Monday, December 31, 2007

Happy New Year !



To the three leading independent hardware reporting sites. Names are odious, but they will recognize them self. I mean their objective style on hardware reporting.


What will happen in 2008, from 1908 perspective?

“When the expectations of wireless experts are realized, everyone will have his own pocket telephone and may be called wherever he happens to be,” one magazine predicted in 1908.

For your info in the year 2007 there were some 3B mobile devices worldwide. And some 6B of people. Pretty close to definition of everyone.

The New Year on alien planet !

Unthinkable has happend. With 2 feets only telescope.
10 times bigger telescope should see Earth size planets around another stars?
Well, provided that those planets are frequent in the Cosmos.

Thursday, December 27, 2007

Intel's Poulson delayed indefinitely !

Delayed? yawn. Now we have breakthrough Tukwilla, somewhat later Nehalem and brilliant Xeon.
3 brilliant processor in the same year? Do you fill that onla one really good was quite enough to kill AMD? Once forever. Or something there just isn't so brilliant. Like Poulson architecture.

Intel delays again !

Intel-ST memory venture delayed

OMG Intel's TLB bugged too ! ! !


However, "independent" journalists somehow forgot to stress it. And Intel's VM inexplicably stalls. :)
Anyway please a very frequent use of No Fix item in a breakthrough product.

And finally, there are errors in the Char also TLB-Bug (AX47), which, like AMD's TLB-Bug, though under rare scenarios "should occur, and for which there is a workaround by BIOS.

We should prepare to live with TLB bugs "under rare conditions" forever. Right?
Intel's corporate strategy is to invest in journalists not in a boring B3 like bugs taming.
And why should do it? There is a BIOS workaround. To work around 6MB L3.
And that moves power consumption significantly down. Right? :)

AMD goes triple core first. In March 2008.

Seems quite natural, after the yield problems in making quad cores in a pretty big chip area consuming 65nm technology.

AMD Phenomen(al) troubles


The Phenom Plug-In Upgrade Doesn't Work

Tuesday, December 25, 2007

The Queen Youtubed !


Two years ago, she famously confessed to never having even used a computer, but now Britain's Queen Elizabeth II is keen to show the world that she is in step with the times. To prove it, she's become the first monarch to have her own YouTube channel.

By the way, we should remember that The Queen coronation in 1953 was the first direct TV broadcast event in UK. Actually the first such event at all, because at that time VCR of any kind simply didn't exist. Seems that nasty computers were just a dark learning episode in the Queen Earthly life. God save the Queen !

Hehe, AMD K10 become glorious, literally !



Finally, a truth on bad hardware
With some Phenom 9900 eng. sample benchies. However, one question is urgent: Is AMD 45nm delayed too?


Monday, December 24, 2007

Marry Biometric Christmas Big Brother !

FBI Prepares Vast Database Of Biometrics
$1 Billion Project to Include Images of Irises and Faces
There are 900,000 federal, state and local law enforcement officers who can query the fingerprint database today, they said.
Civil liberties? Are you kidding? Face your future slaves. Down.


Now AMD delays because of Intel's delays?

And Intel has delayed because AMD had previously delayed?
Ha ha ha.
They are competitors in delaying or all of this is ridiculous?

We are eagerly waiting for the next Intel's move. Let me guess. Might it be yet another delay?

On Energy Dissipation In General Purpose Microprocessors


Abstract—In this paper we investigate possible ways to improve the energy ef ciency of a general purpose microprocessor. We show that the energy of a processor depends on its performance, so we chose the energy-delay product to compare different processors. To improve the energy-delay product we explore methods of reducing energy consumption that do not lead to performance loss (i.e., wasted energy), and explore methods to reduce delay by exploiting instruction level parallelism. We found that careful design reduced the energy dissipation by almost 25%. Pipelining can give approximately a 2 improvement in energy-delay product. Superscalar issue, however, does not improve the energy-delay product any further since the overhead required offsets the gains in performance. Further improvements will be hard to come by since a large fraction of the energy (50–80%) is dissipated in the clock network and the on-chip memories. Thus, the efficiency of processors will depend more on the technology being used and the algorithm chosen by the programmer than the micro-architecture.
From:
Energy Dissipation In General Purpose Microprocessors
IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 31, NO. 9, SEPTEMBER 1996

