I'm starting to wonder what was it that made MiniITX stick as a major standard form factor, but every attempt to go smaller from Nano/PicoITX when Via/Centaur was a voice in the wilderness pushing tiny desktop form factors to miniSTX and another recent one whose name escapes me but which was intended as a "computer module" for third party all in one chassis.
At some point it becomes a lot more expensive to make things smaller. I suspect Mini ITX has reached that point for regular, somehwat modular PC components.
At this moment, yes, but the thing that's holding it is the real estate and dubious compatibility. For example, mini-STX has one big advantage over mini-ITX, which is fixed CPU socket location, and this is a must. But I have a feeling, that mini-STX isn't aimed at individuals, but rather large-scale system integrators, that can afford to manufacture their own chassis, or at least have them made to spec. Mini-ITX is the smallest we can do, given size of human fingers, using standard connectors. Even having angled ATX24 and EPS, and offloadong VRMs to daughterboards, and using underside of PCB (which is mostly out of ATX spec, btw) won't remedy this situation. Only way is to shift connectors: remove ATX24 altogether, and leave DC-in 12V jack only, switch to SO-DIMMs only, move SATA to angled connectors only, move PCI-Express to NGFF format (except main GPU port), and we're kinda better, but that requires changes, like limiting max CPU power to fit inside power delivery subsystem abilities (ruling out 160W little monsters like ASRock's X99-ITX/AC), and forcing GPU to get all the juice trough PCIe 6- and 8-pin connectors, which isn't viable now. Also moving all the IO to edge of Mini-ITX board will further decrease compatibility with chassis without much space around.
In the end, I may be able to build E5-2699v4 with GTX1080 in 20x20x20cm cube, but it will be more engineering the solution, rather than system assembly.
Mini-ITX is the place, where you really, really clearly have to define your purpose and use case scenario(s), and then specialize, and this is hard choice for most consumers. It's impossible to build super-powerful PC in this format that will be quiet, and truly small, yet powerful PC will be either very loud, or extremely hot - and really no amount of money is going to change that, until you talk to ODMs like Kontron (who will happily design a motherboard for you).
ITX products have to fill a lot of niches, and in the end they need to be quality engineering applied, yet will sell in low numbers. Ever tried build an ITX PC, and recreate that build half a year later? Forget about it.
There is a less-known standard - thin-mini-ITX. I myself have built such a system for my living-room: fanless, Asus board (H81T) and an undervolted i3-4130T. It mostly has laptop parts: SO-DIMM, Mini PCI-Express slot for wireless, DC power connector. There are just a few products available, it's kind of like a niche of a niche market.
+1 for Thin ITX. My HTPC/torrent server has that same Asus board with a Xeon 1230Lv3 (25w) passively cooled.
There are even more extreme options, such as the Xeon 1220Lv3, which is 13w I think, but also incredibly slow, and at the time I was building a few years ago, unavailable and probably expensive. The Xeon 1230Lv3 cost $200 on eBay a few years ago, which is totally acceptable for what is essentially an underclocked i7, and to date, the most powerful sub-25w CPU you can buy. All the low-voltage Broadwell-based Xeons are 35w or higher :\
One reason others haven't mentioned is that below MiniITX size you're constrained by PSU size and the ATX12V connector. Smaller sizes really need external power bricks and a smaller power connector, but nobody's put in the work to get agreement/consensus/standards for that. Groups are too interested in pushing their own voltage, own connector, etc.
Why? Because there is (sadly) no market for it. Even look at the mATX standard. Most people who buy mATX end up putting it into a mid to full tower case. I have a mATX in a Gen1 Cosmos... it looks downright silly in there, but I already had the case. But that is the issue most builders have... you can go smaller and smaller, but at some point you have to start getting custom small-batch cases, PSUs, etc that just cost too much. ITX is about as small as you can go while still using standard parts.
And then there is a flip side to this: If you have to go smaller than ITX for some reason, then it is probably for a very specific applicaiton which has other thermal and size constraints to contend with. The simple matter is that once you get so small you are not doing standard work, so the standards fall apart. Small batch manufacturing of boards, cases, and PSUs adds up quick (though not as bad as it use to be).
