Intel claims hyperthreading produces 900%+ boost

The Inquirer ^ | December 6, 2002 | Mike Magee

[JameRetief]

Truly remarkable results gobsmack us

A SHARP EYED contributor to Ace's Hardware Forum has noticed that Intel is making some truly remarkable claims for the 3.06GHz Pentium 4.

A demo on its site – it requires Flash to view – compares a spreadsheet calculation. That takes five minutes 39.9 seconds on a 3.06GHz Pentium 4 without hyperthreading while with hyperthreading, Intel claims it takes 35.58 seconds. [ Actually, if you see the demo, it is supposed to take the non-hyperthreaded P4 an even longer 6 minutes 15 seconds. (5:40 + :35) ]

This is a remarkable feat and leads us to believe that hyperthreading is better than anyone could possibly have anticipated.

Indeed, with the performance of your processor boosted by over 900%, truly no-one should be without one.

You can run the demo from this Intel page, while the Ace's thread starts here. µ

[JameRetief]

This is simply unbelievable. I would not have believed Intel would have put up such a sham "benchmark" if I hadn't seen the demo myself. *tsk* *tsk*

[JameRetief]

Marketing Department Gone Wild, ping.

[pseudogratix]

Intel Shows Speedier, More Secure Chips

Banias mobile processors, hyper-threading technology, and a new security initiative highlighted at Developer Forum.

Tom Krazit, IDG News Service
Tuesday, September 10, 2002

http://www.pcworld.com/news/article/0,aid,104807,00.asp

[pseudogratix]

http://slashdot.org/search.pl?query=hyperthreading

[js1138]

That's some calculation.

[goldstategop]

Is a processor chip with hyperthreading worth the extra price tag or is it just all hype??

[error99]

"Is a processor chip with hyperthreading worth the extra price tag or is it just all hype??"

Sounds like a load of "hypermarketing" to me...

[js1138]

I think a 9X boost in performance would merit the front page of every newspaper in the world. Something is amiss with the claim.

[Technocrat]

Well, some tasks are extremely well suited to parallelism. Most are not :)

[VeritatisSplendor]

If it can improve performance on ANY task by 900% then it will have a significant impact on many others, though not necessarily by the same proportion.

You would be surprised how much can be "parallelized" -- most software is designed for serial processing because that is how most programmers have been trained to think. But the number of tasks that an ordinary computer does that are ESSENTIALLY serial is small.

I agree with you though, that cooking up a benchmark that maximizes the performance improvement can be very misleading.

Cool screen name. It's been 15 years since I read those books, I'll bet a bunch more have come out since then.

[Eastbound]

"I think a 9X boost in performance would merit the front page of every newspaper in the world. Something is amiss with the claim"

I sent the link to this to my son. He wrote back:

"HyperThreading at this point is mainly Hype...aptly named. The applications need to also support HT. Here's a blurb from a comparison site:

"Benchmarking the Pentium 4-3.06 GHz is a very difficult proposition, as its main benefits are easily apparent using standard performance tests. Unless a program supports multithreading, then any performance benefits are extremely hard to quantify. The benchmarks that do support multiple threads showed some serious gains, while standard single-threaded tests were on-par with a 3.06 GHz CPU. The more ethereal advantages are when actually using a HT-enabled system, and multi-tasking with resource-hungry applications. While the numerical results are not consistent enough to publish, we found speed jumps of 5-15% when performing duties like simultaneous MPEG and MP3 encoding."

Now the kicker:
Intel 3.06GHz cpu: $710
Fastest AMD XP 2700 / 333MHz fsb: 323"

FWIW

[Eastbound]

Meant to ping you, but your screen name wouldn't take the first time around.

[js1138]

5-15% is a pretty good benefit. Most high-end applications are multi-threaded -- Photoshop, Autocad, any professional graphics or multi-media editing program.

[dennisw]

I read hyperthreading gives an average 20% boost. That's it and you'll pay for it. The 3.06 price is $709. Almost twice what the next one down is, the 2.8 at $377.

[HamiltonJay]

Ah yes, the "arbitrary and creative truth" that is marketing got to love it. Sorry guys, while hyperthreading is decent technology, and can certainly help with tasks that can be broken into independent seperate non dependent parts, 900% increase? Uh... No.

[js1138]

Intel's first string CPU always cost $700. This is to get the big bucks from the rich and desparate. I always buy a couple of notches down. But in the last two years, I'v bought nothing but AMD.

[LivefromVegas]

It would be smart of them to put this on thier Xeon chips, anyone running a dual or Quad CPU Server would be would have 4x's the processing speed of anything out there on the X86 base

[dennisw]

Intel's first string CPU always cost $700.

True. It's in Dell, Gateway etc but I'll bet they only pay half that $719. If you build your own you would be a fool to pay for the best Intel. I have two home built AMD computers. I will upgrade one with  the 2100+ Athlon (~$87) which will be more than enough for my purposes

[gura]

Finally, something that'll run Battlefield 1942 well!