Low power Sandy Bridge Mini-ITX build

[Ashex]

Obviously it's your decision, but as I said before, even with the Winmate rather than the PICO, and even with the 5670 (AMD TDP for reference card = 64W) rather than the 9800, you'll have basically zero (or less!) headroom.

Personally, I wouldn't even consider doing this.

Solid point, as I mentioned earlier I'm going to but everything but the PSU and measure the consumption using a spare PSU I've got with me. If peak load is <150W I'll go with the PicoPSU as it can handle a peak of 200W (according to the product page). If it's <130W I'll get the Winmate. Any higher and I'll pick up the Q07/Q08 and go with a full-size PSU.

[ces]

Solid point, as I mentioned earlier I'm going to but everything but the PSU and measure the consumption using a spare PSU I've got with me. If peak load is <150W I'll go with the PicoPSU as it can handle a peak of 200W (according to the product page). If it's <130W I'll get the Winmate.

I think your approach is correct, but check with Electrodaucus on this board about which has the higher peak 12v capacity... PicoPSU or Winmate. I would expect it to be Winmate as opposed to PicoPSU, but Electrodaucus will know for sure.

[ces]

ATI just came out with a 6450 (two versions). If you find that you need to save a few more watts, you might want to look at them.

[Ashex]

I took a look at the 6450 and the performance is pretty much on-par with the Sandy Bridge GPU :/

I've purchased all the parts excluding the PSU. Disappointingly the 2500T isn't available through many retailers so that was purchased separate from everything else, but I should have all the parts in a week or so.

[Vicotnik]

I think your approach is correct, but check with Electrodaucus on this board about which has the higher peak 12v capacity... PicoPSU or Winmate. I would expect it to be Winmate as opposed to PicoPSU, but Electrodaucus will know for sure.

Since the picoPSU only passes 12v through from the brick, it really depends on the brick. With a beefy brick like the 18A Dell DA-2 it's possible (with some cable work) to power a system that pulls more than what the picoPSU is rated for.

[ces]

Since the picoPSU only passes 12v through from the brick, it really depends on the brick. With a beefy brick like the 18A Dell DA-2 it's possible (with some cable work) to power a system that pulls more than what the picoPSU is rated for.

I stand corrected.

How big can you take the picoPSU? I have a q07 case and an MI008 case, and I am struggling with the darn PSU getting in the way of a decent cooler. if I could pump up the wattage of the PicoPSU, I could just pull the PSU.

What is the largest brick available, with a compatible plug? I don't have an appetite for messing around with cables.

[Ashex]

That sounds favorable for me, I dug around and found this post that went into some detail on wiring it to a similar PSU. When I get home tonight I'll take a look at the PicoPSU in my desktop to see how that would work with it.

Without a real pin-pin layout I'm not confident about going this method but I'll see if I can sort that out.

[ces]

That sounds favorable for me, I dug around and found this post

After looking at the post I personally have even less appetite for messing with the cables

How far can you take this without messing with the cables.

[Ashex]

How far can you take this without messing with the cables.

I pulled up the manual for the PicoPSU and it can accept an input up to 12V 18A.

The easiest way would be to do what the OP in that other thread did and pull a male connector off a board and hack it to fit the Dell DC connector. After that you'll take the wires and solder them to the points of where the DC in pins are on the PicoPSU.

It sounds a bit daunting but it should be relatively trivial to do. You'll need to short pin 5 to ground on the Dell DC connector so it will stay on. Technically you could wire up a circuit so when you press the power button on the case it will turn on the DC supply and after a short delay (100ms should do it) power on the PicoPSU although I'm not sure how shutdown would work so it's easier to just leave it powered on.

[Vicotnik]

That sounds favorable for me, I dug around and found this post that went into some detail on wiring it to a similar PSU.

Ah, I remember that thread. Helped me a lot when I first put together my first picoPSU + DA-2 system.

After looking at the post I personally have even less appetite for messing with the cables

How far can you take this without messing with the cables.

Yes it's a hassle. But it's fun. For me anyway. :p

Without messing with the cables you will probably run into trouble with a power hungry system. At first I just identified the pinout of the DA-2, then removed the original plug and used a barrel connector. The picoPSU was untouched at this time, I only messed with the brick. But I got a huge drop on the +12v rail. It was something like 12.1v at the brick (opened it up to measure) but close to 11.5v at a molex. I don't remember if I had any real stability issues but it seemed close.
I then did some cable work. I shortened the cable coming from the brick (the Dell DA-2 cable is looong), removed the barrel connector (right now I don't have a connector at all, the brick is permanently connected to the computer but I have a few good connectors I can use, specified for 15A or something like that), and also feed the ATX +12V 4pin connector directly from the brick. The +12V rail is now at a much nicer level, like 11.9 or something.

The picoPSU needs tweaking to work fine on a relative high power system. I will go Sandy Bridge myself soon and I will probably use the Intel graphics and might go with the Winmate instead of the picoPSU. I will probably lose some efficiency but I will get a filtered +12V rail.

[Ashex]

The picoPSU needs tweaking to work fine on a relative high power system. I will go Sandy Bridge myself soon and I will probably use the Intel graphics and might go with the Winmate instead of the picoPSU. I will probably lose some efficiency but I will get a filtered +12V rail.

What sort of tweaking do you think will need to be made?

[Vicotnik]

Like I said, I like to remove the weak barrel connector that comes standard with the picoPSU. It's rated for 5A or something if I'm not mistaken. I also feed the ATX +12v 4pin connector directly from the brick. It helps to have a few spare things lying around, I usually cut the ATX +12v 4pin connector from a broken PSU found in the electronics recycle bin. A multimeter to check voltages is also nice.

