Where are the component US15 systems?

[movax]

Just wondering if anyone has had any luck finding an Atom + US15 chipset board out in the wild, that isn't part of the Fit-PC2 or similar. Looking to build a very low power router, yet the existing Atom boards of course still exceed 30W thanks to the 945.

I found this board, but it's been slipping release date wise since late 2008 or so. Where are these guys?!

 

[movax]

Failed at dredging up some more information on their release dates to mass-market

Seems like the new Ion board idles ~30W or so, but the AMD Geode boards still beat them, with an idle of around 5W. Probably going to start a search for a Geode-board with two PCI slots, but if I must, I'll put a dual-port NIC in one PCI slot.

e: Well, damn, not too badly priced, one open PCI slot, and draws less than 10W. Now just need to find a rackmount for it, or just mount the thin client to one of the internal rails.

 

[__Miguel_]

Have you considered this Intel board?

It's basically a desktop netbook Atom board, with both PCI and mini-PCIe slots. The TDP of the whole system should be way below 30W...

Cheers.

Miguel

 

[RVWinkle]

Have you considered this Intel board?

It's basically a desktop netbook Atom board, with both PCI and mini-PCIe slots. The TDP of the whole system should be way below 30W...

Cheers.

Miguel

As the op stated, current atom boards have high TDP due to the 945 chipset.

Edit: Oops, that one seems to break the mold. I'm not sure if I've seen that one before.

 

[__Miguel_]

As the op stated, current atom boards have high TDP due to the 945 chipset.

Edit: Oops, that one seems to break the mold. I'm not sure if I've seen that one before.

Yes, regular 945GC-based Atom boards are power guzzlers, but that particular board uses the netbook 945GSE chipset, plus the N270 Atom CPU, instead of the usual 230. Both the chipset and the CPU have lower TDPs than normal, AND the CPU even has lower power states than the regular desktop variant, which means idle and average power consumption should drop even more.

There is a reason why you haven't heard of it: the board was launched about a month ago, tops.

In short, you can expect power draws from a system based on this board on the whereabouts of the Eee 2xx series nettop, but you can probably end up with a cheaper system... lol Especially because you don't need a PSU, only a power brick (mini-box.com recomends ~25W 12V bricks/wall plugs, so that should tell you just how little a system can sip with this board).

Cheers.

Miguel

 

[movax]

Woo - very nice. Only complaint is the low-pin count PATA connector - have to order *another* PATA<->CF adapter now, grr.

Not too bad a price either. Just wish it a had a PCIe x1 slot as well, but hey, I'll take PCI as well

e: Oh yeah - no moving parts == awesome. No whiny, crappy little 40mm/60mm fans.

 

[__Miguel_]

Woo - very nice. Only complaint is the low-pin count PATA connector - have to order *another* PATA<->CF adapter now, grr.

Nice finding, movax. I'd swear it was a regular IDE port... But it seems the ICH7M can only provide the 2.5'' PATA connector, not the normal version.

Also, the "mobile" PATA connector has 4 more pins than the standard one... Not very "low pin count"... lol But I get where you're going to... hehehe

Not too bad a price either. Just wish it a had a PCIe x1 slot as well, but hey, I'll take PCI as well

Well, technically, there IS a PCIe slot there. Only caveat is that it's not a normal one, it's mini-PCIe. AND you can fit a normal notebook wireless card (the regular backplate even has three antenna holes for a "Draft-N" card). You can probably build a VERY powerfull router with this board, if you manage to get a Draft-N mini-PCIe wireless card with AP mode support... Or you can fit just about any other mini-PCIe card available (not sure what other options are, though...)

e: Oh yeah - no moving parts == awesome. No whiny, crappy little 40mm/60mm fans.

Actually, I don't know why people stress about the 40mm fan on Intel Atom boards... The only problem I've seen with those is that they're connected to the wrong fan connector. The CPU fan connector hasn't got voltage control, but the SYS fan header has. I put the fan on the SYS header, hard-coded it to 50% and I can barely hear the whole system (including a 4-year old ATX PSU with an 80mm fan) inside the box...

However, it IS a nice thing not having moving parts. I just visited Intel's website, and the total max TDP of CPU, northbridge and southbridge for that board is 11.8W (2.5+6+3.3, and yes, the northbridge TDP is more than twice of the CPU TDP...), which means even with a 25W power brick you still have plenty of power to spare. Wow!

Cheers.

