Sign in with
Sign up | Sign in
Your question

Another Dothan Review

Last response: in CPUs
Share
June 4, 2004 1:15:12 PM

<A HREF="http://www.gamepc.com/labs/view_content.asp?id=dothan&p..." target="_new">Click!</A>

I don't understand Intel! Make these Dothan available for the desktop fast! These would kick the AMD butt!

--
What's the <b><font color=green>AMD Mobile Athlon 64</font color=green></b> overclocking potential? <b>It's huge!</b> Humm... Maybe not that huge...

More about : dothan review

June 4, 2004 2:48:49 PM

Well, we're seeing the same limitations in Dothan that put the original P3 out of the running years ago.

1) low clock speed. the P6 core in Dothan always seems to tap out on clockspeed quite a ways behind AMD's parts at the same trace size. 130nm only goes to 1.3GHz (vs AMD's 2.4GHz), 90nm is only going to 2GHz so far. By the time 90nm Dothan catches up to 130nm AMD64 in clockspeed, AMD will probably have 90nm ramped and going.
2) slow FSB speed. maybe Intel can correct this.
3) yield, possibly? one of Intel's chief weaknesses in the P3-vs-K7 days was launching competitive parts made of anything more than paper. Intel can get away with low yield on laptops, but producing Dothan in Intel's classic desktop and server volume is a bit of a pinch. Having a whopping 2MB of cache isn't exactly going to help yields either--see P4 Elusive Edition.

Still a very nice chip though. It could easily be my chip of choice for 32-bit blades and media-center PCs.

<i>Look! Up in the sky! It's a bird! It's a plane! It's...

...an asthmatic werehamster?

<LHGPooBaa> Well, @#!& on me.</i>
June 4, 2004 3:36:29 PM

Quote:
low clock speed. the P6 core in Dothan always seems to tap out on clockspeed quite a ways behind AMD's parts at the same trace size.

I partially agree with you. We have seen overclocked DOTHAN @ 2.4GHz in some reviews. So, consider Dothan are binned for their MOBILE quality, Intel could probably make faster Dothan based desktop CPU with higher Vcore (A Dothan @ 1.5/1.6 Volts is probably stable at 2.0GHz). An considering the Dothan 1.7 versus the A64 1.8 results. Desktop Dothan at 1.8GHz would be a good Athlon 64 2800+ competitor.

Quote:
slow FSB speed. maybe Intel can correct this.

Slow FSB could easily be scaled UP. Dothan slow FSB is part of the mobile market low power requirements. Lower FSB use less power, dissipate less heat and permits ynch mode usage with low power DDR200/266/333. And considering the GOOD performance of Dothan with slow FSB, imagine them with a 533/800FSB, they would probably gain a few percentage in most benchmarks.

On the other hand the 2Meg of cache compensate (to some extent) for this slow FSB.

Quote:
Intel can get away with low yield on laptops, but producing Dothan in Intel's classic desktop and server volume is a bit of a pinch. Having a whopping 2MB of cache isn't exactly going to help yields either

Intel might even consider deactivate 1Megs of cache on them to get better yield and still get decent performance from this core paired with a higher FSB.

Overall, the Dothan will probably save Intel in the mid-term. Surely, Intel will not push the Dothan to the desktop as fast as we would like. There is a simple reason to this. Intel Netburst architecture at 90nm must be profitable, Intel can't kick out the Prescott, they would loss a lot because of the R&D money invested in that architecture. And the Prescott experience will serve Intel in their move to Dothan @ 90nm. Intel are currently polishing their 90nm process. AMD haven't yet officially shows fully working A64 @ 90nm process. In a technological point of view Intel is currently the leader. AMD have the performance crown, but Intel might regain it sooner than we think if they can get working Dothan @ 90nm running at 2.4GHz and more.

--
What's the <b><font color=green>AMD Mobile Athlon 64</font color=green></b> overclocking potential? <b>It's huge!</b> Humm... Maybe not that huge...
Related resources
June 4, 2004 4:23:36 PM

Lately I don´t understand what GamePC is doing with their reviews... and this Dothan review is a confirmation. I found meaningless things. For example, look at this (UT2004 Botmatch Rankin, in GamePC P4EE 3.4 review):

<A HREF="http://www.gamepc.com/labs/view_content.asp?id=p434ee&p..." target="_new">http://www.gamepc.com/labs/view_content.asp?id=p434ee&p...;/A>

A64 2800+ outperforms P4C 3.2 by a 8%.

OK, in this Dothan review, A64 2800+ seems anemic in this bench. The same P4C 3.2 smokes A64 2800+ by a 15%. Well, all we know how strong all A64 are in UT2004.

The UT2004 P4EE review, seems correct. Obviously, this UT2004 Dothan review, doesn´t.

What is happening here?


With Halo bench, in GamePC P4EE review A64 2800+ is 8-9% slower than P4C 3.2:

<A HREF="http://www.gamepc.com/labs/view_content.asp?id=p434ee&p..." target="_new">http://www.gamepc.com/labs/view_content.asp?id=p434ee&p...;/A>

And now, in Dothan review, nobody knows how is 13-14% slower...

What is happening here?


AND, another incredible and rare (and now I say biased situation): Photoshop filter test. Look at this (GamePC P4EE 3.4 and A64 2800+ reviews):

<A HREF="http://www.gamepc.com/labs/view_content.asp?id=p434ee&p..." target="_new">http://www.gamepc.com/labs/view_content.asp?id=p434ee&p...;/A>

-P4C 3.2 (Dual channel): 169.8 seconds
-A64 2800+ : 208 seconds.