Then there is another big issue; the technology keeps changing too fast. Not to demeane the people who do the work, but it is not difficult to slap a bunch of parts on an ATX board and ship them. There are lots of known quantities (thermals, inter-board interference, etc.) and it makes the work fairly straight forward (my dad does board design... I know it isn't quite that simple). But when you get to small scale parts you start pushing things to limits. Just how close can you put the CPU to the RAM slots without interference? Do you need a GPU or other expansion slot? Need to add SATA/m.2 or just integrate it? So you figure everything out, publish a standard... and a new chip comes out, or a new RAM standard, or more traditionally mobile parts can be used now. It upsets the apple cart and you have to start over from the ground up. Meanwhile ATX is 20+ years old and just keeps on working. Unless you have a very specific application, there is just no payback time for ultra small form factors. It is a lot of work, and short lived. No time to get an ecosystem built up around it.
That's an opinion... According to a Google study only 3.2% of all DIMMs will experience at least one correctable or uncorrectable bit of data per year of constant use.
For servers ECC is important but for most workstations ECC is overkill.
You're forgetting the No True Scotsman's Workstation fallacy. A Real Workstation (tm) is used for mission critical work where a 12.2% or 22.9% (4 or 8 dimms) chance of an error/year is totally unacceptable. :eyeroll:
And the OS will respond to the single corrupted bit by immediately performing a hard shut-down, thus protecting you by ensuring that way more than one bit gets corrupted.
ECC means that the single bit error will be corrected transparently to the rest of the system. Only much rarer double bit errors can bring the system down.
And while a system shutdown means a loss of work since the last save, it's also a problem that announces itself so that it can be immediately corrected vs the corruption going initially undetected and tainting every result that builds off of it until the rot spreads to the point of producing something obviously wrong.
Immediate shut-down, which is not what should happen with ECC errors, won't cause data corruption if one is using a good journaling file-system and has hardware caching configured properly. It may cause data loss back to the last committed transaction, but it won't leave data in a unknown state.
"Random" memory errors are rare. Usually errors indicate a subtle hardware fault of some kind...and even simple parity is much better than nothing, since you can take corrective action.
HP equips some of their USFF's with i3's instead of Xeon's to add ECC support. For light applications, an i3 is more than sufficient, and in business I'd take an i3 with ECC over an i5 without it.
That isn't to say you can't get a Z workstation from HP with an i5/i7 without ECC, but they are pretty rare. Most of the high end HP i5/i7's (before Xeon becomes the defacto configuration) are the Elitedesk 800's, and they don't dare call those workstations like Lenovo does with their P (and even M) series.
No wonder Lenovo is losing the sales crown. People are finally realizing that quality doesn't come cheap. And Lenovo is cheap, especially compares to HP. But IT departments have spoken and enterprise is moving back to HP and Dell.
You should fact check before you troll. First, I would like to see you get a Z workstation from HP with ANY 7th gen Core i processor AND ECC memory. The current gen of Core i does not support ECC memory. That's the only difference between Core i and Xeon processors, and now that Intel has mobile Xeon, they will likely close that loophole that allows certain SKUs to support ECC.
Second, HP likely sells almost as many workstation-class machines with i5/i7 as they do Xeon E3/E5. Based on volumes in the WS market that entry space does =/> volume as the higher-end systems. And last time I checked, they're still called HP Z2 and Z240 WORKSTATIONS. Oh, and try to buy a Z2 with a Xeon AND discrete graphics. Please. Try.
Third, when has Lenovo referred to an M Series as a workstation? Please do tell.
Finally, Lenovo isn't losing the sales crown. If HP manages to seize it for a quarter, it is NOT in the corporate space, it's in consumer, where there generally aren't IT departments "speaking," they're buying based on... cheap.
Entry workstation. It is an ENTRY workstation because it is certified to run specific (primarily single-threaded) workstation applications. ECC memory, while recommended, is not mandatory, nor is it even the majority share of memory in the entry space.
