Only thing worse than a chip analyst is a chip journalist

Senior Analyst? What kind of computer eng. is that, give us a break please. To most of us a senior analyst translates to educated idiot. Most analysts are self serving morons surpassed only by journalists who print their drivel as fact, they are not technically based.

To my knowledge Risc is not a favorite of the wintel school of computer design, yet the article claims that wintels were there first, BULLl. Intel and AMD are going to have to do something as their current desk top cpu's are running out of legs. The bottom line is some one has a better mouse trap and the analyst and journalists who favor M$ and Intel are the real liars. All of the Wintel 64 bit chips listed are work stations and not desk top units. Apple doesn't claim that the G5 is a work station, it is listed as a desk top

. Thomas Carley

Allegations that desktop Alphas couldn't run common PC applications are false. Alphas, including the Multia/Universal Desktop Box line, ran Microsoft Office, Adobe Photoshop, Corel's WordPerfect Suite...

Workstation Alphas were introduced in 1992. 64-bit Alpha desktops were introduced in late 1994, more than eight and a half years ago. The Multias were personal computers in every sense of the word, and they coexisted with DEC's workstation products.

Additionally, the desktops were powered by Alpha processor models with names like 21164PC when the workstations held chips with designations such as 21264, notably lacking the "PC" suffix.

Apple simply is not the first to the 64-bit PC/desktop space, no matter what they claim.

Pete

Clockspeed is the number of times a processor carries out an action in a second. But the power of a single 'action' a processor carries out varies immensely from processor to processor, with CISC chips having very simple actions while RISC chips have more complex ones (that can take several clock ticks to carry out). This is further complicated by the fact that the long queues in modern processors mean that when all is going well they can do complex things every clock tick, but if they fail to keep the queue filled they can spend several ticks waiting for more information to arrive.

Oh, not to mention SIMD and MMX which both perform a single instruction on multiple pieces of data in a single tick.

For a single model of processor, higher clock speed is better than lower clock speed, but the size of cache (to keep the queue filled) and memory manager (ditto) have huge effects too. For different models, all bets are off.

Andy D

Peter N. Glaskowsky Editor-In-chief, Microprocessor Report

http://www.instat.com/analysts.asp

So everyone can now throw a stone or at least laugh at his picture. Is that his own hair?

To be fair " clock frequency per se is essentially meaningless", it is only comparable for CPUs that execute the same number of instructions per clock cycle, even this is a simplifictation however it's vaguely useful in that context....

Love,

Fairy

Hi,

I was hoping this matter would be laid to rest by now, but unfortunately you published another article that is likely to discredit Apple fo no reason whatsoever. This annoys me because Apple is a very positive, if small force in the IT business whose influence far exceeds it's market share, to the benefit of us all.

I have just read the Peter Glaskowsky article you pointed to, and found it to be a very poor article indeed. He makes several claims and mentioned several chip technologies which, besided being factually incorrect, have nothing to do with Apple's G5.

He states (correctly) that Apple calls the G5 a "personal computer" and a "PC", and goes on to mention a number of claims Apple made at the developer forum announcement. Up to this point I agree with him, it is also my perception that Apple made these claims, such as "worlds fastest personal computer", "highest clock rate 64-bit CPU" etc.

Then he goes on to state that Apple is lying and that the claims are not true. However, he DOES NOT substantiate his own claims and does not mention specifically which of Apple's claims are untrue or "lies". He drags a lot of unrelated terms into the argument such as Alpha, RISC, SPARC etc., which have nothing to do with Apple's claims.

He apparently allows others to make "marketing statements", but does not allow Apple the same marketing leeway. In the past many years Intel has been hyping up clock speed, and the non-techie PC buyers now equate clock speed with performance. It seems fairly logical to me that now that the G5 has a reasonable clock speed, that Apple would use that as a marketing tool. Apple would be irresponsible not to.

Then, he goes on to confuse "personal computer" with anything else that has a 64-bit cpu in it and fits on or near your desk. Apple's claim to being the first "personal computer" with a 64-bit cpu is, and remains TRUE. There is not, and there has not ever been a 64-bit computer that was intended as a general purpose, mass-market personal computer.

Of course, it all depends on your definition of "personal computer". In Apple's case, we should look at where Apple intends to sell the G5, which is to the general public, graphics professionals and film, photo and audio specialists. They want to run Microsoft Office, PhotoShop, Final Cut, iTunes and other personal computer applications on it. To about 99.999% of the personal computer buying public, the G5 is a personal computer.

