It's good to see that SPCR is following the P-M development.
The HSF is a pitty again. Why did they have to stick with a 40mm HS? A 60mm one would be much better. The mainboard is expensive enough...
Nevertheless, the P-M is specified to 100°C cover-temperature. Running the CPU with this high temps should not harm the chip.

One typo in specifications comparism table: both GbE ports of the new board are cocnected to the the chipset via PCIe, as stated in the specifications.

Unfortunately, two questions remain:
Why haven't you done any power consumption testing? What about EIST?

Power consumption is identical to the 855GMe boards. 50-ish W under full CPUBurn load. Idles at 38-ish W.

The various "Speed Step-ish" apps work fine, but aren't any more stable than when used on the 855 boards.

My general opinion of these dynamic speed/Vcore adjustments on Pentium M desktop boards is: why? The idle power consumption is already extremely low, and dropping the thing down to 6 x 100 @ .7V reduces it another 2-3W. Big deal. It also makes no difference in the idle cooling requirements and the instant it ramps up to full speed, the cooling requirements are the same as when these "Speed-Step" sort of apps are not present. Useful for notebooks, sure. For desktops, I just don't see it, especially since all the existing apps are a bit flaky.

Just so you know,

On the aopen support forum, a thread dated 6/14 tells us that (I'm paraphrasing 'cause I don't have time to link):
____

An Aopen Tech informed a concerned user that they now have a new updated heatsink available for the 915gm. They were then told to call customer support regarding this.

People have followed up and called customer support regarding this but were told they had no knowledge of any new or updated cooler for the 915gm, but they could check back next week...
____

UPDATED: Paraphrased no more, here is the link to the thread:
http://club.aopen.com.tw/forum/viewmess ... geID=76336

so far no one has gotten any more info.


Speedstep:
Ralf, regarding speedstep: the wattage savings when going from 533fsb dothan (27tdp) 2ghz @ 1.35 down to 1.1 or lower is more than 2-3 watts, it's closer to 10watts. That could could mean the difference between fanless or fanned cooling in various smaller form factor htpcs. I agree that if you are already underclocking the wattage savings dissappears but if people want to run at stock speeds undervolting still is a big help.

One thing that impinges on this is the efficiency of the PSU. Virtually every PSU I have tested shows rapidly decreasing efficiency as power draw drops, especially below 60-70W DC output. It is quite a steep curve that gets steeper as you keep going lower. (It might be better for power bricks that are rated for <100W; don't know, have not tested them)

This means that you will not be gaining a linear improvement in overall power draw as the CPU power demand is dropped. So if the CPU power drops from 27W down to 20W, becuase the PSU power efficiency is also dropping (from say 60% at 50W to 50% at 40W), you might only see a 5W decrease in AC power draw. So the overall heat in the system is not much changed.

Also, as RH already said, because SpeedStep applies only when the processor is not being worked, the temp improvement is not helpful. It's at load that the CPU cooling is an issue, not at idle. With the stock HS for this board, fanless operation is absolutely out of the question, judging from RH's results, if you anticipate pushing the CPU to full tilt even only very occasionally.

For these reasons, I agree with RH that Speedstep really is a non-factor for this board with any known P-M CPU.

i am so sad about the heatsink mounting holes..
this board was looking almost perfect for what i wanted next, but to have to fight it all the way to get a real heatsink on there is just a pita. (why do they do it?! such a waste. ARGH! </rant>)

like the reviewer, i'm also hesitant about only one IDE channel and lack of ps/2. i like to copy cds 'on the fly' which means need two cd drives, probably on pata cables, but i would also want the quietest 2.5" drive in this system, which afaik is unavailable in sata. am i missing some 'modern' way out of this catch, or should i just wait for a newer mobo or sata 2.5" hd?