Let me conclude: How Intel will save energy at Yorkfield's 12MB of L3? No way, so they decided to copycat AMD as soon as with Nehalem, unlicenced copy of Barcelona.
Beside, I had no idea that energy efficiency of X86 say depends on software and benchmarks used. Is that why we have recently "only selected" benchmarks allowed?

Sunday, December 23, 2007

Intel unable to compete AMD with 45nm quad core prices ! ! !


I knew this might happen. 12 MB L3 and 150W TDP (design collateral damages) for only $ 1,5K. The sale will be brisked, so Intel will not be able to satisfy demand. :)
Beside, Intel is crusified at 3.2 GHz clock after its TDP of 150W. In their unbeatable 1000 times lower power consumption high-k 45nm process. So, 400 Mhz of clock rise costed Yorkfield some hefty $1000. Nice future outlook. Seems nothing better before Nehalem. Beside, why should Intel deliver in the end copycat like Nehalem, if Yorkfiekd is a good and perspective processor? Think about it.

Saturday, December 22, 2007

Intel in serious troubles with the Yorkfield FSB !!!!

It simply doesn't work with X38 chipset. And there is bunch of those unsold at the mobo manufacturers. Now, they have left a choice to:
1. Throw X38 based mobos in can (and all their investment too)
2. Or, Intel to postpone Yorkfield 2 months until those incompatible, ill fated X38 mobos are sold.

The second option is just what is happening. Until then, AMD will ready Phenoms at 2,6 GHz and 2,4Ghz. Hard match is ahead.
Yorkfield at 2,83 Ghz against the AMD at 2,6Ghz.

And The Inquirer found no competition at Intel?
In this race Yorkfield might be slightly better in performance, but AMD's 65nm (and yield) is quite mature and prices will go down and down. However, the same design choice (huge caches) that drives Intel performance up, will drive its factory prices up too. In that situation will be hard for Intel to compete AMD. Not to mention market niches where Intel has no the answer like 3 core one.


Microsoft pushes customers to pay for unusable DRAM !

Until I had do downgrade some 40 Vista laptops in my company I didn't believe it is possible.
The reason? They were so slow and so unreliable, the downgrade was the only choice. HP representative informed us that 1GB of DRAM those lapotops had installed, is not recommended Vista's MINIMUM memory size.
Their proposed solution? 4GB of DRAM upgrade and Vista SP1. Then, many things will go faster and better.
Nice, but seems STILL twice slower than the same hardwer under XP+SP3. Right?.
Beside, 32bit Vista is unable to recognize more than 3GB of memory !
How to use all my installed, but partially unusable DRAM? To install 64bit Vista, significantly slower?.
And all the necessary patches. Then you will finally get great, slow, DRAM upgradable Vista system.
But hmmm, do you need it at all?

X86 computer on a chip


1,5 Ghz and 1GB of DRAM. Who needs more? Intel will add in a 2 -3 years coin sized 8 GB SSD to finally get a fully functional palm top size general purpose X86 computer .

Friday, December 21, 2007

83 HP notebooks killed

Atctive X flaw, as you perhaps know.

However, some are already patched.

Thursday, December 20, 2007

AMD survived -15% operating margin in last 4 quarters !


Finally, a truth on bad hardware

Vaporware of year, 9 years ago: Windows NT Wins

Vaporware 1998: Windows NT Wins

All the similarities with Vista are of course intentional.

XP with SP3 is more than twice as fast as the Vista SP1 !!

