Thursday, August 30, 2007
AMD K10h in B0 revision since May 2007
With SS4A support. Before that, it passed A0-A2 revisions. B0 intentionally left some of its untamed bugs for B1 and subsequent B2 revision removal. B1 should tame the bugs and B2 revision tweak the parameters, at the first place clock, of course. One core revision requests some 4 months, thus on September 10th we will see announced B1 revision (that is currently, due to inherent instabilities locked to 2Ghz clock) .
When B2 for volumes? Probably only in 1Q 2008. AMD is obviously in panic delay 3 -6 months.
And marketing director is the only guilty? Interestingly to note, AMD added in its preliminary design L3 cache support only in December 2006. Obviously, competitive pressure from Intel has helped a bit. Then the final, thoroughly tested product is not reasonable to expect earlier than in December 2007. That is why there is no official K10 benchmarks yet. Phenom X4 should later run at 2.6-2.8GHz clock and at 125W TDP, until 45nm implementation arrives. Well, 30W per 2,8 Ghz core is not so bad.
Wednesday, August 29, 2007
Damaged hardware: Space Shuttle hole
Tuesday, August 28, 2007
First Overclocked Barcelona benchmarks
AMD's 3GHz K10 to break 30,000 3DMark06
Report seems plausible, but you will have to wait till 3Ghz Barcelona.
By the way, after the test someone has broken into tester car and had stolen all of its equipment.
Probably some nervous AMD fan. Or Intel's? . Who would know.
To me, the most intriguing part of the report is the use of 32 bit Win Me as the underlaying OS. You know WIn is much lighter OS, much less system interrupts and consequently much better benchmarks (that is why I still work on Win98 !).
3Dmark06 test requires XP. And minimum clock size of 2,5 Ghz.
Does it mean that before overclock single cored AgenaFX Barcelona has run at its real clock?.
When running at that 2.5GHz, the 3DMark06 score ended at 23.768. In that case 20% overclock seems quite normal under ordinary (non LN based) applied system cooling methods. However, seems that benchmark value, after overclock, jumped more than should be expected for the given overclock value. Instead just opposite, as might be expected under normal test conditions. That make me a bit suspitious on overall tester findings. But, it is up to you, to accept them or reject.
Now, lets compare Barcelona SDMark06 with Intel's Barcelona killers speeds :
In the end, lets see 3DMark06 on AMD X2 4400+ @ 2700mhz :
And A64 FX 55 @ 2600 Mhz:
And conclusion. The Inq staff add into the test system TWO high end GC. That pushed overall 3Dmark06 up to its limits. Processor itself can't be much better than it is at Intel. Laugh for Win Me ultimate gaming. Applause for >10X better 3Dmark06 result than on A64 FX-55. Impressive gaming performance advancement in only 3 years.
Think about your next cars 10 times faster. That might be only jet airplane. Congrats for NEW World record, stolen from Intel's LN cooled 5,1 Ghz overclock. Even less overclocked than Intel performance stars, K10h seems better. If proved so, that is my final conclusion. But, wait a bit to ensure youself. Oherwise, 2X improvement is hard or impossible to get.
Process variability below 65nm is more than 20%. And rising.
That is why AMD K10 is in delays. Everything is fine, but when you engrave 4 cores in one chip manufacturing variations will make their collaborative functiong pretty unpredictable and full of the glitches.
Only Intel has no problems even in 45nm. They invented perfect process, or they are perfect liars?
Monday, August 27, 2007
AMD Phenom logo, Barcelona will start two cored?
Where are the hell Barcelona benchmarks?
Please, enough with logo paperwork, we want real products or benchmarks. Now. Beside, seems that new Opteron might start with 2 cores. I don't see X4 mark at new Barcelona logo. Barcelona will start two cored? What a disaster. Seems that 65nm silicon can't bring out 4 cores, what else to say. That means in the worst case 4 cores only in 1Q 2008?.
Let me remind you, AMD released its dual-core Opteron server/workstation processors on 22 April 2005, and its dual-core desktop processors, the Athlon 64 X2 family, were released on 31 May 2005.
Was it normal 4 cores to expect in April 2007. Add evident K10 delay of 6 months. That means app. October 22nd 2007 target date for 4 cores Barcelona launch. And half of November for Phenom's. Yeah, I allow some K10 delays, because 4 cores are not in complexity terms simple linear scaling, like it was two core design (compared to single cored one). But, will competitors allow yet another possible delay? Thus, seems resonable to assume ONLY 2 cores Barcelona EARLY PRElaunch on September 11th. Though, I can't imagine hardware public angry voices in that case. However, one AMD's bad marketing excuse might be that September 11th is simply bad launch date target.