All the other 64-bit machines, including the BOXX and the Alpha's of yore, are WORKSTATIONS, and not intended to be sold as personal computers. Only the Alpha's come close because you could also run Windows NT on them. However, that was about all you could run on them. There were no other PC applications available, not even Office. That's why they never sold well. The general public did not consider them a PC, besides the price was too high.

Claims made by marketing departments should be taken in context, in the same manner as we all do for other products, including those from Intel and Microsoft. When Microsoft claims to "do more with less" as is the current Server 2003 advertising campaign, we all know this to be factually untrue. Why don't you take Microsoft to task for this?

So, I would like to hear from either you or Peter, where Apple has been telling lies.

Regards,

Maarten

"By the way, if clock frequency per se is essentially meaningless, why has it made Intel, AMD, DEC, Sun, HP and others - including the Microprocessor Report -- gazillions of dollars? "

Clock frequency is meaningless, as any metric would be, if the limiting conditions of the system exist elsewhere in the system.

I.E., when data throughput is limited by bandwidth, then increasing the clock frequency above a certain level will no longer improve performance. Increasing bandwidth will increase performance.

One hopes the wonderful free market will encourage development of systems, code and peripherals to take advantage of the graciously provided excess clock cycles our chip manufactures are providing.

name withheld to protect the innocent etc

AMD could help speed nuclear holocaust

Hey Mike,

David Brown should get a life. The onus here belongs on the U.S. government which is the entity that approves or disapproves the sales of supercomputers to China.

What I have to wonder is where Mr. Brown was when Silicon Graphics sold two supercomputers to China in 1996.

http://www.nti.org/db/china/supercom.htm

"China through normal trade channels bought some supercomputers from the United States and their purpose is for weather forecasting, earthquake prevention and other scientific research…This is a totally commercial action…The Chinese side…strictly abided by the terms of the contract…This kind of purchase went through a comprehensive examination and approval process".

Or when the following incident happened:

http://www.nti.org/db/china/supercom.htm

Despite the controversy, on 16 September 1999 the US announced new export regulations for encryption technology which were expected to boost exports of computer technology, including supercomputers, to China. The new regulations, which were implemented on 12 January 2000 state:

"Any encryption commodity or software of any key length can now be exported under a license exception (i.e., without a license) after a technical review, to commercial firms and other non-government end users in any country except for the seven state supporters of terrorism.

Or maybe he should look at China's own developmental capability before the Opteron ever existed and before AMD ever thought about selling processors to China.

In January 1996, it was reported that China included the Dawn-3000, a new model of its Dawn-series mass parallel programming (MPP) supercomputers, as part of its ninth "Five-Year Plan" (1996-2000). The Dawn-3000 will have a 3,000 MTOPS capability.

In June 1996, China announced the Yinhe (Galaxy)-3, a new high-performance supercomputer developed by its military. The Yinhe-3 has a 13,000 MTOPS capability, and is equipped with an expandable multiple processor parallel system. The Yinhe-3 is 10 times more powerful than the earlier Yinhe-2, but only one-sixth the size.

China has also developed the Yinhe Super-Simulation Computer (YHSSC), successor to the Yinhe All-Digital Simulation Computer System Type-2 (YH-FZ2). The YHSSC has been used by the China Aerospace Corporation (CASC) for simulations to improve its rocket systems; by the Southwest Research Engineering and Design Academy (First Academy) to design the Qinshan-2 power reactor; and by the China Aviation Industries Corporation (AVIC) to develop a next-generation unmanned aerial vehicle (UAV).

In January 2000, the State Intelligent Machinery Research and Development Center of the Chinese Academy of Science Computer Institute announced the development of the Shuguang 2000-II SuperServer. The supercomputer is reported to be capable of performing 110,000 MTOPS.

So my question is Mr. Brown with all of the previous technology transfers, missle booster technology, nuclear warhead design, supercomputers, encryption technology, and the probable theft of nuclear simulation software, that it is only now that AMD is getting ready to sell China some processors, that you are complaining? The sale of which, will be ultimately approved or disapproved by the U.S. government not AMD, which has asked for the proper licenses.

Pat

Mike,

I have to take serious exception to your last paragraph in the above article where you characterized Barbados as a US colony.

We are not now and never have been a US colony or flunky. We are currently on the US blacklist for numerous things that we did not do the US way. We did not support the Iraq war. We will not sign an agreement exempting US nationals from being hauled before the new world court. We were one of the first countries to establish diplomatic relations with Cuba against the wishes of the US and I could go on. As a matter of interest the US constitution was based on the Barbados constitution.

Please wash your mouth with soap for saying this.

Michael

[I do apologise to Michael and other readers from Barbados. Mike Magee

The only reason that the single P4 could outperform the dual P4 has to be the faster FSB.