Miguel


P.S.: Did you take a look at the motherboard's manual? It has LOTS of "reserved" headers, including something that looks like an LVDS connector, and several other abnormal-looking (not usual 2x2, 1x3 or 5x2 config) jumper pins... What is that all about?

 

[movax]

The manual seems to make it sound like it supports both 3.5/2.5 connectors; oh well, I got an adapter for it anyways.

Not sure on the headers - lots of legacy ports implemented as headers only, but not sure what the reserved ones are for. I hope the 2A brick I picked from mini-box will be enough (I'm thinking it'll be fine, since I will have a CF drive, not a physical HDD).

 

[__Miguel_]

It would be nice to see just how far one could go with this board... Those "Reserved" jumper blocks are itching me... lol

Nice to see you can use interesting stuff like VFDs without having to route the wires through the back (or hack away with clamps on the internal wires...). Very usefull for a NAS, right?

As for the power consumption, even considering 5W for the memory (not happening, it's probably half of that), another 5W for the HDD (most recent review over at Tom's Hardware on notebook HDDs puts them all under 5W peak power draw, and even 500GB behemoths can peak at under 3W), plus a constant ~12W TDP for the chipset and CPU still leaves you with 2W available.

But of course, you'll be WAAAY under that figure, probably 15~20W will be the absolute max on that system. Especially with a CF card instead of HDD...

Cheers.

Miguel

 

[BecauseScience]

As for the power consumption, even considering 5W for the memory (not happening, it's probably half of that), another 5W for the HDD (most recent review over at Tom's Hardware on notebook HDDs puts them all under 5W peak power draw, and even 500GB behemoths can peak at under 3W), plus a constant ~12W TDP for the chipset and CPU still leaves you with 2W available.

The manual quotes 12.28W for all the following:

2 GB DDR2/533 MHz SO-DIMM
20-inch LCD display through the DVI-D port
USB keyboard and mouse
LAN linked at 1 Gbps
DOS booted via network (PXE); system at idle
All on board peripherals enabled (serial, parallel, audio, ...)

It also gives the power budget for various other things:

SATA DVD-R/W                           0.29 A      3.48 W
3.5-inch SATA hard disk drive          0.525 A     6.24 W
2.5-inch PATA hard disk drive          0.3 A       3.6 W
Intel Z-U130 USB Solid State Drive     0.31 A      3.72 W
Wireless PCI Express Mini Card         0.15 A      1.8 W
PCI cards on riser                     0.19 A      2.28 W
Host-powered USB devices               0.72 A      8.64 W

These numbers are from the test bench. They're not theoretical estimates.

I'm surprised that the D945GSEJT doesn't have a thread of it's own yet. It's the first Intel Atom offering I've been excited about.

GSE + N270 + onboard power + gigabit lan = WIN

My only complaint is the PCI slot. I would have preferred PCIe even at 1x or 4x. Of course you can go with an adapter like the PEMINI2X1 from these guys:

http://www.adexelec.com/pciexp.htm (scroll to the very bottom of the page)

Bus adapters/extenders tend to be low volume, high priced items though. When there's no price on the website...

Oh, and it would be nice to have the option of a version with the 330 as well.

 

[__Miguel_]

The manual quotes 12.28W for all the following:

2 GB DDR2/533 MHz SO-DIMM
20-inch LCD display through the DVI-D port
USB keyboard and mouse
LAN linked at 1 Gbps
DOS booted via network (PXE); system at idle
All on board peripherals enabled (serial, parallel, audio, ...)

Hmmm, very interesting indeed. That's actually not much more power draw than the CPU+chipset TDP... Impressive!

Though, truth be told, you can shave some Watts off that figures just by disabling unused ports. If I remember correctly, a guy using a GA-G33M-DS2R cut his idle power draw in about 5W just by disabling unused legacy ports. And that's nothing to sneeze at...

SATA DVD-R/W                           0.29 A      3.48 W
3.5-inch SATA hard disk drive          0.525 A     6.24 W
2.5-inch PATA hard disk drive          0.3 A       3.6 W
Intel Z-U130 USB Solid State Drive     0.31 A      3.72 W
Wireless PCI Express Mini Card         0.15 A      1.8 W
PCI cards on riser                     0.19 A      2.28 W
Host-powered USB devices               0.72 A      8.64 W

Man, I should have read the manual more closely. My D945GCLF also had some of those figures.