It depends on what you mean by workstation. If by workstation you mean a station where you do work (office, school, retail, etc) then this is a perfectly applicable device. If by workstation you mean a work-horse server being used as an end-user device, then no, this is not for you. It is a glorified laptop meant to stay in one place. You aren't going to be running solid works or other high compute programs on this thing.... they couldn't do it even if you wanted them to. But such workstations are not going to be so tiny either.
A coworker has the Dell iteration of this and it's dead silent. It's the size of a paperback book with a real quad-core i7. What more could you ask for?
So, I am no fan of Lenovo. Outside of a few high-end laptops I find most of their products to be disappointing. That said, the school district I work for bought the previous gen of these mini-desktops last year and they worked out rather well. They are cute, small, quiet, and ours (i5 with SSDs) were nice and quick. Very happy and would purchase again. But the best part is that they don't seem to compromise on port selection like Intel NUC devices do. Where our NUC devices have 3-4 USB ports and 2 video outs, these have 6-8 USB, 3 video out options, audio jack, ethernet, wifi (w/ antenna), etc. I would still use a NUC for my wife's next PC. But in a school or business where a device is more likely to be thrown in different environments and use cases over it's life as programs change I would absolutely pick something like these again to keep options more open.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
26 Comments
Back to Article
DanNeely - Tuesday, June 20, 2017 - link
I'm starting to wonder what was it that made MiniITX stick as a major standard form factor, but every attempt to go smaller from Nano/PicoITX when Via/Centaur was a voice in the wilderness pushing tiny desktop form factors to miniSTX and another recent one whose name escapes me but which was intended as a "computer module" for third party all in one chassis.DanNeely - Tuesday, June 20, 2017 - link
... today have failed to gain any significant traction.MrSpadge - Tuesday, June 20, 2017 - link
At some point it becomes a lot more expensive to make things smaller. I suspect Mini ITX has reached that point for regular, somehwat modular PC components.Vatharian - Tuesday, June 20, 2017 - link
At this moment, yes, but the thing that's holding it is the real estate and dubious compatibility. For example, mini-STX has one big advantage over mini-ITX, which is fixed CPU socket location, and this is a must. But I have a feeling, that mini-STX isn't aimed at individuals, but rather large-scale system integrators, that can afford to manufacture their own chassis, or at least have them made to spec. Mini-ITX is the smallest we can do, given size of human fingers, using standard connectors. Even having angled ATX24 and EPS, and offloadong VRMs to daughterboards, and using underside of PCB (which is mostly out of ATX spec, btw) won't remedy this situation. Only way is to shift connectors: remove ATX24 altogether, and leave DC-in 12V jack only, switch to SO-DIMMs only, move SATA to angled connectors only, move PCI-Express to NGFF format (except main GPU port), and we're kinda better, but that requires changes, like limiting max CPU power to fit inside power delivery subsystem abilities (ruling out 160W little monsters like ASRock's X99-ITX/AC), and forcing GPU to get all the juice trough PCIe 6- and 8-pin connectors, which isn't viable now. Also moving all the IO to edge of Mini-ITX board will further decrease compatibility with chassis without much space around.In the end, I may be able to build E5-2699v4 with GTX1080 in 20x20x20cm cube, but it will be more engineering the solution, rather than system assembly.
Mini-ITX is the place, where you really, really clearly have to define your purpose and use case scenario(s), and then specialize, and this is hard choice for most consumers. It's impossible to build super-powerful PC in this format that will be quiet, and truly small, yet powerful PC will be either very loud, or extremely hot - and really no amount of money is going to change that, until you talk to ODMs like Kontron (who will happily design a motherboard for you).
ITX products have to fill a lot of niches, and in the end they need to be quality engineering applied, yet will sell in low numbers. Ever tried build an ITX PC, and recreate that build half a year later? Forget about it.