According to a Florida performance testing software developer, XP SP3 is not only 10% faster than XP SP2, but more than twice as fast as Vista SP1, claims that Microsoft disparaged within days.

SP3 update is the size of only 70MB, but not before H1 2008.
Stright back into the future.

School claims it was largest human-power computation in history


School claims it was largest human-power computation in history.
I suggest further power efficiency improvement. Or entirely new class of computing.
To use horses instead of people for power generation. Human work for that kind of work seems to me a bit Orwellian.

Wednesday, December 19, 2007

And the winner for the Lie of 2007 is The Inquirer


Once distinguished computer site, is the winner of the Great 2007 Lie Award.
Here is their Award winning headline:

Intel to delay 45nm quads - report, No competition
Intel has already notified its partners that it will push back the launch of the three CPUs to February or March next year. The three CPUs that Intel plans to delay are the Core 2 Quad Q9300, Q9450 and Q9550, added the sources.

BADHARDWARE's Award Committee speech:
1. The Inquirer found bad news at badhardware long time ago.
2. Intel is in yield troubles with 45nm and high-k high end processors as badhardware reported
3. AMD touched the bottom, and Intel didn't kill it. Revenge will follow.
4. The Inquirer reported in 2003 that Intel DELAYS further Pentium speed improvements
The last holdup that Intel faced for a major desktop processor launch came during an intense clock-speed battle with chip rival AMD, when Intel announced and then recalled its 1.13GHz Pentium III.

The Inq reason? The same one like now, no competition. Hahaha. What later happened with Pentium, and its competitiveness you know. Later became known that Intel secretly and in a hurry has developed new Core Duo architecture at that moment.
Almost the same situation is now. The only difference is that Nehalem being so late is no more so secret project.
P.S. What about lies of Mr. Bush? Well, he is out of competition,
simply because he has no competition.

Future Disneyland or Disneyland of the future ?


50 years after Disney envisions , we got in reality only melting glaciers . Seems he didn't make prior energy bill for his vision of the future. Ironically, US highways government strategy the first envisioned father of Nobel prize winner Al Gore, who the first envisioned what his father actually did.
What I envision more clever is that in the future information should run and people stay at home. Alas, my son soon realized that Internet servers already started to melt glaciers too (thanks to Xeon inside).
Free lunch is still not invented, but anyway that should be without any doubt the biggest invention of all times.
What we actually need are info highways based on reversible computers. And when those will start to crunch?
Hmmm, eh, weeeell, .... say in 50 years.

Tuesday, December 18, 2007

Americans dependant to deficit


Projections show that by 2070, total Government expenditures are projected to be 50 percent of GDP. Such levels of expenditures have only been witnessed once before,during World War II .
By the way, The Cost Of Bush: $32 Trillion translates into a current burden of about $175,000 per American

The Brits addicted to computer !

Brits spend 24 years of their life in front of the computer.

Third of their average life, even without irresistible Vista gadgets still applied on their displays???
Sign of the Beast?

Intel fails with 32nm generation ! !

As a result, NIST predicts that current photoresist formulations will not be able to achieve the necessary sharpness for chips at the 32-nm node.
Seems that world would be locked a lot of time in 45nm. TIC, TOC, TIC, TOC ?. My ass tic toc soon.

The Free Lunch Is Over: 4 x 3GHz < < 12 GHz

A Fundamental Turn Toward Concurrency in Software

This article appeared in Dr. Dobb's Journal, 30(3), March 2005.
What has changed in the past three years? Seems nothing.
X86 platforms are simply late. AMD as a fore runner in the field is late some 6 months.
Intel is late a year behind AMD. Microsoft is late with Vista. ANd no one even tried to optimize and integrate multi core hardware and software as a mainstream event, taking in account compatibility problems.
Thus, what will really change until 2012 ? Seems nothing.
A few rare classes of applications are naturally parallelizable, but most aren’t. Even when you know exactly where you’re CPU-bound, you may well find it difficult to figure out how to parallelize those operations; all the most reason to start thinking about it now

Monday, December 17, 2007

Sun's Rock in H1 2009

Sun hoped to ship the 16-core SPARC dynamo by the end of next year. Now, however, we're hearing that early versions of Rock have struggled to perform during testing. So now, Sun expects to ship the chip in the first half of 2009.

