Journalism is a trade rather than a profession, a bit like bricklaying
|
|
|
Vendors can do much more to stop viruses
Cisco and ISPs SHOULD block viruses leaving their network, even at their borders - that would go a long way to helping. The technology exists TODAY. Do it.
Microsoft and Anti-Virus companies need to look at themselves serverely. Our clients had to WAIT for their virus company update to get hold of the new definitions for the new variant of W32/Mytob. Even then they had to do it manually at first.
Surely it's possible to write a virus scanning application or email program that detects that 500 emails are leaving a machine per minute with the about the same executable attached. Why would you then have to wait for virus definitions?
The kicker in all this is that the "new" virus definitions identified the virus as W32/Mytob.gen (generic).
If all of the the above suppliers tried harder, the crap traffic on the internet would drop, and sales of their products would drop (Cisco and Anti-Virus companies especially). Crucially, ISPs would see a huge drop in incoming traffic which they charge to users whether they requested it or not.
I'll bet Microsoft market Longhorn as their "most secure operating system EVER". It had better be, but that begs the question - why didn't you get it right last time! "Scriptable everything" has been a huge mistake in my opinion.
It's in their interests to do nothing.
Regards,
Connor
Dave Y claims: > So, there is no magic, you can't possibly > do more than one instruction in one clock > in one unit.
This is actually not correct. As an example, the VMX unit in the PPC970 is capable of a peak rate of 8 operations per clock cycle (four 32-bit floating-point multiply-add instructions per cycle).
As described in other places, the Cell processor chip employs multiple processing elements, each of which can execute multiple operations per clock cycle (in a SIMD style).
The Cell processor has significantly greater internal parallelism than the POWER5 processor, but this is only one small aspect of what would need to be considered in a comparison of the two processors and their value in specific computing environments.
John D. McCalpin, Ph.D.
HPC Performance Lead
IBM eServer System Performance department
Just because you don't understand how the cell gets the claimed 218 gigaflops doesn't mean it's impossible.
So I'll try to explain it for you without using words that are too large for your obviously limited vocabulary and mental capacity.
There are 8 vector units on the chip (+1 for redundancy) -- 1 is in the PowerPC core, and 7 are in the SPUs. Each one can do 8 floating point operations each cycle. -- a vector multiply-add (for 4 deep vectors) is 8 floating point operations, and each vector unit can complete one each cycle. (Yes they take longer than that to get through the pipeline, but that's what pipelining is all about -- you can read these things in books, you know -- you CAN read books, can't you?)
So the cell chip can do 64 floating point operations each cycle in the vector units. The scalar fp unit in the PPC core can do another two -- it can complete a single FMA each cycle -- making the total 68 per cycle. So take 68, and multiply by 3.2 billion (That's the clock speed, if you remember.). I'll wait while you reach for your calculator... See -- it comes out to 217.6 billion. Round it to the nearest whole number and you get IBM's claimed number of 218 gigaflops. (Yes, these are single precision floating point operations.)
Cheers
Ian Ameline
(I always love a good flame :-)
Love in an Intel bunny suit and Love in an Intel bunny suit part II
Dear Misty,
"I collapsed in tears. Would I ever see Babs again?"
You've kept us who are concerned with Augusto and Babs' relationship in suspense for more than half a year. Will you bring an end to my agony of waiting?
Sincerely
Thomas Yiu
Yeah, nice story about the picketers at intel Dupont. Except that it's too bad these idiots picketed the wrong place.
Turns out that Intel's Dupont Campus, has nothing to do with Hyperthreading, or competition to Athlon at all. Dupont works on Enterprise Systems, ie: Itanium. Maybe next time, they should go further south, and try for the Hillsborough Oregon campus instead.
Then at least, maybe they would have a clue :-)
Name supplied
More Dave
Hi,
In response to Dave Y's letter.
(1) You have no clue what the heck you're talking about.
The Cell processor is designed to offer "supercomputing like" performance.
It's a team project from Sony/IBM/Toshiba. Its made on 90nm SOI process from IBM's FishKill facility. (Same with AMD K8 and PowerPC G5).
If your pee-brain bothered to order your hand with the mouse to click on the Sony Press Conference on video stream, you know that's the intention of the Cell design.
