First Dothan Review & Benchmarks on Tom's Hardware - Page 3

A couple of random thoughts.

I can see why Dothan 2.0GHz was recently added to the list (originally only 1.7GHz and 1.8GHz were to be released). A comparison of Dothan 1.8GHz with Banias 1.7GHz would have lead to an anti-climax, especially after so much waiting.

The relative speed increase between 1.7GHz and 2.0GHz is 17.6%. So if the top Banias 1.7GHz were compared with a hypothetical Banias 2.0GHz, CPU benchmark increases of about 17% would be perfect.

Since Dothan 2.0GHz yields an increase of about 22%, I won't complain, although I was expecting a bit more. Previous rumours only indicated improvements of "up to 15%" at comparable speeds, so under ideal circumstances, one could only hope for an increase of at most 35%.

As for prices, it wouldn't be realistic to expect Dothan 2.0GHz to cost the same as Banias 1.7GHz did last week, i.e., $423. Of course, $637 for Dothan 2.0GHz looks steep, but on the other hand, you now jump from 1.7GHz to 2.0GHz for $214 ( = 637 - 423), whereas last week going from 1.5GHz to 1.7GHz cost $182 ( = 423 - 241). If you are less ambitious, you can now get Dothan 1.8GHz for the old price of Banias 1.7GHz. Thus, as with previous Intel's releases, there is a premium price for the top model, and the second-best model (here Dothan 1.8GHz) may be more cost-effective if you don't care about (or can't afford) the best performance.

There is little need to discuss Banias 1.7GHz vs Dothan 1.7GHz, since they cost the same, and Banias 1.7GHz will probably disappear from the market very quickly.

As noted ealier on other boards, and finally here, the fact that THG's gaming benchmarks show no difference between Banias 1.7GHz and Dothan 2.0GHz may be due to these benchmarks' being essentially insensitive to processor speeds, at least at 1.7GHz at higher. We must wait for comparisons on more CPU-intensive games. There is no Dothan-related magic here; it may well be that Banias 1.7GHz was already fast enough for THG's benchmarks.

Now a couple of specific replies.


rincewind:

THG's tests indicate nothing about the cache influence. You seem to believe that it somehow slows down Dothan in practice. This may mislead some users.

The 21W for Dothan and 24.5W for Banias refer to "Thermal Design Power"; this doesn't mean "average dissipated power" (see Intel's documentation for Pentium M). What is much more important here is that, although the frequency increased from 1.7GHz to 2.0GHz, Dothan doesn't require more sophisticated cooling (contrary to some previous rumours), and doesn't eat up batteries faster.


Enderet:

Where did you see Dothan 2.0GHz being beaten by Banias 1.7GHz?

In the UT2k3 benchmarks, the 1.7 beats the 2.0, in the SXGA benchmarks, as well as in the UXGA benchmarks.

Edit---> ALso in the second Aquamark 3 scores.

toms hardware 'hints' that the doubled cache leads to performance decrease... yet they do not outright blantly say so in the tests...

that was my own opinion based on previous knowledge/analysis on p4's with larger cache and athlon64's with smaller cache... if u dont think having a larger cache leads to decreased cpu performance and that there isnt any latency created by having excess... well...i'd suggest u read up on how amd is going to increase their clock speeds with their newcastle design

hardwareanalysis.com has an excellent review explaining how having a larger l2 cache creates latency and decreases cpu performance slightly by 5-10%

http://www.hardwareanalysis.com/content/article/1686.8/





when i mentioned the TDP of dothan/banias that was to emphasize my point on how unsiginficant that change had in regards to extended mobile battery time. (I was NOT implying that dothan would consume mobile battery time faster, only stating that because of the lower TDP, dothan only gained 5 more minutes according to the tests... which was so unsignificant)

i would be careful when u say dothan performs 22% better, because that is only in video rendering/audio encoding/workstation benchmarks

for all other intensive purposes, tomshardware benchmarks show that there is no performance increase in gaming (and i was merely stating that banias performed slightly better over dothan in some tests, and that was probably due in part to 'latency' created by having a doubled cache size)

all other benchies show only a 5% gain when comparing the latter

Enderet:

I guess you are just joking here.

In UT03, look at the figures: (28.47,45.15,71.57) and (17.2,33.31,56.85) for Dothan, (29.55,45.6,71.53) and (17.84,33.68,56.8) for Banias. Even if THG gave averages for five runs (say) and provided standard deviations, I don't think these numbers would be distinguishable on statistical grounds. Of course, THG doesn't expect its readers to know about statistics, and simply mentions "measurement errors".

For Aquamark3, Dothan looses by a whooping 0.7%!

Anyway, if you think that these figures justify your statement that Dothan "gets beaten by the 1.7 Banias at several things", think again.

yes, banias beats dothan in some benches, but its so unsignificant between the two that the price difference of 300 dollars and 750 dollars is outrageous

thats the point were trying to make

not glorifying banias as bein the better cpu, just the more value cpu

dothan is an evolution of banias to continue the pentium-m line, thats all it is, there is nothing revolutionary about it

lol, i just caught that... did you just divide 2000 by 1700 to get that percentage?

scaling megahertz numbers do not necessarily equate the same actual performance number, especially when you need to factor into the different design variables between banias 1.7 and dothan 2.0. (larger cache, strained silicon .09 micron manufacturing process, etc)

and that point is made in THG's reviw.....

22% better in encoding/rendering/workstation
0% difference in gaming performance
about avg 5% better in all other performance tests...

that would be a more accurate way of comparing the two

It's almost like an AMD vs. Intel or Mac vs. Windows thread in here.