Bad Hardware 2007 choice is tadaaam Vista !

Five years in the making and this is the best Microsoft could do?

It's not that Vista is awful. The integrated security and parental controls are nice, and the Aero interface is as whizzy as it gets. Searching and wireless networking are ...

AMD offers quad cores for dual core price ! ! ! !

Round a price of AMD coming in, especially the price of the Phenom.

That’s right! With effect from Monday 17th December you will be able to harness Quad-Core Power at a traditionally Dual-Core price!"

Sunday, December 16, 2007

Intel's 45nm suckers

3DMark 06 benchmark shows Intel's 45nm generation is on average 2% faster at the same clock, compared to its 65nm generation. The only significant improvement is lower power consumption.
Some 15% max. That will enable Intel to push its clock some 7-10% up from the current generation.
That means in the whole next year Intel will move its performance max 7-10%. Or its clock from current 3 GHz to 3,3Ghz. AT the same time, AMD will move up from current 2,4 Ghz for the same value 10% plus IPC improvements enabled by B3 revision. And with additional IPC in 45nm Shanghai. Year 2008 is obviously AMD's. Regardless of all stories.

Intel 45nm DT CPU Pricing ( 1Ku ) Q1, 2008

Model

Speed

L2 Cache

FSB

11-Nov

Jan 2008

Quad Core

QX9650

3GHz

12MB

1333MHz

$999

-

Q9550

2.83GHz

12MB

1333MHz

-

$530

Q9450

2.66GHz

12MB

1333MHz

-

$316

Q9300

2.50GHz

6MB

1333MHz

-

$266

Dual Core

E8500

3.16GHz

6MB

1333MHz

-

$266

E8400

3.00GHz

6MB

1333MHz


$183

E8300

2.83GHz

6MB

1333MHz

-

-

E8200

2.66GHz

6MB

1333MHz


$163


Saturday, December 15, 2007

Exclusive: Benchmarks of 2,6 Ghz AMD Phenom 9900


See the benchies themself here. BEING still revision B2 Phenom 9900 is performance comparable to Intel QX6800 at 2,66 Ghz. In B3 revision, Phenom 9950 should be comparable to QX at 3,2 Ghz and at the approximately same TDP.
Bad hardware thing at Phenom 9900? At 1,3+V applied, TDP is 140W .
Thus, no chance for anything faster, before, you guess Phenom serie 10 000 in 45nm, the next year.
Bottom message: Never try to use at Intel processors program chunks size of 256KB (or larger). You will get a sudden drop of performance of 100%. That is why new Intel multicore is performed miserable in supercomputing applications.

Is Microsoft Hyper-V a VMware Killer?

Microsoft Hyper-V could receive a warm welcome, especially from small and midsize businesses, according to Gordon Haff, an analyst at Illuminata. "In situations where companies have invested a lot in Microsoft software training with CRM, Sharepoint, and other applications, the low energy path is to go with Microsoft," he said.

Let me remind you. All you can squeaze out from future AMD Shanghai and Intel Nehalem multi core processor is based on virtual computing. One server board might contain 4 or 8 independent OS instances, you previosly needed 8 separate cases. That could save a lot of polar bear lifes. Right?
I mean: Asian datacenter energy use to double by 2010
The region excluding Japan will require electricity equal to output from two new 1,000-megawatt power plants by 2010 to run datacenters.

With a new 45nm high-k, Intel introduces new problems too!

And contrary to AMD hides them.
Mizry mentioned new high-k defect types that must be reduced or eliminated in the name of transistor reliability, including those elusive oxygen vacancies, and in true Intel fashion he announced their suppression.