And who will be responsible for that latest debacle? Recently fired AMD's marketing director. In practice, no one. However, we want someone's blood. And tears. And nothing less than that.
We all know Phenom will start two cored, when its time come, but Barcelona too? OMG. Real disaster. Or, perhaps, you would more like wait 4X BArcelona on October 22 nd. And than perhaps again wait and wait . Who would like it?
Friday, August 24, 2007
AMD Barcelona logo announced
You should see it officially at Singapore event.
INQ has revealed the new logo of the upcoming quad core Opteron processors based on Barcelona core, scheduled for launch on September 10th this year. We heard that AMD partners has gotten B1-step Barcelona for compatibility testing with better and newer AGESA code compared to the B0-step Barcelona. There is also B2-step Barcelona which is bug free and then BA-step Barcelona for mass production. AMD plans for two 2-way Opteron at 1.9 and 2.0GHz at launch and higher clocked parts at 2.1 to 2.4GHz coming later this year. Thus, B2 -BA K10h revision by the end of this year?
B2 revision itself is glitches free, 6 months late and clock locked to 2Ghz? BA revision or step is for 2,6 Ghz Phenom?
AMD outlines timeline for new $3.2B fab after Phenom launch
AMD spokesman Travis Bullard said the company's plans for a new fab in New York "are on track."
"AMD has drafted an initial design review of the potential fab. We are pleased with the outcome of the process and feel that we have the right draft plans ready to go," he said.
Bullard said AMD now has two decisions to make: First, to determine the right timeline, then to initiate the project.
"AMD expects to make a decision on the timeline for this project by the end of the year," he said.
BADHARDWARE comment: After this order of moves seems that it is easier to make new big $3.2B fab then one new small but decent multicore processor design. Yes, I mean on Intel too. Even more than on AMD. So much on design complexities.
Thursday, August 23, 2007
Intel - AMD clock relativity theory
Intel's 3Ghz + 12MB cache server against 2 Ghz 2MB Barcelona. Intel's 3,4 Ghz + 6MB desktop against 3 Ghz Phenom. Plus 45nm inside . Seems like pretty expensive way to counter EFFICIENT K10h architecture. And if Nehalem doesn't hit its target in one shot, that will bring Intel into the troubles. Because 3Ghz server consumes 120W TDP, and 3,4Ghz desktop is already in clock diminishind return area like ill fated Prescott was.
L2 & L3 Cache
512 KB L2 cache x 4 cores
2 MB L3 cache shared
512 KB L2 cache x 2 cores
2 MB L3 cache shared
AMD Phenom rides again at 3Ghz clock !
And it is not hot. My arse is not so hot. Do you see 5 fans in the linked article picture? Obviously Phenom has 45nm manufacturing technology in mind and its time will be coming in 2H 2008. Then it will shine in its full glory.
But, about that time its competitor Nehalem might be ready? On the other side impressive is usage of its 4 cores, something that Intel can now only dream about. In the end, let me say nice processor whose time is coming. The only problem I see in this moment might be its yield in volume production, but in the next 12 months it should be solved in any case. however, in this moment AMD will only announce B1 core that is short in clock for some 600 Mhz. Simply, manufacturing volume yield at this moment simply doesn't allow higher level of B2 revision of K10h core production. That is why Intel decided to glue its established 2 cores and delivers them as 4 cores product. Ha#
Lets compare. Intel's mobile Penryn, 2x2 cores in 45nm will consume some 89W TDP, at 3Ghz in 2008. But, AMD Phenom shows that now, and will deliver it in volume earlier, by the end of this year. However, to be competitive in a lower geometry with competition means pretty good architecture. And K10h is obviously just that. Though, pretty complex too.
So, Phenom will deliver you in some not so distant future, without any doubt, some real performance for a decent price. Here is an ASUS mobo ready for Phenom:
Wednesday, August 22, 2007
AMD Ruiz believes the bad tide is turning
The company's next-generation server chip, Barcelona, will be released Sept. 10. A desktop PC version of the product, called Phenom, will arrive before the end of the year.
I admit, goal was pretty high and noble, but competitors were blood thirsty and note crucified Hesus, err. Ruiz in the mild shadow in the picture left.
And a new "platform" of computer and graphics chips for laptop computers, dubbed Puma, is anticipated for early next year. It will benefit from AMD's October purchase of graphics chip maker ATI Technologies.
The currents also might be changing on the antitrust front. Last month, the European Commission filed antitrust charges against Intel that parallel those AMD alleges against its rival in a civil suit in the United States. Intel says its conduct is lawful and pro-competitive.