nirolf - Tuesday, June 20, 2017 - link
There is a less-known standard - thin-mini-ITX. I myself have built such a system for my living-room: fanless, Asus board (H81T) and an undervolted i3-4130T. It mostly has laptop parts: SO-DIMM, Mini PCI-Express slot for wireless, DC power connector. There are just a few products available, it's kind of like a niche of a niche market.nirolf - Tuesday, June 20, 2017 - link
An article about it: http://www.anandtech.com/show/4394/more-on-intels-...DanNeely - Tuesday, June 20, 2017 - link
Thanks. That was the standard I was blanking on.Samus - Tuesday, June 20, 2017 - link
+1 for Thin ITX. My HTPC/torrent server has that same Asus board with a Xeon 1230Lv3 (25w) passively cooled.There are even more extreme options, such as the Xeon 1220Lv3, which is 13w I think, but also incredibly slow, and at the time I was building a few years ago, unavailable and probably expensive. The Xeon 1230Lv3 cost $200 on eBay a few years ago, which is totally acceptable for what is essentially an underclocked i7, and to date, the most powerful sub-25w CPU you can buy. All the low-voltage Broadwell-based Xeons are 35w or higher :\
bug77 - Tuesday, June 20, 2017 - link
My guess is customizability is the reason. Sub miniITX cases simply do not allow much.jensend - Friday, June 23, 2017 - link
One reason others haven't mentioned is that below MiniITX size you're constrained by PSU size and the ATX12V connector. Smaller sizes really need external power bricks and a smaller power connector, but nobody's put in the work to get agreement/consensus/standards for that. Groups are too interested in pushing their own voltage, own connector, etc.CaedenV - Saturday, June 24, 2017 - link
Why? Because there is (sadly) no market for it.Even look at the mATX standard. Most people who buy mATX end up putting it into a mid to full tower case. I have a mATX in a Gen1 Cosmos... it looks downright silly in there, but I already had the case. But that is the issue most builders have... you can go smaller and smaller, but at some point you have to start getting custom small-batch cases, PSUs, etc that just cost too much. ITX is about as small as you can go while still using standard parts.
And then there is a flip side to this: If you have to go smaller than ITX for some reason, then it is probably for a very specific applicaiton which has other thermal and size constraints to contend with. The simple matter is that once you get so small you are not doing standard work, so the standards fall apart. Small batch manufacturing of boards, cases, and PSUs adds up quick (though not as bad as it use to be).
Then there is another big issue; the technology keeps changing too fast. Not to demeane the people who do the work, but it is not difficult to slap a bunch of parts on an ATX board and ship them. There are lots of known quantities (thermals, inter-board interference, etc.) and it makes the work fairly straight forward (my dad does board design... I know it isn't quite that simple). But when you get to small scale parts you start pushing things to limits. Just how close can you put the CPU to the RAM slots without interference? Do you need a GPU or other expansion slot? Need to add SATA/m.2 or just integrate it? So you figure everything out, publish a standard... and a new chip comes out, or a new RAM standard, or more traditionally mobile parts can be used now. It upsets the apple cart and you have to start over from the ground up. Meanwhile ATX is 20+ years old and just keeps on working. Unless you have a very specific application, there is just no payback time for ultra small form factors. It is a lot of work, and short lived. No time to get an ecosystem built up around it.
fazalmajid - Tuesday, June 20, 2017 - link
No Xeon means no ECC DRAM. How is this a workstation?Einy0 - Tuesday, June 20, 2017 - link
That's an opinion... According to a Google study only 3.2% of all DIMMs will experience at least one correctable or uncorrectable bit of data per year of constant use.For servers ECC is important but for most workstations ECC is overkill.
DanNeely - Tuesday, June 20, 2017 - link
You're forgetting the No True Scotsman's Workstation fallacy. A Real Workstation (tm) is used for mission critical work where a 12.2% or 22.9% (4 or 8 dimms) chance of an error/year is totally unacceptable. :eyeroll:benzosaurus - Tuesday, June 20, 2017 - link
And the OS will respond to the single corrupted bit by immediately performing a hard shut-down, thus protecting you by ensuring that way more than one bit gets corrupted.DanNeely - Wednesday, June 21, 2017 - link
ECC means that the single bit error will be corrected transparently to the rest of the system. Only much rarer double bit errors can bring the system down.And while a system shutdown means a loss of work since the last save, it's also a problem that announces itself so that it can be immediately corrected vs the corruption going initially undetected and tainting every result that builds off of it until the rot spreads to the point of producing something obviously wrong.