Bigger chip area, more of those vacancies and ... complex Yorkfield is delayed !!!. That is what actually happens.

Delete "boot.ini"

It might appear from the installer log that we made the mistake of explicitly deleting \boot.ini instead of just boot.ini. The former would delete the file from the root of the current drive whereas the latter would only delete from the current working directory. It is not quite so; the mistake was to assume that the file would be deleted from the current working directory without giving the full path. In fact, this is the code in the installer script that caused the problems:

Section "EVE-ONLINE"
SetOutPath "$INSTDIR"
Delete "boot.ini"
Delete "manifest.dat"

File "boot.ini"
File "manifest.dat"
File "resDX9*.stuff"
SectionEnd

Guess is this a part of Vista application software ?
The documentation for the Delete function says the file should be specified with a full path but in fact it must be specified with a full path, like so:
Delete "$INSTDIR\boot.ini"
Delete "$INSTDIR\manifest.dat"

Friday, December 14, 2007

AMD's latest roadmap


Finally, a truth on bad hardware. Shanghai is improved core K10.5 with better IPC, larger L3, made in 45nm . Not bad. More promising seems a hope that big swallow named Barcelona is finally digested.


Revolutionary Sound DRAM

A new way to store information inside a fibre optic cable by turning it into a high-frequency sound wave has been demonstrated by researchers in the US.
By the way,
Does anybody remembers mercury sound delay memory lines? Seems we are back in the future.


AMD Humbles Itself Before Wall Street, Adds Roadmap Details


Though market share remains the same.
AMD also said that it will launch "Montreal," a quad-core server processor, in 2009.
This announcement makes K10.5 core more likely in 2009 than in 2008.

AMD will be profitable by Q3 next year.

Thursday, December 13, 2007

AMD shares deep down

The Inq reports.

Researchers claim computer records for speed, scale, detail and parallelism

On Opteron 2,6 Ghz. Not on Intel's quad core beasts. As BAD HARDWARE predicted.

Sun Niagara 2 open sourced !!

Sun Niagara 2 CPU Now Open Source


With all its 8 cores.

AMD is looking for a new President ?

Dirk Meyer, former Intel's emplyee?
Dirk wouldn't pay 3 times more for ATI than it is worth?
Perhaps some other hard questions?

AMD is looking for FEOL 45nm high-k engineer

PRIMARY PURPOSE: This position is for a technical individual contributor within the ASTA partnership at IBM's facilities in East Fishkill, NY. The individual selected for this position will work together with AMD and ASTA engineers in IBM's 300mm development facility. Employee will participate in joint development of advanced FEOL etch processes and modules. Projects will include development and optimization of etch processes for STI, Gate (incl. high k and/or metal gate), Si recess, spacer dielectrics, and other FEOL etch modules in 45nm and 32nm technologies, linking with other process engineering and integration groups. In addition to development experience, experience optimizing etch processes for manufacturability will be important.

Wednesday, December 12, 2007

AMD to announce high-k 45nm K10.5 core in H2 2008 ?

Too ambitious, though the best thing I have ever heard, but if make it, AMD might bury Intel.
With a grain of salt we should take this and IBM's decision on High-k only at 32nm node in 2009.
On the other side, yet another delay sequel might definitely bury AMD itself.
I think less radical approach will evolve. With K10.5 somewhere in 2009.

Intel's X38 and X48 chipsets comparison


X48 seems pretty pale, without DDR 1333 support.

Microsoft Releases Office 2007 SP1, XP SP3 Release Candidate to the Public

Microsoft has two presents for its customers today with the release of Service Pack 1 (SP1) for Office 2007 and Service Pack 3 (SP3) Release Candidate 1 (RC1) for Windows XP.

According to the Office 2007 SP1 whitepaper [DOC], Microsoft took steps to enhance its productivity suite in the areas of stability, performance and security.