Ruiz, in a visit with Mercury News reporters and editors, said he is confident the new products will help AMD regain lost ground. And he claimed the technical glitches that held up Barcelona are behind the company. Barcelona will be competitive with whatever Intel throws at it, he said.
Here are excerpts from the conversation:
QHow have Intel's allegedly illegal business practices affected your business?
AWe have great products. I think we demonstrated that with (the popular chip) Opteron. I think we're going to continue to demonstrate that with Barcelona. What's missing is our ability to be rewarded for having done those things right. That's the impact. We should have a significantly stronger position in the market.
QAMD is paying calls on antitrust regulators in the United States. Do you believe they should initiate an antitrust case against Intel?
AI think they should. The role of government is to protect citizens. There are laws being abused and broken. There is a monopoly that's abusing its power. I think it's the role of these regulatory agencies is to ensure that doesn't occur. From that point of view, they should take an interest.
QWhat would be an appropriate remedy in the Intel antitrust case in Europe and to your civil suit in the United States?
AThe thing we're after is the same thing our customers would like: a field in which they can make choices on products and technology strictly on the merits of the product and not because they were coerced in choosing one over the other. To me the remedy we would like to see is one where the practice of coercion and intimidation stop.
QWhat about a financial penalty?
AI think there should be, but that is not the No. 1 challenge for us.
QTwo years ago, Intel signed an order in Japan agreeing not to engage in anti-competitive practices alleged by the Japanese Fair Trade Commission. Has that helped?
AJapan found there were anti-competitive practices being exercised by Intel. In my opinion, they haven't stopped them totally. We know that there are things still being done that fall into the category of (what) shouldn't be done. When a customer tells you they have this plan with you and says, "I like it, I love it, we're going to do it," and 90 days later they change their minds because of an occurrence that happens to be along the lines of what (Intel) is not supposed to be doing . . .
(Editor's Note: Intel said it believes its business practices in Japan are fair and lawful, and that it is defending those practices in civil court.)
QWill AMD's new Barcelona chip change the competitive landscape when it is launched Sept. 10?
AI expect us to regain position in the server market because Barcelona is a very strong product. It won't have the huge impact Opteron had. If you remember at the time Intel had no new product (to compete against it). So it will be more challenging. What we're anticipating is that Intel and AMD are going to be very competitive in every space going forward. We've become big and strong enough and credible enough that we are going to flip back and forth in leadership.
QAt the same time, AMD has struggled with Barcelona. What has been the problem?
A: Frankly speaking, the issue with Barcelona has been that it is later than we wanted. We had hoped Barcelona would have been earlier in the year. We are suffering from that.
QWhat caused the Barcelona delay?
A: It's awfully complicated. We're doing something that nobody has ever done. As strong as our competitor is, they have not done a quad core on a chip (four chip cores, or brains, laid out on a single piece of silicon instead of on two pieces attached together). . . . Every time we ran into a gotcha (or technical glitch), it created a six-week or so hole in the schedule as we went back and fixed it. We hoped we wouldn't get many of those, but in the Barcelona case, we got more than we thought. By the time we got through fixing them all, we were six months-plus late from where we originally wanted to be. Picture below has been taken at Barcelona city in March 2007. As you could see, famous 4 cores Barcelona towers then weren't quite completed yet. Please note quite original direct connections between the towers. That would slash your visiting time in half compared with Intel's standard built twin towers. (Computer) architect behind this unusual design is famous Gaudi.
QYou joined AMD in January 2000 after 22 years with Motorola. Do you have any plans for retirement?
AI'm 61, and I haven't thought about retirement yet. I'm sure the time will come. At this point I feel young. I can still ride a bicycle 200 miles a week. I'm not an egomaniac. The time will come when we need to transfer power to a new generation of leaders. It's not my intent to be here forever. At this point, however, I am not planning any imminent retirement.
50 000 BADHARDWARE hits
See the hits counter on the right, below. 50000 fireworks video.
Toshiba announces 32GB SD flash
To be released on January 2008. However for October this year you could choose 16GB Class 4 one or 8GB micro SDHC. Finally, one backup drive for your laptop or palmtop, nail sized.
Some idiotic rants on AMD's SOI
ZD "expert" finds that AMD's SOI is broken. He even consulted another expert idiot to confirm his thoughts. Is it true?
However, it is correct that AMD 65nm is late, it is correct that it is more complex process than Intel's simple bulk process and thus heavier to volumes tune, it is correct that AMD's cache is worse than Intel's.