One thing does bothers me... Are those average, idle or load power draws? 6W, por one, seems a tad too much for idle (or too little, if you're using large multi-platter drives), and too little for load figures on a 3.5'' SATA drive... AND if you go the GP way, you'll probably end up with sub-5W idle anyway... lol

I'm surprised that the D945GSEJT doesn't have a thread of it's own yet. It's the first Intel Atom offering I've been excited about.

Mee too. I'd also like to see a D945GSEJT+Z-U130+4965ABGN/5x00 combo for a very atractive price, it would be a MAJOR blast for a home server...

My only complaint is the PCI slot. I would have preferred PCIe even at 1x or 4x.

Well, it really boils down to two things: production cost (my guess is fitting a PCI slot is still cheaper than fitting a PCIe slot, albeit the more complex PCB routing involved) and target market.

These boards are just about spot-on built to be used in NASes, routers, dumb terminals and other such appliances. Which need to be cheap, and PCI is still cheaper to aquire to OEMs...

For instance, you can actually build a 23-drive NAS with this board alone (not including the internal USB headers with boot capabilities and external drives, of course). Using one of these and four of these (extreme solution, of course, but it has been used for similar, though less extreme, approaches, like here.

The point being, you can actually build a VERY respectable NAS with lower costs than you would have if PCIe was to be used.

You can argue a 2-port, PM-aware, PCIe 1x SATA2 controller is actually cheaper than the 4-port, also PM-aware, PCI one, and that would still add up to a very respectable build, but I was actually considering those €300+, 8-port+, dedicated hardware RAID controllers (now that Ciprico hit the dust, there are no more decent low-cost alternatives...), which btw would not matter very much if you were to use WHS or a similar OS (it's CONSUMER hardware we're talking about, after all... lol).

In short, PCI has been around for a while, it is still very cheap to work with, and it can actually satisfy the needs of most people who would actually buy the thing in the first place...

Now, do any of you guys have the faintest idea of how to (easily, for a guy with NO programming skills) implement a WHS-certifiable VFD and SATA drive lights? There are very specific requirements, and it would actually be very fun to have something like this being used for a full-fledged, WHS-certifiable machine...

(And yes, I know that WHS machines can't have video ports. Nothing that two well-placed covers wouldn't handle... hehehe).

Oh, and it would be nice to have the option of a version with the 330 as well.

That's not going to happen. The N270 is a mobile part, same used in netbooks. The 330 is a desktop-only part. You might need to wait until the next generation Atom CPUs for that option to appear.

Shame they didn't go for the N280, though... But I think Intel actually doesn't want that one out in the open, if I remember correctly they have already killed it...

Cheers.

Miguel

 

[BecauseScience]

One thing does bothers me... Are those average, idle or load power draws? 6W, por one, seems a tad too much for idle (or too little, if you're using large multi-platter drives), and too little for load figures on a 3.5'' SATA drive... AND if you go the GP way, you'll probably end up with sub-5W idle anyway... lol

From the manual in section 2.7.2 directly above the table:

Maximum load refers to the incremental power demands placed on the power supply,
augmenting the minimum load configuration into a fully-featured system that stresses
power consumption from all subsystems.

There are notes on most lines stating the nature of the load. On the dvdrom it says "Load: DVD playback". On the 3.5 SATA disk it says "Load: continuous read/write benchmark". I didn't reproduce all of those. The manual is readily available on Intel's website.

For instance, you can actually build a 23-drive NAS with this board alone (not including the internal USB headers with boot capabilities and external drives, of course). Using one of these and four of these (extreme solution, of course, but it has been used for similar, though less extreme, approaches, like here.

Very interesting read!

Well, it really boils down to two things: production cost (my guess is fitting a PCI slot is still cheaper than fitting a PCIe slot, albeit the more complex PCB routing involved) and target market.

...

These boards are just about spot-on built to be used in NASes, routers, dumb terminals and other such appliances. Which need to be cheap, and PCI is still cheaper to aquire to OEMs...

...

In short, PCI has been around for a while, it is still very cheap to work with, and it can actually satisfy the needs of most people who would actually buy the thing in the first place...

...

That's not going to happen. The N270 is a mobile part, same used in netbooks. The 330 is a desktop-only part. You might need to wait until the next generation Atom CPUs for that option to appear.

The choice of PCI and lack of a dual core option are 100% about Intel protecting it's more expensive solutions.