Iranian are even worse than we suspected ?

Iranian Scientists Develop Country's Most Powerful Supercomputer

And Intelligence Services are better than we expected. They even sometimes tell as the truth. Click on the pic to enlarge it. Iranians smuggled AMD processors for supercomputers? They hadn't choice. Intel's Yorkfield is delayed, and beside made in Israel. ANd for their computer they had use the best procesors. And everyone now knows who is the best. Beside, it is pretty hard to stop someone to buy 216 processors at 216 different locations. Right?

The Government is however careful more than expected, censoring Iraq invasion Wiki article. Why?
Because Iranians are quite close to Iraq. And Government can't allow unsecured web talks.
So, the GOvernment revision was made in 2005 but has only come to light now, that software can tell who has been attempting to mess with your mind on Wackypedia.

Tuesday, December 11, 2007

BAD King Pentium is dead, long live the King


At 2,2 Ghz. Man, I thought for a moment this must be the year 2200, error 2002 , not 2007/8 !
Let me be stright. Pentium at 2,2 Ghz was a good, efficient processor. The problem is in the year. 6 years passed and Intel has nothing better to offer ? Soon we will see dual and triple core Phenoms at 2.2 Ghz and at very low prices. Intel actually does now what AMD will do with Phenom in 45nm. But without TLB bug bonus. By the way, I write this using the almost 5 years old Celeron at 2,4 Ghz. Gosh. But, this new product must be for the customers born in 1995. They don't know too much. Right?

Naked pictures of all the Intel's Royal family members are exposed here.

AMD taped out 45nm parts !

The 45nm, we consider it Rev C of the device. So all the learning, all the hard knocks that we had on Barcelona, we're going to apply it to Shanghai. I just read on Friday, I have to admit to my delight because misery loves company, that our competitor's 45nm part seems to be delayed too. So, errata. And it happens, because these are advanced products, right? So, test vehicles so far on 45nm look good. That's why my confidence level of being able to say generally that we'll have silicon on 45nm is pretty high.

BADHARDWARE: Good. January samples are mature alpha revision. Soon after B0 will follow, and after B3 experience in 65nm, 45 nm B1. However, I am not sure that AMD will ever again try to announce B1 revision, like it did with Barcelona. Let me conclude, 45nm B2 is in H2 2008. Just as AMD promised.
Probably in September 2008. And it should be Barcelona 65nm shrink. No new experiments, no new complexities. Perhaps only new, lower prices. Because 45nm processor area is almost half of the 65nm one.

Singapore will use Gigabit Internet in 5 years

Holy...
I am now satisfied with 100 Kbps. Think in terms about 10 000 times lower value. What would you do with 10 bps ? Morse telegraphy. That way would look like my current connection in 5 years.
From the Singapore perspective.

Intel delays 3.2 Ghz Yorkfild !


And will keep it forever at 3.2 Ghz , after its melting 150W TDP. Obviously, 2008 is not Intel's year.
And will not offer anything else new until Nehalem. Until then Intel should have very bad moments in price/performance war against the AMD.

Monday, December 10, 2007

Microsoft manager arrested

Former Microsoft manager facing trial for billing fraud
Former Microsoft Program Manager Carolyn Gudmundson has been charged with 11 counts of wire fraud and seven counts of mail fraud related to a complex reimbursement scheme she had set up, resulting in her making almost $1 million in fraudulent funds.

During her time as Program Manager for Microsoft's MSN Division, Gudmundson was responsible for the registration and maintenance of domain names for both Microsoft and travel site Expedia. According to the US Attorney's Office for the Western District of Washington, Gudmundson had a three-part scheme in order to defraud Microsoft. Since she was authorized to use her personal credit card to pay for Microsoft's domain names and submit requests for reimbursement from the company, she altered receipts to show higher prices than what she had actually paid and included those in her expense reports. Gudmundson also submitted invoices to Expedia for domains that had already been paid for through an agreement between Microsoft and Expedia.