i still use ps/2 mouse AND keyboard and like it that way (was never able to get suspend to ram working properly when they're plugged to usb?!). hopefully i'd have better luck with S3 wake from usb going properly on a newer board, but is there pins on the board for ps/2 mouse incase? i guess not right, otherwise they'd probably include the plug on the same piece of metal as the keyboard... still kinda strange that they bother to give you one but not the other

can you tell a bit more about the onboard audio, did you try it out? anything good or bad to say? what type of plugs are those 6 i see, they look a bit like miniplug which i don't like much - can you get digital audio out through one rca cable instead (eg a spdif panel included)? what about optical out (toslink)? and what is the 7.1 all about, never seen it or heard of anyone who has 7.1 speakers setup, will the user who is already happy with his 5.1 speakers have any woes trying to connect this up and get everything working just as it should with a 5.1 sound card?

i saw some things you mentioned but it's hard to get the right idea..so did you do any (quantitative) noise measurement on the fan? if not can you give a guess, it's about as loud as an L1A at x volts.. actually i'm pretty sure i'd try to get rid of it no matter what, but it would be useful info for the community i think

SATA notebook drives are already out. See our new Fujitsu 80G SATA notebook drive review here: http://www.silentpcreview.com/article250-page1.html Others will be coming very fast. Notebook makers are already offering models with SATA drives and I am sure they are not all Fujitsu -- it was just the first one offered to retail.

FYI, I've requested AOpen for a sample of that fan to be sent to me so that it can be measured / recorded and added to the review. RH in LA is too far away for him to make a casual jaunt up to me / my auidio lab in Vancouver BC.

Interesting review. I hope you guys stay on the ball, if AOpen decides to release a better HSF - which they obviously SHOULD DO - I'd love to see the review updated.

The interesting thing is, before I went to Computex, I saw a picture of the ATX version of this board ( http://usa.aopen.com/products/mb/i915Ga-HFS.htm ) with a socket 478 HS mounting frame, not the tiny HSF on the i915GMm-HS. But at the show, the samples shown had the small HSF. Now, I see there is no picture of the i915Ga-HFS on the AOpen site.

This may mean that they are retooling the ATX version of this board to put a 478 socket frame on it... and hopefully do the same with the mATX version. Even if their engineers think the small HSF is good enough, NONE of the users think it is good enough. In this case, they must consider that the customer is right. (They are not ALWAYS right, but in this case....)

I WAS talking about a 100% cpu usage scenario; I already agree that undervolting when the cpu is underworked is of little practical benefit (UNLESS it lets you get past the fanless operation threshold).

I thought Ralf was talking about software clocking utilities in general. Most of these utilities (RM Clock, Centrino Hardware Control to name two) let you define both the ramping up of cpu speed AND the voltage. So with that software you can seriously undervolt even when the cpu usage is at 100%. The example i quoted is actually my real-world Pentium M laptop undervolt: 2ghz at 1.068 volts is MUCH cooler than the default 2ghz at 1.356.

Moreover you can also manually control the throttling instead of let the software manage it, so conceivably for an HTPC you can set it to some fanless speed when playing back video or web browsing, and set it to run max speed when doing media encoding or gaming. (I personally like manually controlling this stuff because the throttling softweare I've used tends to ramp up even when it doesn't need to).

Regarding PSU heat generation. That is a good point, but now you've expanded the scope to a total system design perspective and from that viewpoint, we are also presented with other alternatives: like you mentioned, perhaps an external power brick could be used, or use a psu whose effciency ratings with such a low power draw remain good. OR take all these issues into account when designing your ultimate low-power, silent, living room HTPC.

I guess what I'm saying is: I wouldn't dismiss the utlity of EIST/SpeedStep,RM Clock,etc. It's a tool that is available and it's up to the end user building their own system to determine if it's useful for their setup or not. I don't think you need to do use it in your reviews though, it's the kind of info that forums are good for (since the mileage will vary for everyone it's best if end-users compile the experiences).


[/quote]

Good review, Ralf. I do think "Price" belongs in the Cons. Also, why did you use the standard 21W and 26.8W for the undervolted C/W calcs? The resulting C/W values are misleading at best.

Oh yeah, Good Review!! Told me everything I ever wanted to know about that board and just how difficult it is to find a good heatsink for it...

Great review, especially since it's what I've been seeing since I bought one of these boards.