PCIe is cheaper and easier to implement than PCI. A 1x slot is less than one inch long while 32 bit PCI is many times that size. PCIe 1x is only 36 lines whereas 32 bit PCI is 62. That giant card connector and 62 lines take up much more board area than a (superior) PCIe 1x slot. Additional board area costs more. Additional lines cost more. Sometimes if you have too many lines you are forced to add additional layers to the PCB. That adds a ton of extra cost.

There's a similar effect on chip cost. Each signal that's brought out on a pin costs. On-chip transistors are dirt cheap. If you can do something to reduce pin count you do it even it means using up a few 1000 extra transistors. This makes chips that interface to 1x PCIe cheaper (in general) than ones that have to interface to PCI.

When you reduce signal lines you save money. This is one of the reasons everything is moving from big fat parallel busses to serial. There are other reason too of course.

The embedded / industrial guys do want a dual core Atom on a board like the D945GSEJT. Here's a quote from Logic Supply's announcement of the D945GSEJT. (Logic Supply caters to the embedded / industrial market.)

Our customers are waiting (pleading) for a dual core atom paired with a mobile chipset.

http://www.logicsupply.com/blog/2009/03/25/first-glimpse-at-intels-new-mini-itx-platform-johnstown/

Intel could easily do a dual core Atom on Socket 441 if they wanted to. The trouble is they don't want to. I predict that we'll see multicore mobile Atoms right around the same time that multicore ARM chips begin to come into their own.

Intel only wants to sell Atoms into markets where price and / or power efficency makes it absolutely impossible to sell the customer a regular desktop / moble chip. Intel's Atoms are a finger in the dyke, and that dyke is holding back a tidal wave of ARM's.

Shame they didn't go for the N280, though... But I think Intel actually doesn't want that one out in the open, if I remember correctly they have already killed it...

The N280 is alive and kicking in the new ASUS Eee 1000HE. There is little performance difference between it and the N270 though. http://computermonger.com/intel-atom-n280-vs-n270-review.html

 

[__Miguel_]

There are notes on most lines stating the nature of the load. On the dvdrom it says "Load: DVD playback". On the 3.5 SATA disk it says "Load: continuous read/write benchmark". I didn't reproduce all of those. The manual is readily available on Intel's website.

Thanks for the info. Right now I'm studying for my bar exam (well, actually, it's the oral exam, but I wanted to put it in a way non-continental people could easily understand), so I don't really have that much time to begin with... If everything turns out well, next Wednesday night everything will be all right

Very interesting read!

I though you might find it interesting. The bottom line is, short of Vista's impressive transfer speeds over Ethernet (100MBps+ in Vista-to-Vista with adaptative packet sizes), you will NOT be limited by PCI bandwidth on a NAS (or just about any other use this small motherboard can handle)

The choice of PCI and lack of a dual core option are 100% about Intel protecting it's more expensive solutions.

So true... A dual-core mobile Atom would probably kill the sub-€600 Intel notebook offering altogether.

PCIe is cheaper and easier to implement than PCI. A 1x slot is less than one inch long while 32 bit PCI is many times that size. PCIe 1x is only 36 lines whereas 32 bit PCI is 62. That giant card connector and 62 lines take up much more board area than a (superior) PCIe 1x slot. Additional board area costs more. Additional lines cost more. Sometimes if you have too many lines you are forced to add additional layers to the PCB. That adds a ton of extra cost.

There's a similar effect on chip cost. Each signal that's brought out on a pin costs. On-chip transistors are dirt cheap. If you can do something to reduce pin count you do it even it means using up a few 1000 extra transistors. This makes chips that interface to 1x PCIe cheaper (in general) than ones that have to interface to PCI.

When you reduce signal lines you save money. This is one of the reasons everything is moving from big fat parallel busses to serial. There are other reason too of course.

I know those arguments. Really, I do.

However, to this day, and at least here in Portugal, PCIe offerings are very limited. You get the odd sound or TV tuner card, but that's about it. NICs and other stuff are rare. AND they are always much more expensive than the PCI counterparts.

That makes no sense at all, to me. But it's what happens. Weird...

Intel only wants to sell Atoms into markets where price and / or power efficency makes it absolutely impossible to sell the customer a regular desktop / moble chip. Intel's Atoms are a finger in the dyke, and that dyke is holding back a tidal wave of ARM's.

LOL, and, again, so true.

I think the 330 was actually one of those "I can do it" things, not supposed to have much market attractive. But the limitations of the 230 made it that much more attractive, along with the economic downturn.