Offer an extremely powerful multi-core CPU that is able to do things in realtime, all in one package. (Technically it is multi-core with all those SPEs).
Then supplement this with the most powerful GPU that Nvidia can make. (That's made from Sony's 90nm process).
(2) Have you even seen the PS3 in action?
Even without Nvidia's RSX GPU, it can handle itself quite well. Physics, AI, software rendering...All without even breaking a sweat in real time.
A Sony rep even demoed an exploding petrol station...The explosion, physics (thermodynamics, projectile motion fragmentation due to the explosion, the haze effect, etc), were all done by the Cell CPU alone, in real time.
There is so much excess power left over, that you can play a game while doing video chat with your friends or even have someone else browse the web (using PS3's ability of a dual screen setup).
Have you seen any other CPU decode 12 High Definition video streams simultaneously? If you have, links please, I would like to see it.
(3) The 2 TeraFLOP claim is the total floating point performance of the Cell CPU AND Nvidia RSX GPU.
The Cell itself does NOT offer 2 TeraFLOPS!
(4) The goal of the PS3 is to offer movie-like graphical quality games at 60fps.
What does this require? A CRAPLOAD of processing power (especially graphics), if you want to do this in real time. And Sony/Nvidia did wanted this done in real time. (In fact, they have)...They even modelled London, with people walking around, cars driving buy, the well-recognised red double-decker buses moving around.
I really like to know how YOU would squeeze enough performance into a PS3 chassis, fullfill Sony's demand for lifelike graphics, and keep frame rates at 60fps. (and to top it off, support 2k x 1k resolution). Then put all that power in one reasonably small enclosure.
Please, enlighten us with your technical prowess and engineering knowledge.
(5) You can't compare numbers based on two different designs. You're comparing a multi-core solution to your dual-core SMP example.
Yes, both the G5 you mentioned and the Cell CPU are fundamentally PowerPC architectures, but they are very different implementations.
According to Ken Kutaragi at the Sony Conference for the PS3, the Cell CPU's Floating Point performance at 3.2Ghz is 318 GFlops. This is twice that of the Xbox 360.
(Yes, Ken paused for a brief and sniggered at the fact that his "baby" has twice the floating point grunt of the MS design...And some of the audience giggled at that thought too).
This still is a significant amount of grunt to play with, coming from a multi-core CPU.
The only thing "stupid" is your inability to find out what that 2 TeraFLOP claim involved.
Not only did you not properly find out what the "marketing use" of 2 TeraFLOPS applied to, you merely continued to ramble on with your dribble, dragging in the G5 as your example.
Seriously, next time you send a whinge to TheInquirer for public viewing...Check the facts, because you look like a damn fool to a million (?) or so visitors of TheInquirer.
The question mark is there, as I have no information on exactly how many hits theinquirer.net actually get per day.
Regards
Name supplied
Intel shows imagination, signs graphic licences
The chumps at Intel had first crack at the purchase of 3DFX BEFORE they filed chapter 11. They passed up the better graphics technology from 3DFX than they could have ever hoped for. Bit Boys were a company whose IP was purchased by 3DFX. This IP allowed 3DFX to slam way past the likes of nVidia and ATI. It allowed for the use of slowere memory due to the tilized method of processing the video. The tilized method was designed around low cost hand held GPU controlers. Intel is a joke.... They couln't find their collective arses with both hands....
Had they purchased 3DFX.. this may have taken processors in a completely different way. The reason for this is that GPU manufactures rely on much higher performance EVERYTHING. This could have been a way to leverage mutiple GPU's / CPU's on a single die. Intel like the rest doesn't see the future. The future is a one chip system... Its only a matter of time.....
Ron
The Yankees and Bargain Buys
No not the baseball team. I'm from the mainland US, and read your webpage every day, first thing, but it sucks everytime you guys have like deals of the day, or anything else that can only be taken advantage of buy mainland UK peoples. Now, I drank with a couple English peoples while I was staioned in Hawaii, and as a whole, you guys were the funniest lot to drink with, on account that you could keep up, or outdrink me (girls included, very impressive), and actually hated France more then I did (something I didn't think possible), so come on, show some love and include a deal that has a $ sign in front of it every once in a while, or some other way to let us (I'm sure there's others) americans who rely on your website for news feel more at home, thanks.
Mark
Share this:
Share