But, it is not correct AMD's process is broken. The proof? AMD will switch to 45nm in H2 2008.
Probably nothing new, only shrink of an old 65nm. But, how will AMD successfully shrink something intristically broken, if ZD's idiot is correct?. No way. So, his lies are not so naive, and their purpose in a week when
AMD benchmarks will be announced, are not in any case without one back intention.
Let me conclude: AMD's complex process was simply late for volume production. But, not once again.
Ever. Because, its metal gates for 32nm should be simpler process.
Now, it is Intel's turn to make bad move with its 45nm metal gates. That could be literally broken, because
as you know, joint process is easily physically breakable.
Sorry, but AMD haters should accept the fact that AMD now has product delays behind CEO meetings topics.
Oh, by the way, Barcelona launch is at Barcelona and NY. Though, I don't think September 11th launch date in NY is quite happy choice.
Monday, August 20, 2007
Why AMD K10h will beat Intel in large number of applications?
Barcelona K10 CPU can compete successfully with Intel processors working at the same clock speed in a large number of applications and win the competition.
Sunday, August 19, 2007
AMD prepares Barcelona benchmarks at Singapore event on August 25th
AMD's claim: Benchmarks are not available for dissemination yet. She is still trying to obtain confirmation from global HQ for their release date. Once she gets the information she will release the information to you prior to the date. Sorry about this but this is the best we can do at the moment.
Saturday, August 18, 2007
Intel's first 45nm processor at max 3,16 Ghz
Intel recently increased the number of 45nm-based CPUs it plans to launch for desktop PCs to nine, according to sources at motherboard makers.
Within the nine CPUs, five will be dual-core processors (Wolfdale) and the remaining four will be quad-core (Yorkfield). All will adopt a 1333MHz FSB. Model numbers for the upcoming products are still undecided, noted the sources.
Four of the five Wolfdale processors will have core frequencies of 3.16GHz, 3.0GHz, 2.83GHz and 2.66GHz, while the frequency of the remaining one is unknown. All five Wolfdale processors will feature 6MB L2 cache, detailed the sources.
Three of the four Yorkfield processors will have core frequencies of 2.83GHz, 2.66GHz and 2.5GHz with the remainder as yet unknown. Three will include 12MB L2 cache while the 2.5GHz version will have 6MB.
Intel will introduce the ability to increase the clock multiplier in steps of 0.5 in the upcoming 45nm range to increase its model choices, pointed out the sources.
With Intel aiming to eventually scale the 45nm range up to a maximum core frequency of 4.0GHz, the sources estimate that at least four more CPUs will appear at a later time with frequencies higher than the initial 3.16GHz.
Intel declined the opportunity to respond to this report, saying it cannot comment on unannounced products.Ha, I remember this kind claim on 10 Ghz Tejas Pentium. Can we call its 4Ghz 45nm version Unobtanium? For example, on too unannounced Merom you know everything , a lot of months ahead.
Friday, August 17, 2007
Theory of financial relativity: When 2% drop is bigger than 2,43%
About credit caused several of our fave hi tech shares to bleed on US stock exchanges.
Google (GOOG) fell by $11.62 (2.28%), Intel (INTC) by 46 cents (2%) and AMD (AMD) was down 29 cents (2.43%).
Actually, Intel's fall is smaller in percents , but their shares have bigger nominal values, so today Intel's shareholders lost much more than AMD's. And prevailing public opinion is that currently Intel is shaped good, and AMD bad. Bullshit.
Yes, but if you don't look at your pocket.
Today AMD dropped yet another price slash (Opteron down for 30%).
Wednesday, August 15, 2007
Intel: Barcelona "may challenge us on specific server/workstation benchmarks."
Though, Intel believes that its upcoming 45 nm Core 2 Duo product line will keep a "broad leadership" position in the industry.
I second that. But , until then "poor" 1,9 Ghz Barcelona may challenge Intel on specific server/workstation benchmarks?
So, Intel inevitably needs 45nm to counter attack? One 45nm fab is worth at least $5B. Good God. I found Barcelona new advanced architecture much cheaper way for obtaining top price performance ratio.
Yes, Intel aims to regain "clear performance and energy-efficiency leadership" with Nehalem in 2008.
But, what to do in the meantime? AMD will shrink its genuine 4 core die in 45nm in the second half of 2008.
And its prices. Were are Intel's margins (declining by years) deseprately needed for new expensier and expensier
lower geometries fabs. Seems that something stinks in Intel's economy mathematics. Margins are lower and lower, expenses higher and higher. In one moment investors should ask themself, is Intel a barrel without the bottom?. Where are our returns? How certain are they? And if Intel fail, who will buy so big company?