I found the same problems everyone else did, I think.

Before I even mounted the heatsink, I removed the pink thermal adhesive with some mineral spirits, then spent an hour lapping the bottom of it. Never mind the obvious machining marks (of which there were a LOT) -- the bottom of that sucker was NOT flat. I mounted it to the CPU (2.0GHz P-M at 533FSB) and found idle temps in the mid to upper 50's C, and way hotter than that under load. I eventually wound up buying a Thermaltake Slim Volcano 10 (all copper) sink with a 60x15mm fan, drilled four holes through the bottom to match the mount pattern on the board, and put that on instead. It's a lot better -- I'm now at ~42C at idle and 60C at load -- but it's still not quite right. I may have it mounted unevenly.

The other problem I found is my own damn fault, but one I wonder if others accidentally run into. This is a MicroATX board, which I, for one, wanted to put into an MATX case. I did (a SuperFlower SF101, which I love), but I found out the hard way that, even though the shiny-new Gigabyte 6600GT fit well into the PCIe slot, the 250w MATX power supply that came with the case couldn't supply the juice to the card (how am I supposed to conquor the barbarian horde in Rome: Total War if the machine keeps freezing?) It's hard to find a more powerful PSU in MATX form. I guess my point is that this board lures builders into building a power-hungry machine in a very small box with a too-small power supply.

Also, I'm pretty sure that the SiI controller only works in RAID mode, NOT in single-drive mode. I didn't want RAID, but I did want SATA-II, so I was a little disappointed.

There's a PS2 adapter that fits on a motherboard header and lets you expose a port in the expansion slots. And, the floppy drive header is not standard, but a cable came with my retail version of this board.

FWIW, the final config:
* i915GMm-HFS board
* Pentium-M 2.0GHz @533MHz FSB
* Thermaltake Slim Volcano 10 heatsink (mod)
* 1Gb OCZ memory at PC4200, 3-3-3-8
* Gigabyte NX6600GT128D on PCIe 16x
* Two Hitachi 7k80 SATA-II 7,200k drives
* Sony DVD-RW drive (forgot the model number)
* Hauppage PVR150 TV Tuner card
* TTGI/SuperFlower 250w power supply (soon to be an Enermax EG285SX-VB -- 20A on the 12v bus, vs. 10A)
* Pro-9 card reader
* 2x Panaflo 60x25mm fans, low-rpm version


-j

AFAIK, the new SS-300SFD Active PFC 300W from Seasonic are the most powerful SFX PSUs on the market today: http://www.seasonic.com/product/pc_sfx.jsp They are probably up to the task you're seeking.

I can think of at least one successful instance of the heatsink being replaced: This was the board used in Puget System's fanless, watercooled system. Puget obviously did a custom mounting job to get the waterblock on the CPU.



CPU is under the waterblock on the left.

wim: Another 2.5" SATA drive to consider is Hitachi's Travelstar line. I believe there are a few models in this line on the market.

hi MikeC, yes i did read the sata notebook drive review and i remember this bit: for a 2.5" drive to be louder than any 3.5" is unacceptable imo. good to hear that others will be coming very fast, i'm looking forward to seeing samsung's offerings..

do you think you'll be reviewing this i915Ga? looks pretty good, except max disk is only 144,000,000GB . fingers crossed they get that 478 mounting frame on there. i wonder why the mATX board has dual ethernet but the ATX board only got one? (yeah i'm one of the nutters that actually uses both)

Take a look at Tom's Hardware Guide's review of this motherboard published today. Their results are very contradictory to those discussed here. On this page they state the following:

I find this very interesting. Personally, this is what I expected from Pentium M in the first place. The high temperatures reported in SPCR's article seem unnatural to me, though I'm not saying they're faulty, of course. But I do think this matter needs some further inspection.

The high temps that We got were matched identically by many posters in the mysteriously missing AOpen forums. One of those posts is paraphrased earlier in this thread. AOpen tech also gets these same sort of temps.

