Intel’s Dothan: The Little Engine That Could?
There’s a topic all abuzz in tech forums right now. I’m not talking about SLI, or the “Revision E” Athlon 64’s, or even dual core chips. It has nothing to do with toaster-oven Prescotts, massive supercooling solutions or the giant power supplies that juice these machines.
No, it’s a little guy, with a cooling solution reminiscent of a VIA C3 and a power draw that’s low even for the mobile arena. Clockspeeds are around those of Socket A Athlons, and though it’s not meant to be a desktop platform, some are claiming it to be THE desktop platform.
And this is coming out of Intel? Whodathunkit?
Meet Dothan, the codename for the current revision of Intel’s Pentium M chip. He doesn’t run much faster than 2GHz, works off a piddly 400MHz single-channel bus (533MHz versions have just been released), and he can still play with big bad P4 EE’s and Athlon FX’s. Did I mention he overclocks well too – even on the kind of aircooler that would look wimpy on a graphics card?
Perhaps I exaggerate a little, but this just sounds too good to be true, and in some ways it is. In fact, if someone wants to send me one to test, I’m all for it!However, Dothan is not the overclocking community’s savior. It will not bring back the days of 50% + overclocks on air, at least not to the general masses. And most importantly, it will not give Intel back the gaming performance crown. (It might in Fugger's system, but there aren't many of his kind around
)
Never mind the “how” of this miracle chip – extremely efficient prefetch and branch-prediction mechanisms, effective use of larger caches, shorter pipeline designed for high IPC (instructions per clock), use of low-leakage transistors on 90nm, the best parts of the P6 and Netburst architectures – ok well, that’s another article . . .
Plus, no decent computer engineer talks about CPU architectures without doing his research, and this one needs to do his research first – right now I might as well be talking out of another, less-often-used orifice . . .
Back to the point, there are at least two considerations preventing a Dothan in every enthusiast’s machine – current problems, which are primarily logistical/physical, and future problems, which are more of the “paradigm shift” kind.
Yay Vaporware! Well, almost . . .
Currently, motherboards for Dothan’s socket (a 479 pin socket, which is not officially called Socket 479, and looks physically similar to S478 but is very different electrically) are hard to find, expensive, and not engineered to the standards of enthusiasts today. Meaning PCI/AGP locks don’t work, voltages don’t go high enough, components don’t take the torture we usually throw at them, ad nauseam.
The only current chipset for the Pentium M is Intel’s own 855 series, and it only supports single channel DDR1 up to 333/PC2700 and AGP 4x - no DC DDR1/2, no Serial ATA, no PCIe, and if someone can rig it with SLI I’d be hella impressed
As for the actual boards, they suffer from numerous limitations, seemingly from the chipset itself (but who’s to say? Talk about untested technology . . .) FSB’s will only get up to ~175, and when you have a downwardly adjustable multiplier and a CPU begging for memory frequency, that can be a bit of a hindrance – even more so if the default FSB is 133 (x4).
Not to mention these boards aren’t cheap, and while the CPU’s don’t take a trip to the Bahamas on your credit card like an EE or FX would, even the low end ones will run you upwards of $200, and with a lower multiplier you have less headroom. Add the fact that you currently pay close to $300 for an arguably less-than-ideal motherboard, and the proposition doesn’t look so enticing. Even the 2.1/400 Dothan has a TDP of just 21 watts, but the current boards only allow mounting the proprietary cooler, and though it does its job admirably, a larger cooler would be more ideal for numerous reasons.
Now some of these problems look to be solved fairly soon. There’s a 855 board on the horizon accepting S478 coolers, 533MHz FSB which gives a hearty performance boost, slowly increasing demand which will reduce the pricing and improve the quality of motherboards, and even Intel’s codename Alviso chipset, which looks to bring dual channel DDR1/2 to the Pentium M. Don’t forget – this sucker is still meant to be a mobile chip, so the timeline for the desktop will lag behind that of mobile solutions, and it’s anyone’s guess as to exactly how much lag will be involved at each turn.
That’s not what keeps me up at night:
Or at least sleeping soundly with an A64 under my pillow
Unless some Intel employees/execs have an aneurism, Dothan and its family will remain on-the-go and out of your gaming rig, at least for practical purposes.
Intel seems to be in no hurry to support new tech like PCIe on the Pentium M, and doesn’t have a reference spec for desktop platforms. Since it’s a mobile chip, Dothan, like the rest of the mobile offerings out of both Intel and AMD for the last few years, could be considered a “handpicked” CPU – it’s hard to say what Intel’s real production capacity, effective yield and profit margin is on these (send me a link if you know more!) and as such it might not be feasible for Intel to slap them in tomorrow’s desktop.
Yes, the marketing department would have to jump some hoops too, but I’d be willing to bet Intel couldn’t make these profitably for the mass-market desktop, even if they wanted to, and signs seem to say they don’t wanna.
Flip the coin, and you still have a CPU that’s designed to run today’s software. Barring 64bit, HT and/or dual core (which I hear is coming), Dothan seems to be good at sequential, scalar-based processing (games, anyone?). That’s not the direction CPU manufacturers want software to go. The trend, and future of consumer processors according to current roadmaps, is parallelism, whether it be thread-level or instruction-level or even OS level (way off topic, but virtualization sounds way cool – multiple independently-running operating systems on the same machine!). Dothan is simply not intended to be a vector/streaming-data processor. And with benchmarks on the slim side, I couldn’t say whether it handles this type of application well regardless – again, links appreciated here . . .
Years from now, the CPU peeps want software to split up, delegate and reorder data execution so multiple things can be done at once – not as an ideal, as the standard. That may be over-zealous, but with frequencies clawing at the wall, programmers may not have much choice. Dothan won’t fit well in this environment, and adding cores somewhat negates the power consumption advantages.
Bottom line?
This is a commentary, an opinion piece, so let the flames fly!
I think most enthusiasts will be happy with the mainstream platforms, and let the tinkerers and extremists play with semi-exotic platforms like Dothan. I’m certainly not against Intel throwing out their leaky-faucet PressHots and sending in the mighty Dothan to whip A64 with a wet towel (man that stings . . .). But I don’t see it happening, and I don’t think it really needs to. Things will be quiet in 2005 for the most part, but that’s true for the gaming industry as well, and for a similar reason – the calm before the storm. Video games are awaiting the new platforms, and CPUs are awaiting the Gemini revolution. Soon your shady XXX store will be advocating two-headed everything, regardless if that brings you jollies personally. (Please, pretty please DON’T tell me what does, I’m NOT interested . . .)
There are a number of reasons why Dothan is intriguing, but I’m interested to see what you think – so here’s my little devious plan. I’m going to conduct an informal poll. Send me a PM with the title “Dothan Article” and tell me what it is that excites and interests you about this CPU. In the meantime, I’m going to do my research like a good little boy, and find out what makes Dothan tick, tick, tick . . . (well that’d be the PLL clock generator you idiot) – now I’m just talking to myself. Here’s your friendly neighborhood computer engineer signing off . . .
And yes, I wish I was Spiderman. Kirsten makes a hot readhead. Then again, who doesn’t?
-- Matt




However, Dothan is not the overclocking community’s savior. It will not bring back the days of 50% + overclocks on air, at least not to the general masses. And most importantly, it will not give Intel back the gaming performance crown. (It might in
)



Reply With Quote
Any day now . . .



Bookmarks