If Intel finds it in their heart to finally drop the hedious 9xx chipset series for Atom for a Gxx series one, then the extra TDP headroom (not that it it's that necessary to begin with, but oh well...) will probably allow a dual-core version of the chip.

However, Intel has already said the ULV mobile CPUs are being hit hard as it is, so unless Nehalem goes mobile I don't think we'll be seing anything like that. Sadly, of course.

The N280 is alive and kicking in the new ASUS Eee 1000HE. There is little performance difference between it and the N270 though. http://computermonger.com/intel-atom-n280-vs-n270-review.html

Yeah, I know. But that's just about the only option. AND the G40M is nowhere to be seen, too, which would allow lower TDPs and longer battery lives...

Oh, well, my guess is 32nm will bring us very interesting options... Until then, hardly...

Cheers.

Miguel

 

[BecauseScience]

However, to this day, and at least here in Portugal, PCIe offerings are very limited. You get the odd sound or TV tuner card, but that's about it. NICs and other stuff are rare. AND they are always much more expensive than the PCI counterparts.

Wow, I had no idea.

If Intel finds it in their heart to finally drop the hedious 9xx chipset series for Atom for a Gxx series one, then the extra TDP headroom (not that it it's that necessary to begin with, but oh well...) will probably allow a dual-core version of the chip.

However, Intel has already said the ULV mobile CPUs are being hit hard as it is, so unless Nehalem goes mobile I don't think we'll be seing anything like that. Sadly, of course.

...

Yeah, I know. But that's just about the only option. AND the G40M is nowhere to be seen, too, which would allow lower TDPs and longer battery lives...

Oh, well, my guess is 32nm will bring us very interesting options... Until then, hardly

Pine Trail might bring a smile to your face in Q4...?

I'll very close to ordering a D945GSEJT. I'm tired of waiting.

 

[__Miguel_]

Wow, I had no idea.

Yeah, that explains our different opinions on the technology, right?

The only real PCIe advantages are on the 2-port SiI-based PCIe 1x controller, which cost about half of that the 4-port PCI version. Albeit with half the ports, of course...

Gigabit NICs, on the other hand, cost about €15 in PCI version (Realtek-based), and about €50~€60 for PCIe (usually Intel is the only one available). Go figure...

Pine Trail might bring a smile to your face in Q4...?

I hope so.

Though other problems might appear in the meanwhile on my private life... (read: most likely)

Cheers.

Miguel

 

[movax]

On vacation right now in CA, but UPS tracking tells me that the parts are waiting at me for home. Will post build log, perf figures, Kill-A-Watt numbers, etc.

My plan is to make it an ipcop based router, but I'd be willing to test some other things for you guys. Could run some iperf tests, etc.

 

[__Miguel_]

On vacation right now in CA, but UPS tracking tells me that the parts are waiting at me for home.

What?!? Why aren't you at home by now to pick up that thing and begin assembly? Now THAT would be a SWEET vacation for me... lol

OK, joking aside, have a nice vacation.

[Off-topic]Interesting UPS waits for you even if you're on vacation. Registered mail and courier-based packages only wait about 5 days before returning the package to the sender, in Portugal...[/Off-topic]

Will post build log, perf figures, Kill-A-Watt numbers, etc.

Hmmm, I'm drooling already...

Btw, how come just about every one of you guys has one of those Kill-A-Watt thingies? I haven't even seen anything like that here in Portugal... Not that I've searched high and low for them, but they seem to be extremely common in the US...

My plan is to make it an ipcop based router, but I'd be willing to test some other things for you guys. Could run some iperf tests, etc.

Well, since IPCop only really needs a USB pen drive (and small, at that), my guess is maximum power draw will be about 10W, probably less most of the time. Handling routing tables and such is not THAT CPU-intensive, after all...

Don't forget to disable everything IPCop doesn't need (LPT, COM, storage subsystems not used, etc., etc.). You might be able to drop idle power draw by a significant amount, like 10~20% (considering the maximum power draw, of course).

As for tests, iperf seems nice. And, if you have the time and parts, a Windows 7-to-Windows 7 (or Vista) file transfer. No use installing Vista on this board (hence Windows 7 appearing), but it would be nice to know if ~100MBps LAN transfers are doable with this board, and acording to SmallNetBuilder only the Vista kernel is capable of that (everything else will stop at ~80MBps).

Cheers, and congrats.

Miguel

 

[BecauseScience]

On vacation right now in CA, but UPS tracking tells me that the parts are waiting at me for home. Will post build log, perf figures, Kill-A-Watt numbers, etc.

