This is a post from TabletPC Reviews forum that I want to share here because it says what I have been saying here since the day I saw for the first time the benchmarks of a Q1 Ultra. There are rumors about Samsung dropping the production of the Q1 Ultra at 600 MHz (A100) due to the really poor performance.
I was somewhat cautiously optimistic about the Intel's 2007 UMPC platform known as McCaslin, along with it's 90nm Stealy processor (officially known as A110 and A100), but apparently, it's gonna be replaced early next year with Menlow UMPC platform as all the recent news stated.
I say cautiously optimistic because up until now, Intel powered UMPCs had to make due with low voltage or ultra low voltage versions of the mobile CPUs designed for much bigger devices and their bigger batteries (namely portable or ultraportable laptops). This resulted in either not getting enough power for longer battery life, short battery life with acceptable power, or both short battery life and bad performance because of the size constraints of the UMPC platform. But with the Stealy CPU in the McCaslin platform, Intel was seemingly about to address the UMPC problem with a line of dedicated processors for the job (with it's own designation apart from the regular Pentium brand name).
But "cautiousness" of my optimism went to full blown caution when I later read that Stealy is essentially a Dothan Pentium M with much lower clock and tiny L2 cache of 512KB. Basically, The power savings comes from much smaller die size due to huge reduction in L2 cache size, reduction to 90nm from 110nm, and capping of the clock speed to a PDA like 600 to 800 MHz. Having read this fact, I questioned whether A110 could even out perform the 900MHz Celeron M in the Samsung Q1.
Now, the evidence seems to bear out my suspicions: Intel has moved up Menlow from late 2008 to early 2008. Why would they do this if McCaslin was up to the task?
I think the real reason for McCaslin was to stop momentum of UMPC OEMs switching to Via and it's C7-M until Menlow became available, and Menlow, not McCaslin is the first true dedicated UMPC processor. Let's looks at the reasons:
* LV and ULV Pentium M and Celeron M are expensive processors that eats up a huge chunk of the cost of a UMPC. They were designed for $2000+ ultra portable laptop market. They cost nearly $100 and up to the OEMs (do not confuse the cost of LV and ULV P-Ms to their regular counterparts. there is a huge gap there). This is big problem for UMPC OEMs trying to keep the cost of UMPCs well below $1000, so many have made deals with Via and it's much cheaper C7-M line.
* Stealy is a quick shortcut to provide a cheaper CPU based on current Intel processor line up. No new design here. Just cut the speed and get rid of much of L2 cache. Quick and easy. Would it provide a good performance at a much lower power? That's not the point. The point is to provide the OEMs a cheaper CPU alternative to fight off Via C-7M.
* Stealy will be replace in only about 9 months time. We have still yet to see Stealy equipped UMPCs in the wild in June of 2007. That's a mighty short run for a new CPU. And it's not using the latest fabrication process, which would be 65nm. Smaller fab process benefits battery challenged devices like UMPCs most of all. Why would they not use it for Stealy? Because that would take more time and cost more money. Not very conducive for a quick stop gap solution. BTW, Menlow and it's Silverthorne CPU will be using the cutting edge 45nm process (Silverthorne also promises high end Dothan Pentium M performance at only 2 watts. Stealy eats up 3 watts to do 800MHz of castrated Dothan performance).
As it stands now, my suspicion is that Stealy will prove to be a very poor choice for those consumers looking for a UMPC solution that addresses a lot of our complaints. I believe A110 equipped UMPCs will perform quite poorly compared to the previous models using the LV Pentium Ms. And most are still quite pricey (A decently equipped Samsung Q1 Ultra with A110 will cost around $1300).
I personally think the best thing to do for most of us will be to sit tight until Menlow platform comes to the market in 1 half of next year and skip the McCaslin/Stealy UMPCs of 2007.
The other day Hugo says that the secret of working with a slow processor is "Wait, you open a program and wait until it's completely open and then you use it or open the next program." But my question is, are not you wasting your precious battery in that unproductive time?
How important is it really to have several applications open quickly at the same time? Isn't the generalized use case that you want to have open 4 or 5 open and be able to switch between them smoothly as you're working? Opening time isn't a big factor in this use case.
ReplyDeleteFor what it's worth, I don't see a major improvement in my user experience between my Samsung Q1p and my OQO. I'm running Vista Business on both and routinely open 4 or 5 applications, generally a couple at a time. These are applications like: Outlook 2007, OneNote 2007, Flex Builder, Flex server, Putty and Internet Explorer 7 running weighty Flex applications. Both the OQO and Q1p meet my needs in terms of loading up the applications and readying my development environment. I see no real difference when switching application contexts once I have them all loaded.
It will be great when Menlow is released as it will spur compeition between AMD, Intel and VIA.
The A100/110 is also positioned against the AMD Geode LX800 and LX900. I have not seen any comparisons with these CPUs.
ReplyDelete