6MB
Front Side Bus 1,066MHz 1,333MHz 1,333MHz
3DMark06 CPU 4,047 4,944
3,068
3DMark06 Overall
11,252 11,969 11,020
Mainconcept H.264 (sec) 89.05 72.75 119.22
Cinebench 10 10,551 12,900 7,061
Cinebench 9.5
1,541 1,917 1,125
Half-Life 2 Lost Coast (fps)
153.14 213.03 206.06

Divx 6.6 (sec)

38 18
22

Shareholders, pic down is how Intel tweaks Cinebench 10 results for 45nm processors ! ! !


However, all hardware arse holes talking these days missed to notice one important thing:
AMD has to fight with Intel in two gladiator arenas SIMULTANEOUSLY. Luckily, it has now two fabs to be tweaked for two different types of market requests: high performance and low pertformance. But AMD previously has to conquire market share to make a sense of it.
And AMD did it, regardless of enormous shares fall collaterals.
So, some journo slots don't forget: AMD is alive and well kicking !.

Intel to demo 45nm Harpetown at Computex ! ! !


Nothing less than that above is promised by The InTELREquirer ( in more technical terms actually this pic below)

and somewhat below is what has been actually delivered at the Computex show under the Windows Vista :


Pic above is ie. how actually Intel's V8 monster was bussy during the benchmarks. 4 uncorelated dual cores. Too much dependancy of overall performance on useless core steroids, that's all. Intel X86-64 is hopelessly optimized towards two cores, and that is where the whole higher performance story ends.
More about the topics here , in advance. And I am afraid this pics below are the problem cause: Old, good 45nm hardware trap.



By the way, thingies like Coppermine bug leak are now sealed forever. Bug handling is now extremely clean and careful at Intel, thanking to engaged MOSAD guards. The moral is: no more leaks, no more ship wrecks. All associations to legendary Itanic ship are however, INTENTIONAL.

Tuesday, June 05, 2007

AMD demos Barcelona at Computex !



At Computex on Tuesday, the company plans to demonstrate three servers running Barcelona, a quad-core server processor. Three system builders-- Supermicro, Tyan and Uniwide -- will be running servers using Barcelona in a variety of configurations, said Steve Demski, Opteron product manager.
Barcelona is expected to help end the skid as AMD's first quad-core product. Servers with the chip are expected in the third quarter, while a desktop version will arrive later in the year.
In fact, although we have seen DVT level chips running up to 2.3GHz, we understand there might be another silicon spin (maybe more) of Barcelona that is due later this month. If this occurs, then current rumors swirling around the trade show floor about the mass availablity of product not occuring until the fourth quarter might just be true. However, Bad hardware found AMD Barcelona plans even on 02/01/2007. Tyan-AMD server board will be supporting the quad-core Barcelona with nForce Professional 4000 series chipset. And dual power lanes.


One of the Barcelona-based platforms is a blade design by Supermicro, a dense, 4P system that will features 160 processor cores within a 7U (12.25-inch) enclosure. This will translate into 960 processing cores in each rack, according to a statement from Supermicro, of San Jose, Calif.

Quad core is more power efficient.


Being Barcelona announcement quite certain, let the current prices go down.

This page is powered by Blogger. Isn't yours?