My plan is to make it an ipcop based router, but I'd be willing to test some other things for you guys. Could run some iperf tests, etc.

If you're talking about a D945GSEJT...Sweet! I'm especially interested in kill-a-watt numbers, ftp transfer speeds (Filezilla), and general network throughput under XP. None of that will influence my buying decision though, so no worries...

I'm going to try to use the guts from a P4 M laptop to do the same job. The original laptop idles around 18W. I'll be running it without cdrom, lcd, battery, etc. so it should end up close.

I don't want to spend much getting the laptop board fitted out when I could built around something new for not much more. It it gets too spendy I'll just order a D945GSEJT and be done with it.

 

[ultatryon@ecc]

I just ordered one of these boards as well as the M350 from Mini-box. I wont get it until tuesday however I can't wait to see your performance and power draws.

 

[BecauseScience]

I just ordered one of these boards as well as the M350 from Mini-box. I wont get it until tuesday however I can't wait to see your performance and power draws.

The M350 looks like the ideal choice for a D945GSEJT. It's the ITX case that someone should have been making years ago. Still, it seems a bit overpriced at $40. I really do want one but I don't know if I can bring myself to shell out $40 for what's essentially an 8" square empty box.

 

[__Miguel_]

The M350 looks like the ideal choice for a D945GSEJT. It's the ITX case that someone should have been making years ago. Still, it seems a bit overpriced at $40. I really do want one but I don't know if I can bring myself to shell out $40 for what's essentially an 8" square empty box.

If that was available here in Portugal, it would still be, by at least €15 (prices in Portugal are usually 1:1 in terms of exchange, because we display prices with taxes already added), the cheapest mini-ITX case available. The cheapest one available now is the MI-100, by Apex, starting at ~€56, so the M350 is as cheap as they get.

Also, do keep in mind that smaller cases (and lower volume ones) are actually more difficult to build (machinery and staff need to be more accurate), and also more expensive (low volume). AND that's not a bare case, it has a few interesting twists, like the hideable USB port, and the 2.5'' HDD tray, which add up.

Btw, going back to the original post... Have you guys seen the Fit PC version 2? It's rather pricey (~$400 for the full system, including memory, 2.5'' HDD and Windows XP Home), but it is VESA-sized AND supports FullHD decoding with the Pulsbo chipset...

Cheers.

Miguel

 

[ultatryon@ecc]

So, my machine is together. I disabled the non-needed ports in the bios, and have a Seagate 200gig 2.5" SATA drive hooked up. Using the 5A 12V adapter from mini-box.com, I see a peak draw of 17W on disc spinup, and it settles down to about 12-14W once in Windows XP.

I must say, I am very happy with the system, as its intended to be my "always on" box.

 

[BecauseScience]

So, my machine is together. I disabled the non-needed ports in the bios, and have a Seagate 200gig 2.5" SATA drive hooked up. Using the 5A 12V adapter from mini-box.com, I see a peak draw of 17W on disc spinup, and it settles down to about 12-14W once in Windows XP.

I must say, I am very happy with the system, as its intended to be my "always on" box.

Good to hear that you're happy with your new system!

It struck me just how little power these boards use when I was hooking up an external 3.5 disk the other day. I looked at the specs on the wallwart and was amazed to see that it was 24W at 12V. That's just the right size for a D945GSEJT system. These 24W switching supplies are tiny!

http://www.mini-box.com/24w-12v-2A-AC-DC-Power-Adapter

If I built a D945GSEJT system I'd be using it with UltraVNC (with the special video driver) over gigabit. Is there anyone out there running an Atom with UltraVNC over gigatbit? I'd love to hear how it performs.

I'm currently running a 2.2Ghz P4-M with UltraVNC over 100mb ethernet. Most of the time it's like having a real monitor but things do lag a little here and there.

I sometimes use an E7200 machine with UltraVNC over gigabit and it really is like having an actual monitor. I wonder if the single core Atom in the D945GSEJT is quick enough to deliver that same experience. There needs to be enough horsepower left to run the actual apps too.

 

[movax]

I messed up a bit - I got a 2.5 -> 3.5 adapter, but with the CF adapter I have, the M350 case can't slide shut (also, it came with no instructions, and TBH, I have no idea how you'd actually put a drive in the thing. Power button is flimsy crap as well).

Right now, looking for something to go from female IDE->male IDE (basically an extension cable), or maybe I'll just roll my own (find a dead IDE socket somewhere).

