Dothan Pentium M performance bets?

Message

#1 Post by **K. Eng** » Sun May 09, 2004 6:26 pm

Less than 24 hours to go... Anyone want to wager on the performance of Dothan?

I'm predicting the 2.0 GHz Dothan will achieve SPECint2k between 1300-1400 and SPECfp2k of 900-950.

These are base scores. I calculated this figure using known performance values of Pentium M 1.0 GHz, 1.50 GHz, and a powerpoint presentation I found on a Chinese site concerning per/clock improvement of Dothan v. Banias.

RadMan · #2 Post by **RadMan** » Mon May 10, 2004 7:35 am

will surpass the fastest intel desktop CPU in atleast 50% of benchmarks.

G-Man · #3 Post by **G-Man** » Mon May 10, 2004 7:55 am

RadMan wrote:will surpass the fastest intel desktop CPU in atleast 50% of benchmarks.

You are kidding, right? I guess that's a little bit too much.

G-Man

#4 Post by **JHEM** » Mon May 10, 2004 8:07 am

RadMan wrote:will surpass the fastest intel desktop CPU in at least 50% of benchmarks.

Against a P4 3.4GHz with HyperThread and the 1MB cache? Not a chance!

Regards,

James

Conmee · #5 Post by **Conmee** » Mon May 10, 2004 8:23 am

I think synthetic FPU intensive benchmarks may show the Dothans to challenge and/or exceed the performance of some of the desktop CPUs, but 50% of the benchmarks? I would have to agree with the other incredulous souls in this thread... otherwise, Dothan would be the clear choice over any desktop/workstation Intel CPU except for instances where a user needs superior SSE2 execution (desktop CPUs will still do this better). We'll all see soon enough...

G-Man · #6 Post by **G-Man** » Mon May 10, 2004 10:06 am

There is a review at Tom's Hardware of a WinBook powered by Dothan. http://www.tomshardware.com/mobile/20040510/index.html Of course it's not a TP, but gives us a good starting point what to expect.

G-Man

Conmee · #7 Post by **Conmee** » Mon May 10, 2004 11:17 am

G-Man wrote:There is a review at Tom's Hardware of a WinBook powered by Dothan. http://www.tomshardware.com/mobile/20040510/index.html Of course it's not a TP, but gives us a good starting point what to expect.

G-Man

Based on what I just read, it seems Dothan's only significant advantage is in the fact that you can get more GHz without sacrificing battery power. Perhaps beyond 1MB cache, there is significant diminishing returns... this configuration is also almost identical to the ThinkPad T42 2373-3VU... so for people looking at that model, this article will give you a good idea of expected performance.

Daniel.

#8 Post by **K. Eng** » Mon May 10, 2004 1:22 pm

Supposedly the 1.60 GHz Dothan gets around 1200 in SPECint2k.

http://www.heise.de/newsticker/meldung/47222

Im c't-Labor absolvierte ein Notebook mit der 1,6-GHz-Version des Dothan die meisten Benchmarks rund 4 Prozent schneller als mit einem gleich schnell getakteten Banias. Vor allem bei rechenintensiven Benchmarks fiel der Vorsprung noch deutlicher aus. Insbesonders beeindruckt das SPECint-Ergebnis: Mit 1206 Punkten hängt schon der 1,6-GHz-Dothan jeden Celeron und jeden AMD Athlon XP ab.

G-Man · #9 Post by **G-Man** » Mon May 10, 2004 1:36 pm

Yeah, but according to the article, most of the benchmarks showed only an average of 4 percent increase. That's not much.

G-Man

nikemen · #10 Post by **nikemen** » Mon May 10, 2004 1:49 pm

At the end of the article, it goes on to say that the new dothan are more power hungry than the predecessor and if you are expecting more battery life, forget it.

Hopefully, this won't mean even WORSE battery life.

nikemen · #11 Post by **nikemen** » Mon May 10, 2004 4:21 pm

Looks like speed is speed, so more is better. But, that with the higher speeds Intel has managed to keep the battery life constant at least

http://www.tomshardware.com/mobile/20040510/index.html

#12 Post by **K. Eng** » Mon May 10, 2004 6:03 pm

G-Man, 4-6% performance increase per clock is consistent with doubling the L2 cache size. I was hoping for a bit more but that may have to wait until the Alviso chipset (faster FSB, dual channel RAM) becomes available.

Nikemen - the power usage curve of Dothan is interesting. At max speed it uses less power than Banias, but at lower speeds (1 GHz or so is the crossover point) it uses more power. I think the average power consumption of Dothan is higher than Banias, but that this is somewhat offset because Dothan's cache allows for fewer memory reads and hence less use of the memory system.