You've gotta admit it's not as bad as Prescott. The benchmarks and battery life for the same speed Dothan are essentially equal (within standard deviation) or better than Banias depending on what you're running. At worst, you're looking at a speed bump for the Pentium M line, which is a good thing.

Personally, I do more encoding and video work than gaming, so Dothan is an improvement. I'll take every second I can get when doing multi-hour video encodes. I'm also guessing Dothan's performance increase also extends to decoding, which should be good for High Definition video playback.

Finally, everyone benefits from the price cuts. The 1.7GHz Dothan looks like the best price/performance after this -30% drop. It doesn't look like we're going to see Banias prices dropping much below Dothan, but if they do drop a bit before being phased-out, Banias will be a great bargain.

yes presscott was just a joke

im glad intel is coming to its senses and scraping plans for tejas and jayhawk or whatever that other next generation p4 was

they were hitting a barrier beween increasing clock freq while trying to maintain adaquette operating temperatures...

the presccot 3.4 was reviewed and ran at 76c with the case OFF under fullload, thats damn hot, i could cook food over it

they are turning towards utilizing pentium-m as dual processors since its intels only real 'efficient' cpu

rincewind:

Concerning the cache influence, let's recall your earlier statements:

"however, there is 0% difference in gaming performance when comparing 2.0ghz dothan vs 1.7 banias (this is due to the fact that games utilize more of the GPU rather then CPU, and more cache creates more latency).

and 22% difference in media encoding/workstation performance when comparing 2.0ghz dothan vs 1.7 banias (this is due to its double cache size, not higher clock speeds)."

So your first statement indicates that having a larger cache is bad, while the second statement says that Dothan performs better only due to its larger cache. Isn't that inconsistent, to say the least?

I wrote that "THG's tests indicate nothing about the cache influence" because this aspect has not been tested by THG. Things would be different if THG compared Dothat 1.7GHz with Banias 1.7GHz. Until such comparisons occur, there is little point in speculating on the basis of how cache sizes influence performance in other architectures.

My reference to Dothan's performance increase of 22% was clearly restricted to CPU benchmarks.

As for your subsequent messages:

"yes, banias beats dothan in some benches, but its so unsignificant between the two that the price difference of 300 dollars and 750 dollars is outrageous"

I tried to explain in my reply to Enderet that the data available in THG's review don't support any conclusion about Banias beating Dothan. If you are not convinced, peek at any textbook on data analysis.

And the next one:

"lol, i just caught that... did you just divide 2000 by 1700 to get that percentage?"

Yes, and I didn't say I was comparing Dothan 2.0GHz with Banias 1.7GHz at that stage. Read the whole paragraph, and avoid "reading in" things I didn't say.

BTW, in the Conlusion section of THG's review, the statement

"Even if you only consider the difference in clock speed between the two CPUs, Dothan still offers a 5% advantage."

may be hard to understand; what the reviewer probably wanted to say is that even at the same clock speed, Dothan still offers a 5% advantage.

I am definetely note joking. Sure it only beats it by insignificant numbers, but like ricewind said... and I have also... why pay the extra money for Dothan when a 1.7 Banias gives you about the same performance in games? I sure as hell wouldnt pay 500 more for Dothan, just to get the same performance I would get with a 1.7, in video games.

Edit--> Ultimately, Dothan gets beaten when it comes to bang for the buck.

i dont think u seem to understand that l2 cache is a double edged sword
please dont try to get technical if you dont understand how it works

If you increase the L2 Cache size, you will gain benefits for programsthat require large cache. This would include encoding/decoding/rendering/workstation applications. However, as L2 cache size increases, CPU performance decreases because of latency created in having an excess.

Having said that, the same could be said in a reverse role; Having a smaller L2 cache allows the procssor to perform faster, but at the same time it suffers in performance when you need it for encoding/decoding/rendering/workstation applications. etc.

so, yes, i am inconsistant in regards to L2 cache, because it IS in consistent
its a double edged sword.

ok ur a funny guy coz CPU benchmarks reveal @ THG reveals only that its only 5% better

22% better perfromance is for encoding/decoding/workstation benchmarks

ok listen i dont need to learn how to interpret data
u are just insulting my intelligence now

do u work for intel or something?



i am not stating or have ever stated that banias beats dothan. i only once pointed out an irregularity in benchmark scores where banias beats dothan by a few points/frames (not significant, as i stated earlier)3

my main point, if you didnt catch on by now is that dothan is not revolutionary and the hype behind it all needs to be downplayed coz its only an evolutionary process in furthering the pentium-m line

bloke:

I share your assessment, as evidenced in my first post. Good to see that others are prepared to take a more objective stand on Dothan.

The reason I joined this discussion was to show that the initial Dothan bashing, although understandable on emotional grounds, had been at least partly based on misconceptions and wrong interpretations of THG's review.

Thus, except for lost hopes/illusions on the gamers' side, nothing wrong has happened today, and we should be more positive in appreciating yet another technological advance; the days of significant breakthroughs are gone.

no. he probably did NOT mean at the same clock speed, cause he stated when you consider the DIFFERENCE in clock speed...

nice try tho

I wonder how a 1.8Dothan will hold against a 1.7 Banias?

i want a review on the new 34w 2800 'true' mobile athlon64's

Dothan chip gains early backing from six PC makers

The tomshardware review is probably accurate, and Dothan is probably a disappointment, but, this is just ONE review. We should maybe wait for anandtech sharkyextreme or another site to agree or disagree with this review before passing total judgement. Just a thought.
Andrew
Austin, TX

So then, am I right in assuming from all of this bickering that a 1.7 or 1.6 Banias will outperform a 1.7 Dothan?