 

[ultatryon@ecc]

The 350 only accepts 2.5" drive natively. The holes in the top plate are spaced for the bottom of a laptop drive. Does your CF adapter have some sort of mounting bracket for 2.5"?

 

[BecauseScience]

I messed up a bit - I got a 2.5 -> 3.5 adapter, but with the CF adapter I have, the M350 case can't slide shut (also, it came with no instructions, and TBH, I have no idea how you'd actually put a drive in the thing. Power button is flimsy crap as well).

Other than the power button what's the build quality like? Is it plastic or metal? What does "TBH" mean?

All of these expensive (for what you get) ITX cases are negating the value of the cheap Atom boards.

ITX is crying out for a cheap, simple case...something that sells for $20 retail. It doesn't need a secret hiding place or a dozen little accessories. It needs to be sturdy and decent looking, an 8"x8" box with a hole in the back for an I/O plate. Someone with manufacturing contacts in China could make a bundle. It'll happen eventually but it's already so long overdue.

Right now, looking for something to go from female IDE->male IDE (basically an extension cable), or maybe I'll just roll my own (find a dead IDE socket somewhere).

http://www.southernhillscomputer.com/ide40idsmato.html

There are a bunch on ebay too. This guy had the lowest price the last time I checked:

http://myworld.ebay.com/dshien/

http://cgi.ebay.com/40-pin-IDE-cabl...3257577QQcmdZViewItemQQptZPCA_Cables_Adapters

 

[__Miguel_]

What does "TBH" mean?

I believe I can be of assistance there. TBH = To Be Honest

All of these expensive (for what you get) ITX cases are negating the value of the cheap Atom boards.

Hell, yeah. I've been saying that for YEARS now.

With the MI-1xx series from Apex, I'm starting to see a decent price trend, but right now the M350 from Mini-Box seems to be the cheapest mini-ITX case available. And $35 before P&P and taxes is still not cheap. Period.

Someone with manufacturing contacts in China could make a bundle. It'll happen eventually but it's already so long overdue.

Unfortunately, and as I've said before more than once, mini-ITX has two drawbacks:

1) Low production volume. Standard ATX cases (not Lian-Li, Silverstone or such stuff, the OEM manufacturers) and, to some extent, uATX ones (manufacturers have discovered you can build a single frame and have 20 different case models... hehehe) are so damn cheap because they are produced several THOUSANDS at a time. mini-ITX cases are almost equivalent to Lian-Li and such brands in terms of manufactured volume - probably only a few of hundreds at a time. And, if my Economics classes were correct, lower production volume will mean higher production costs (economics of scale and such). Which sucks. AND mini-ITX frame designs are rarely replicable, as ATX and uATX are (though the MI-100 series from Apex is a step in the right direction).

2) Size and tolerances. ATX and uATX case manufacture require only semi-decent machinery to be built. They're big, with lots of open spaces, have little to no need for size tolerances when assembling, and the ATX form factor has been used in a tower configuration for more than 10 years now. Mini-ITX cases, on the other hand, are tiny, non-standard, and generally a pain in the a** to fit stuff in. If you can't even fit your hand confortably in one (which is the case with most mini-ITX cases), then you'll either have to cough up the cash to get a more programable/dextrous/precise machine OR make the cases by hand. Any way you put it, you'll end up with more production costs, ergo a more expensive case.

Though, uATX was once such a desert: virtually no choice, and the few cases available were at least twice as expensive as a comparable ATX case. Today you can get uATX cases with OEM PSUs for about €30~40, which is about the same as a comparable ATX case. So, hopefully we'll see mini-ITX cases drop similarly in price as market share increases.

Thank Intel for the mini-ITX push. In the old days you either had VIA or industrial designs in the mini-ITX market. Both horribly expensive.

Now... Anyone has connections with Intel, to see when they'll be launching a sub-€100 Pico or Nano-ITX motherboard? Hehehe

Cheers.