easp - Wednesday, June 21, 2017 - link
Immediate shut-down, which is not what should happen with ECC errors, won't cause data corruption if one is using a good journaling file-system and has hardware caching configured properly. It may cause data loss back to the last committed transaction, but it won't leave data in a unknown state.cbm80 - Wednesday, June 21, 2017 - link
"Random" memory errors are rare. Usually errors indicate a subtle hardware fault of some kind...and even simple parity is much better than nothing, since you can take corrective action.Samus - Wednesday, June 21, 2017 - link
HP equips some of their USFF's with i3's instead of Xeon's to add ECC support. For light applications, an i3 is more than sufficient, and in business I'd take an i3 with ECC over an i5 without it.That isn't to say you can't get a Z workstation from HP with an i5/i7 without ECC, but they are pretty rare. Most of the high end HP i5/i7's (before Xeon becomes the defacto configuration) are the Elitedesk 800's, and they don't dare call those workstations like Lenovo does with their P (and even M) series.
No wonder Lenovo is losing the sales crown. People are finally realizing that quality doesn't come cheap. And Lenovo is cheap, especially compares to HP. But IT departments have spoken and enterprise is moving back to HP and Dell.
HeyEric - Wednesday, June 21, 2017 - link
You should fact check before you troll.First, I would like to see you get a Z workstation from HP with ANY 7th gen Core i processor AND ECC memory. The current gen of Core i does not support ECC memory. That's the only difference between Core i and Xeon processors, and now that Intel has mobile Xeon, they will likely close that loophole that allows certain SKUs to support ECC.
Second, HP likely sells almost as many workstation-class machines with i5/i7 as they do Xeon E3/E5. Based on volumes in the WS market that entry space does =/> volume as the higher-end systems. And last time I checked, they're still called HP Z2 and Z240 WORKSTATIONS. Oh, and try to buy a Z2 with a Xeon AND discrete graphics. Please. Try.
Third, when has Lenovo referred to an M Series as a workstation? Please do tell.
Finally, Lenovo isn't losing the sales crown. If HP manages to seize it for a quarter, it is NOT in the corporate space, it's in consumer, where there generally aren't IT departments "speaking," they're buying based on... cheap.
HeyEric - Wednesday, June 21, 2017 - link
Entry workstation. It is an ENTRY workstation because it is certified to run specific (primarily single-threaded) workstation applications. ECC memory, while recommended, is not mandatory, nor is it even the majority share of memory in the entry space.CaedenV - Saturday, June 24, 2017 - link
It depends on what you mean by workstation.If by workstation you mean a station where you do work (office, school, retail, etc) then this is a perfectly applicable device.
If by workstation you mean a work-horse server being used as an end-user device, then no, this is not for you. It is a glorified laptop meant to stay in one place. You aren't going to be running solid works or other high compute programs on this thing.... they couldn't do it even if you wanted them to. But such workstations are not going to be so tiny either.
mobutu - Tuesday, June 20, 2017 - link
I like this, nice. Hopefully it stays relatively cool and quiet.Ej24 - Tuesday, June 20, 2017 - link
A coworker has the Dell iteration of this and it's dead silent. It's the size of a paperback book with a real quad-core i7. What more could you ask for?mobutu - Tuesday, June 20, 2017 - link
That's very very good to hear, thanks!Also, these guys updated their article with the info that the price starts at $799
https://www.neowin.net/news/lenovo-announces-the-t...
CaedenV - Saturday, June 24, 2017 - link
So, I am no fan of Lenovo. Outside of a few high-end laptops I find most of their products to be disappointing.That said, the school district I work for bought the previous gen of these mini-desktops last year and they worked out rather well. They are cute, small, quiet, and ours (i5 with SSDs) were nice and quick. Very happy and would purchase again. But the best part is that they don't seem to compromise on port selection like Intel NUC devices do. Where our NUC devices have 3-4 USB ports and 2 video outs, these have 6-8 USB, 3 video out options, audio jack, ethernet, wifi (w/ antenna), etc. I would still use a NUC for my wife's next PC. But in a school or business where a device is more likely to be thrown in different environments and use cases over it's life as programs change I would absolutely pick something like these again to keep options more open.