That more or less indicates that this particular program is memory intensive in its use. The dual P4 operates on a shared bus with memory contention always a possibility. The Opteron presumably has one bus for each CPU and such programs will always show superior improvements on Opteron, or on any CPU with a similar architecture. (None presently, but that could change in the future.)

It's also worth noting that that are already boards being shown for Opteron with only 1 memory bus for two CPUs. That will handicap Opteron in situations where the program requires more memory than the onboard cache can hold.

Accordingly, I would expect that such boards will show not much better scaling on large, memory intensive progams than the current P4 does. The HT links will help some but will not solve all problems.

Time will tell, but such shortcuts will inevitably affect the perception of how well the Opteron can scale.

Caveat emptor.

Cheers,

Frank

MailWasher blackballs, blacklists the Spamsters

Following your article on mailwasher, you could offer some ready-made filters that could be cut & pasted into an end-users` filters.txt

I could start you off with some if you like (just let me know)

Here`s a little trick i`ve just found in my inbox too (courtesy of mailwasher) - the same email to two different email addresses purporting to be from Holiday Inn, full of identifying links, but from two different originating sources with an "opt out" link - anyone who didn`t know it *wasn`t* from Holiday Inn might be caught out and added to yet more spam lists.

The body of the email contains database riddled links which direct you to sites who are probably offering the spammer money for referrals...

Headers are shown below for your perusal (please don`t repost my email address or these headers though if you follow up the original story :-} )

Colin Wilson

HP says there's no engineering talent in Silicon Valley

dude,

Roseville is *so* not Silicon Valley. HP offered to move me there when my Network Server project was relocated out of Cupertino here in the Valley up to Roseville in the fall of 2002. Like most engineers in the group I declined. Most of us left HP during a recession and stayed in the Valley rather than get stranded in Roseville, where there is ABSOLUTELY NOTHING ELSE if the HP job fell through.

So this HP manager guy may be right, he can't attract top talent to Roseville. He certainly could attract people if the jobs were for Cupertino site, where there are empty buildings (e.g. old Compaq/Tandem space on Tantau Ave.). But as we all know it's about cost. HP could find as many engineers as it wants in the Valley proper (not Roseville) but they will get much more bang for the (Singapore) buck in the far east.

Those guys are good too, didn't they design the original Jornada out there?

"The Quantity",

h/w engineer
ex-hp Network Server Division

FBI can't wiretap voice over IP

Hi Mike,

I live in Serbia & Montenegro (former Yugoslavia). We have only one state owned telecom company. Weird thing happened to a friend of mine.

She was using voice over IP to talk to her relatives in USA. Soon they called her from local pbx office and told her that they will cancel her phone contract (take her phone line) if she doesn't stop doing it. Now if they are not able to spy voice over IP how the hell did they know that she was using it?

Forgot to mention that the majority of internet providers were connected to the www through the telecom's internet network at that time.

Here is what I think:

1) She used to call USA the regular way before. Her bill dropped and they noticed it.

2) They set some kind of monitoring on her phone line to alert them when the line is used for data transfers.

3) They knew the ISP dial-up number she dialed because they can log both incoming and outgoing calls.

4) Once she was connected to an access server at the ISP they knew the IP address she was assigned.

5) They set packet sniffer on a backbone for packets with that IP and discovered what type of data she was transmitting.

Sounds SCI-FI but I can't think about any other way they could knew what she was doing.

Perhaps someone with trained ears can differentiate between modem buzzing while surfing and while making voice over IP conversation? My ears certainly aren't that trained.

As you can guess my country doesn't have the latest tech gadgets the FBI uses every day but if they could manage to do it -- the FBI must be already doing it for quite a long time and now is just seeking the public approval.

More MailWasher
About your mailwasher review, I've been using Mailwasher for a while now, and I can't say I disagree with your finding. Two caveats however: 1) I've been told the bouncing feature is a *terrible* idea; most spammers won't send you a mail from a functional email address. The address in most cases just won't work/exist. Since the bounce message Mailwasher generates is supposedly sent from "postmaster@you.isp", and the bounce messages will be bounced by the mailserver from the non existing spammer's address (get it ?), this will result in a massive flood of emails at your ISP's postmaster address. This doesnt do anyone any good.

2) Mailwasher lacks an autoexecute option (at least last time I looked at it) I would love to have Mailwasher run nonstop on my machine to clean my mailbox while I'm abroad (so my webmail is still useable), or even have my computer boot up automatically every day and run mailwasher, then shutdown, but there is no such option, unless you train your dog/cat to push the right button once in a while :(

regards,

P4man