Miguel

 

[mavalpha]

Considering that this thread has already wandered quite far off-topic, I'll say a few things.
1) The Intel board mentioned here was actually not the first one to use the 945GSE chipset, MSI had one out long before. Same specs with similar features, already reviewed back in December.
2) The Atom 230 and N270 are substantially different. The 230/330 use GTL-derived clock generators, which is why they are compatible with the 945 and 9300 chipsets- GTL is also used in the Core architecture. N270 uses a BIOS generator, which is slower and draws less power, but is incompatible with dual-cores. This is also how the 945GSE cuts its consumption.
3) The M350 is a perfect budget ITX case, I'm just putting the finishing touches on my J&W X2 build. It takes a lot of planning and even more squeezing, but it can pack a lot. (BTW, the 780G is the WORST board for this case. The CPU is dead-centre of the board, so the HSF interferes with a 2.5" HDD. I had to use zipties to offset my drive. Same problem with a pico-PSU and the front USB ports, some bending is required to use the ports.) That said, I have it running off a 40W brick with no hard drive at full speed and a USB 802.11g adapter. The 7200rpm laptop drive pushes it over the top, so I have to use an 80W brick instead. Movax, I don't know why you think the power button is flimsy- it feels perfectly solid to me.

If this build withstands the summer heat, I might put together another M350 build with an E8400/9300, and perhaps a Velociraptor for laughs.

 

[BecauseScience]

Considering that this thread has already wandered quite far off-topic, I'll say a few things.
1) The Intel board mentioned here was actually not the first one to use the 945GSE chipset, MSI had one out long before. Same specs with similar features, already reviewed back in December.
2) The Atom 230 and N270 are substantially different. The 230/330 use GTL-derived clock generators, which is why they are compatible with the 945 and 9300 chipsets- GTL is also used in the Core architecture. N270 uses a BIOS generator, which is slower and draws less power, but is incompatible with dual-cores. This is also how the 945GSE cuts its consumption.
3) The M350 is a perfect budget ITX case, I'm just putting the finishing touches on my J&W X2 build. It takes a lot of planning and even more squeezing, but it can pack a lot. (BTW, the 780G is the WORST board for this case. The CPU is dead-centre of the board, so the HSF interferes with a 2.5" HDD. I had to use zipties to offset my drive. Same problem with a pico-PSU and the front USB ports, some bending is required to use the ports.) That said, I have it running off a 40W brick with no hard drive at full speed and a USB 802.11g adapter. The 7200rpm laptop drive pushes it over the top, so I have to use an 80W brick instead. Movax, I don't know why you think the power button is flimsy- it feels perfectly solid to me.

If this build withstands the summer heat, I might put together another M350 build with an E8400/9300, and perhaps a Velociraptor for laughs.

This is why I've ignored the MSI board:

1. Overpriced by a factor of 2. A single core Atom board just isn't worth $200 to me.

2. Requires a conventional PSU. Any sub 100 watt board that doesn't have onboard converters is a fail in my opinion. I'm amazed that Intel hasn't been building their Atom boards like that from the beginning.

 

[mavalpha]

This is why I've ignored the MSI board:

1. Overpriced by a factor of 2. A single core Atom board just isn't worth $200 to me.

2. Requires a conventional PSU. Any sub 100 watt board that doesn't have onboard converters is a fail in my opinion. I'm amazed that Intel hasn't been building their Atom boards like that from the beginning.

The price is actually the same price as when it launched, when it was unique. I don't understand why they won't drop it, but their loss.

I prefer the picoPSU, honestly. It gives me some flexibility, and the ports on the back panel are too important to me- I'm not sacrificing USB, surround sound or a NIC just for a pretty barrel connector that can go anywhere else on the case anyway.

 

[BecauseScience]

I prefer the picoPSU, honestly. It gives me some flexibility, and the ports on the back panel are too important to me- I'm not sacrificing USB, surround sound or a NIC just for a pretty barrel connector that can go anywhere else on the case anyway.

I've never seen an i/o panel so full that there was no space left for a 3.5mm power jack. It could always be an internal connector though.

picoPSU's do their best to emulate a standard ATX PSU. That's convenient in some respects but it misses out the efficiencies that are possible when DC-DC converters are designed to the exact power and voltage requirements of the board. Without that knowledge you can have things like multiple DC-DC conversions and unused voltages. Even when neither of those are a factor, it's still a win to know exactly what voltages and wattages are needed. It allows the designer to fine tune the components to deliver the best possible power efficiency.

You can see the same thing on the wattage vs efficiency graphs of regular PSU's. There's always a wattage sweet spot. Everything far above or below that spot is at a reduced efficiency.

There's really no reason to schlep all these different voltages around in PC's anymore anyway. Point of use DC-DC converters are small, cheap, and efficient. It would really be easier on everyone (and more efficient) if PSU's delivered a single DC voltage to every component. Everything is already doing it's own DC-DC conversion anyway.