As noted in the review, when a decent heatsink is mounted (in this case w/o even being clamped onto the CPU) temps plummet right to the same levels as we've had with the other P-M boards. Here's a quote from page 5 of the review:



So how do you explain the "unnatural" temps with the little AOpen HSF, and the "completely normal temps" when that HSF is replaced with a known quantity? My explanation would be that the stock AOpen heatsink is worthless.

GamePC.com also had a review of this board and found some obscene temps:

http://www.gamepc.com/labs/view_content ... 915&page=1


-j

Not at idle it's not. I'm not pulling those power draw numbers out of my ass. They come from a Seasonic Power Angel, that my test systems are plugged into 24/7/365. I can see a drop of 2-3W under idle when going from default speed/Vcore to 600MHz @ .700V. As the CPU ramps up from idle to any sort of load, that 2-3W "advantage" quickly disappears.

Alright, I first misunderstood some stuff in these articles. And I certainly was surprised to see a CPU with such low TDP to get so hot. I agree, it was cheap of AOpen to include the smallest possible heatsink that they could get away with. After all, the most robust HS available weigh 10x as much as this one.

Ok, just so I understand: the total power system draw of your P-M system at default speed and settings is only 2-3 watts more than when it was undervolted and underclocked?

I believe what you say but the total system power draw wasn't my focus with those comments and I originally misunderstand what you meant.

Now I'm curious however: do you know how your cpu temps varied during such a test (600mhz @ .7 volts versus default speed/voltage) or was it similarly unimpressive?

Nice review, thanks!

I was rather surprised to see that the HSF does such a poor job: one would think Aopen could at least have spend a bit of R&D time checking whether they were going to have CPUs baking? Surely aluminium and/or copper can't have crept to so high a price?

I am sure the choice of HSF has nothing whatsoever to do with price. You have to remember that AOpen is a BIG company with many divisions, and intracommunications in big companies is often suspect.

Their first desktop P-M board was produced in response to demand from the Japanese market where completeness and small size are prized. Hence all the built-in features. I am guessing that with regard to the HSF, the engineering dept made a decision based more on size than cooling or noise. It's very possible that the engineers who worked on this project are not proficient in the English language and may never have had a chance to read the many reviews that praised the choice of HSF and cooling on the original 855 board. There were also comments made by some AOpen folks about the BTX orientation of this board, with the implication that it's because BTX is coming that the layout of this mATX board had to follow a BTX style -- which makes zero sense, I know.

Anyway, since the review was posted, I've been contacted by AOpen about a new, improved HSF for this board, which is being sent to both RH and myself -- RH will test its efficacy and I will measure / record the noise. Here's a pic of the new HSF:

I'd agree with you on the point of many engineers perhaps not speaking English and hence not being as informed as we here might be.

However, that neither alters the fundamental fact that they are, after all, engineers, and one would expect them to produce a feasible solution.

Nor does it on the point of their English: I'm sure that their previous offering must have been reviewed by at least one Japanese review site (almost certainly many more), and that one of them surely brough up how efficient their previous design was!

Overall, I'd think that this board is really good, and my assessment is that we here should not be overly put-off by the HSF. After all, most of us would change it, even were it identical to their good designs.

But, one of the big problems here is that this isn't a standard-issue heatsink mount. I wound up spending time modifying a stock heatsink to make it work (which isn't necessarily a problem, but I'd rather have something engineered, rather than hacked together by me).


-j

Great review. To bad about the HSF...
Could anybody tell me, or make a test if this motherboard, with the built in graphics, could play a WMVHD 1080p movie? I am planning to build a HTPC and I am not really sure what type of processor to buy for it to be able to cover playback of most formats. Would, for example Pentium M 745 be enough?

we can only hope that a few of the premier after-market HS vendors will see this as an opportunity and produce special versions of their HSs to fit this MB.

Two question how do you mount a different heatsink with this board. And second why is it so expensive.
Thanks

Does anyone know what temp the Pentium M runs at inside a laptop? I'm just wondering because on laptops there isn't a lot of space for cooling (although there is certainly heat dissipation through the casing) and apart from the occasional fan whine I don't hear much either.