BADHARDWARE: What has been doing Microsoft Money finance tracking program all the time? Not good enough?

IBM and AMD will switch to high-K at 32nm node

IBM's laboratory tests in East Fishkill, NY, have confirmed a 45% reduction in total power consumption using their new process, with lower operational voltages. They have indicated this equates to a 30% speedup potential for existing chip applications.

Moving disk 50 years ago


IBM RAMAC with whopping capacity of 4,4 MBytes.
MB not GBytes, it was the year 1956.

Saturday, December 08, 2007

Vista SP1 brings ExFAT with 16 Exabytes of address space


Some billion times bigger address space than with FAT32.

Drug dealer sells US Labs nuclear secrets on USB stick


2006 a drug dealer was allegedly found with a USB stick containing data on nuclear weapons tests.

Los Alamos has a checkered security history, having suffered a sequence of embarrassing breaches in recent years. In August of this year, it was revealed that the lab had released sensitive nuclear research data by email, while in 2006 a drug dealer was allegedly found with a USB stick containing data on nuclear weapons tests.

"This appears to be a new low, even drug dealers can get classified information out of Los Alamos,"

This aspect of ultimate NAtional security should be stressed by President Bush in his state of the Onion address.


Friday, December 07, 2007

Holy s..., Intel's buggy Yorkfield delayed to February


Apparently the "integrity" of the FSB frequency is not satisfactory and may even cause crashes.
However, Intel's shares, 52 week highest still doesn't crash.

Seems that both price war competitors are completely exhausted in all aspects:
profits, performance, stability, prices, authenticity. That was actually BADHARDWARE
goal. Now, I can reveal why. Because I can't look at healthy, rich hardware billionairs.
I actually want their blood and tears, their suffers, just as we ordinary users suffer using
their products. Now, they will exactly know what their users want and needs. And should
be more sensitive in the future.

However, if you want to see what the bug is, wait unti Nehalem arrive.

Thursday, December 06, 2007

AMD TDP is out of control, but there is a solution !


Lets call 79% of TDP ACP. And it nicely fit in current thermal envelopes.
This explains why AMD is going to require a horrible 140W TDP for the 2.6GHz Phenoms, and suggests continuing severe 65nm process problems for AMD. Their parts are literally burning up as they approach 2.6GHz.

Details on Phenom TLB patched performance !


No TLB patch TLB patch
Sandra cache and memory bandwidth 6527 5932 9.6%
Sandra memory bandwidth - FPU 5403 3650 38.7%
Sandra memory bandwidth - ALU 5401 3648 38.7%
CPU-Z memory access latency 63 99 44.4%
WorldBench - Microsoft Office 2003 SP-1 371 399 7.3%
WorldBench - Adobe Photoshop CS2 521 595 13.3%
WorldBench - Firefox 298 536 57.1%
WorldBench - Microsoft Windows Media Encoder 9.0 248 272 9.2%
WorldBench - WinZip 10 305 321 5.1%
picCOLOR overall score 9.74 7.21 29.9%
Valve Source engine particle simulation benchmark 62 55 12.0%
Valve VRAD map build time 182 191 4.8%
SiSoft Sandra Multimedia Integer x16 130697 130648 0.04%
SiSoft Sandra Multimedia Floating Point x8 169434 169373 0.04%
Total average difference 19.3%
Average difference without memory subsystem tests 13.9%

Phenom B2 Firefox is phenomenally 47% slower?

As we've reported elsewhere, AMD does plan to fix the TLB erratum with a new revision of its quad-core chip due some time in mid-to-late Q1 of 2008. Once the new revision is available, the Phenom 9500 and 9600 will be replaced by the 9550 and 9650, with the -50 suffix denoting the updated silicon and higher performance. Most users will want to wait until those new Phenom models are available before paying full price for a Phenom processor or a system based on one.
However, Barcelona should be in a fixed B3 revision in January.

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