<A HREF="http://www.gamepc.com/labs/view_content.asp?id=a642800&..." target="_new">http://www.gamepc.com/labs/view_content.asp?id=a642800&...;/A>

-A64 2800+ : 206 seconds (practically identical result than above).

We can see, there is a correct situation with both A64 results.

NOW, What happening in Photoshop filter test in Dothan review???? This:

-P4C 3.2 (single channel): 174.3 seconds (2-3% slower than with dual channel, so, an understandable situation).
-A64 2800+ : 241.0!!!!!!!!!!!!!!! seconds ??????????? 14% magically slower than in the other two reviews??????????


Of course, thanks to these unsuspected results, Dothan comparing with A64 seems... well... each one can continue the statement for me.


What is happening here?

Sure, Dothan is the way Intel must go, a very very good processor, but...<P ID="edit"><FONT SIZE=-1><EM>Edited by eugeneMC on 06/04/04 06:32 PM.</EM></FONT></P>
June 4, 2004 4:27:17 PM

Since when the cache size reduce the % number of working chip.Redundacy in cache is really easy to use and work fine with any size.

i need to change useur name.
June 4, 2004 5:24:48 PM

Humm... OK!

It's probably because GamePC are running A64 with low speed memory in their Dothan comparison. Usually, laptop are running with DDR266, they pobably crippled the A64 with DDR266 memory to have a "better" comparison to Dothan 400MHz FSB. But it not seems too, based on their TestBed specs... It's hard to understand.

I don't know much about GamePC, but your analysis seems to point that this site is not probably not the most professional review site around!

--
What's the <b><font color=green>AMD Mobile Athlon 64</font color=green></b> overclocking potential? <b>It's huge!</b> Humm... Maybe not that huge...
June 4, 2004 7:09:38 PM

Quote:
<i>Originally posted by: <b>TheRod</b></i>
I partially agree with you. We have seen overclocked DOTHAN @ 2.4GHz in some reviews.

We've also seen overclocked Athlon64s @ 2.8GHz--and that's not even taking 90nm into account.

Quote:
Slow FSB could easily be scaled UP. Dothan slow FSB is part of the mobile market low power requirements. Lower FSB use less power, dissipate less heat and permits ynch mode usage with low power DDR200/266/333. And considering the GOOD performance of Dothan with slow FSB, imagine them with a 533/800FSB, they would probably gain a few percentage in most benchmarks.

Probably so. Of course, so would the Athlon64, given the second channel of Socket939. I guess using the Socket754 part works as an equalizer here...

Quote:
Intel might even consider deactivate 1Megs of cache on them to get better yield and still get decent performance from this core paired with a higher FSB.

True. However, I suspect the 2MB cache is primarily responsible for Dothan doing so well, especially on gaming benchmarks. We've all seen how P4EE benefits from it.

Quote:
AMD haven't yet officially shows fully working A64 @ 90nm process. In a technological point of view Intel is currently the leader. AMD have the performance crown, but Intel might regain it sooner than we think if they can get working Dothan @ 90nm running at 2.4GHz and more.

Actually, I recall AMD showing some running systems with prototype 90nm Opterons a few months ago. In any case, it seems like 90nm Athlon64s and Dothans will hit the desktop at approximately the same time.

FWIW Intel is technically "ahead" on 90nm, although Prescott is obviously more of a forward stumble than a forward leap.

Quote:
<i>Originally posted by: <b>juin</b></i>
Since when the cache size reduce the % number of working chip.Redundacy in cache is really easy to use and work fine with any size.

AMD does cache redundancy well; Intel does not. That's why Celerons and M0-stepping P4's get pulled from standard P4 and P4EE lines; they're typically the parts that failed on the cache testing.

<i>Look! Up in the sky! It's a bird! It's a plane! It's...

...an asthmatic werehamster?

<LHGPooBaa> Well, @#!& on me.</i>
June 4, 2004 9:19:06 PM

>1) low clock speed. the P6 core in Dothan always seems to
>tap out on clockspeed quite a ways behind AMD's parts at
>the same trace size.130nm only goes to 1.3GHz (vs AMD's 2.
>4GHz)

Not quite; first Banias scaled to 1.7 GHz, using extremely low voltages and TDP's, so using a process tweaked for low power, not high clock. I have a hard time believing Banias could not have been produced at speeds above 2 GHz using a process tweaked more for speed instead of low power, and with a much more liberal vcore and thermal enveloppe.

>90nm is only going to 2GHz so far.

Some french nutheads overclocked a Dothan to 2.4 GHz in a laptop, using stock voltage (so much lower than any A64), and they where most likely limited by the RAM (SODIMM laptop dimms dont come in DDR600 flavours, and their DDR400 module was running at 480 MHz, in synch with the FSB). I wouldnt bet any money on Banias being much more limited in clock than K8.

>2) slow FSB speed. maybe Intel can correct this.

of course they can. there is no reason I can think off dothan would not work on a dog standard P4 board using a 800 MHz bus. Its the very same fsb as the P4. The only issue is powerconsumption (for mobile use).

>3) yield, possibly?

I doubt it. If intel can't fab a well known core, that is no bigger than 88mm², they had better closed their doors and start selling tv's :)  The issue with the high clocked P3's was not yield, it was bin splits. Don't forget intel didn't use copper interconnects back then, while AMD had done so since their 180nm process (afair). If Athlon (K7) outscaled P3, it was probably more due to that than anything else.

> Having a whopping 2MB of cache isn't exactly going to help
>yields either-

I'm not convinced yields would be worse than for any other high volume chip intel fabs. Remember Dothan is *SMALL*, at 88mm² its almost two and a half times smaller than willy.. if binsplits are okay (and again, with a bigger thermal enveloppe, they should), intel should be able to produce boatloads of those chips.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
!