And what for? Think with your own head, not with someone's foolish ass.
Monday, August 13, 2007
AMD Phenom FX 4 cores strikes at 3.0 Ghz
The secret of Intel's success !
Bigger production? Well say much less than 50%.
New architecture. No, 0%.
Barnum style marketing? Markting is very important to non educated people.
Introduction of new production processes in time? Yes, definitely yes. But, how?
What is Intel's secret weapon?
Intel has Oregon Development facility named D1D. What does it produce? Literally nothing.
It makes Intel's each new manufacturing process better, so the rest of production fabs make shorten introduction of the new processes. On the other side, that is AMD's weak point.
It needed the whole year to bring production yields to acceptable levels. That is
why AMD drove so long it established 90nm process. How AMD can cope with?
There is only one long time solution to counter with. Its name is LG.
Without fast volumes ramp, AMD will remain only successful engineering company.
Todays 3GHz Phenom Japan presentation gives Vista windows experience index 5,9 .
By definition here that means literally:
Level 5 PCs will dramatically exceed the Windows Premium logo requirements. They will easily run Aero-Glass on multi-monitor systems at high resolution. First person shooters, multiple HD streams, video creation, high end multimedia applications are all characteristic scenarios of people who will be using Level 5 systems. As Level 6 machines are not defined yet, it is possible that the highest performance and capable Level 5 machines may be re-leveled as level 6 in the future (i.e. that may happen if the new hardware is evaluated before the index is updated with the new capabilities).
However, just as success of one processor at market is not solely dictated by its performance, the Vista windows experience index isn't dependent on processor's but on overall system performance. And it is
dependent on its weakest component, whatever it is. However, usually processor is that weakest chain link.
Thursday, August 09, 2007
Barcelona definitely on September 11th.
Hmm, seems quite unpleasant date timing for Intel's CORE DUO monopoly towers.
If so, expect AMD Barcelona's first complete benchmarks in a few days.
It is literally the last moment for Barcelona launch. First K10 revision B0 was taped out in early December 2006. And 9 months is even enough for a baby to be born.
Let me remind you that Barcelona revision b1 was taped out in March and
The Inquirer christened it K10. Actually resurrected it after previously being "officially" announced dead by another The Inquirer luminary journo. And finally, about Computex time, revision B2 was taped out, however not publicly exposed.
Barcelona, save us, please.
Tuesday, August 07, 2007
How hot is Barcelona ?
Barcelona's memory consumes 1/3 of Intel's 100W power
To keep pace in the performance race, seems that Intel gives significant contribution to global warming.
Seems that Barcelona accelerates memory access by intelligent design, not by brute power consumption force.
Saturday, August 04, 2007
DAAMIT: Phenom X4 starts at 3Ghz by Q4 2007
At 3.0GHz, Phenom X4 can defeat Intel Core 2 Quad and Core 2 Extreme so at 3.24GHz Phenom X4 also can be better than 3.33GHz 45nm Yorkfield Core 2 Extreme QX6950. In the Q1 2008 it will exceed this clock. I am curious how far Intel could exceed 3,3 Ghz clock. Other sources second those rumors. Let me remind you, at 3,2 Ghz clock Pentium started with diminishing returns. Does it mean Intel's new deja vue false performance campaign? AMD will continue with reduction in power consumprion after introduction of its 45 nm in H2 2008. WHat Intel can do more in 2008? Nothing.
Though, Intel (s)thinks that it can easily manage 4Ghz clock in 45nm, and 3,66 Ghz in quad core.
Good morning mister Prescott.
Friday, August 03, 2007
Sun to Release 8-Core Niagara 2 Processor
"Sun Microsystems is set to announce its eight-core Niagara 2 processor next week.
Throughput Computing is the underlying strategy of Sun's new family of UltraSparc processors designed to significantly increase real-world application performance while helping to cut the cost and complexity of network computing. These processors maximize throughput [via] chip multithreading. CMT integrates the power of symmetric multiprocessing onto a single chip, allowing a single processor to execute several software threads simultaneously.
43% of Intel’s profits a result of monopoly
$80 biillion figure "translates into an Intel monopoly tax on every consumer who purchases a computer. That's a jaw-dropping figure that helps explain why the European Commission brought antitrust charges against Intel."
|Annual net profit comparison (in million dollars, rounded)|
Wednesday, August 01, 2007
AMD 2Ghz Barcelona benchmark is here !
It significantly beats comparable Xeon 5345 in power performance ratio. Now, we may expect Barcelona
official launch in last August week. Compare yourself AMD's and Intel's performance claims.