Imho, unless you are able to handle a soldering iron and do some basic work with it, the picoPSU is only viable for low power systems. If you have a small system that you can power with a 60W brick, the picoPSU is fine like it is. When you approach 150W or so at peak, not so much.

[Ashex]

Like I said, I like to remove the weak barrel connector that comes standard with the picoPSU. It's rated for 5A or something if I'm not mistaken. I also feed the ATX +12v 4pin connector directly from the brick. It helps to have a few spare things lying around, I usually cut the ATX +12v 4pin connector from a broken PSU found in the electronics recycle bin. A multimeter to check voltages is also nice.

Gotcha, those are the modifications I expected to make, I had assumed you were talking about modifications to the PicoPSU itself aside from removing the barrel connector.

[MnM]

Ashex, where are you going to buy the 2500T from? I am looking in the UK, and they are not widely available as far as I can tell.

Matt

[Mats]

Ashex, where are you going to buy the 2500T from? I am looking in the UK, and they are not widely available as far as I can tell.

Matt

Pick the 2300 instead, that's what I'd do: viewtopic.php?f=23&t=61903&start=14

Here's the power consumption of the 2300, if you lower the speed to the same as the 2500T. It's under 45 W:

[MnM]

Thanks Mats. I had read your posts, and the reviews by xbitlabs and others. I'm not sure about them.

xbit seemed to say that the T series had great performance per watt, but didn't think the same of the S series... well, these are similar, but not quite not identical processors, using different clocking and voltage schemes. Also, the intel specs give different bus/core ratios. I don't know exactly what that means, but they are different. I am planning to use an H67 chipset for a mini-ITX build, and am not sure how much control I will have over clocking and voltages. I haven't played with this stuff for a decade, but my impression is that there isn't much to play with on the H67 chipsets - I could well be wrong though.

I'm building a headless media server. No great graphics needed, won't ever use a tv-out, but would appreciate a bit of speed when I install software, manage the media library, maybe do some bulk tanscoding, etc. And Sandy Bridge has great low-power/low temperature performance, so I want to make the most of that. I want to use this Chenbrocase, which has a built in 120W ps, four low power hard drives (Samsung Eco F4s) and an SSD. The motherboard allows for staggered spin-up of hard drives.

There is an external 180W ps available, I'd be very happy to squeeze this into the capabilities of the internal 120W supply though.

Price wise, the 2400S series are about £10 more than 2300, and £3 more than a 2400, which I can stomach. No idea how much a 2500T costs.

Given all that, the 2500T rather than an underclocked 2300 could be a good idea... if I can source one...

Sorry for the minor hijack Ashex!

[Mats]

I'm not sure why you're having doubts over the 2300? Is it:

1 - You won't be able to lower the multiplier in BIOS?

2 - Even if you can, it will still run too hot?

These CPU's all uses the same amount of power in idle, the power used during full load is easy to measure,
and the calculations I showed you in the last post shows what kind of power draw you can expect from just lowering the multiplier.
You don't have to touch the voltage.

It's quite easy to check if BIOS allows lowering the multiplier, check the manual or some review with BIOS screen shots. Which model is it?
Even if you can't lower the multiplier, the CPU would only use about 50 W at most, which wouldn't make any difference for you.
You get a faster CPU for less money. 99% of the time it will use <15 W, no matter which model you pick.

In Germany, a 2300 cost £123, and a 2500T cost £174. Yeah, that's £50 for something you can do fix by lowering the multiplier yourself.
http://geizhals.at/deutschland/?cat=cpu ... +i5&sort=p

The 2500T is sold without a cooler, it's not a retail part, meaning that you'll get a much shorter warranty. Retail CPU's have a 3 year warranty.
This is because it's made for computer manufacturers primarily, and they have their own warranty.

[MnM]

Indeed, I won't be able to lower the multiplier, and it is peak power draw which I worry about, as I want to benefit from a built-in 120W psu in my case of choice.

Your last comment is contradicted by reviews on this website. It's hard to establish how much would be saved by a lower power processor, but a fully stressed 2500K caused system power draw to peak significantly above idle, the range was from 20W to 121W. I do not want my PC to fall over when, for example, I ask dbpoweramp to do a transcode of my library.

[ces]

Indeed, I won't be able to lower the multiplier, and it is peak power draw which I worry about, as I want to benefit from a built-in 120W psu in my case of choice.... I do not want my PC to fall over when, for example, I ask dbpoweramp to do a transcode of my library.

If that is an issue, the 2100t (or even the 2390T) is a good choice. The difference between the two is that 2390 has turbo.... which for your concerns won't hurt and would still give you a performance boost for short bursts when you need it.

If you are really paranoid, look at the Pentium G620T when it becomes available. It is just two cores without multithreading. It is 2.2gHz to the 2100T's 2.5 gHz.

But even the basic low cost 2100 will probably do you OK. See:
http://www.xbitlabs.com/articles/cpu/di ... html#sect0

The most important factor is probably to get an Intel motherboard which will have the lowest energy utilization of any other mainstream 1155 motherboard.

[Mats]

Indeed, I won't be able to lower the multiplier, and it is peak power draw which I worry about, as I want to benefit from a built-in 120W psu in my case of choice.

Yes, it all depends on which model you buy. For instance, Asus P8H67-I can lower the multiplier AFAIK.

Your last comment is contradicted by reviews on this website. It's hard to establish how much would be saved by a lower power processor, but a fully stressed 2500K caused system power draw to peak significantly above idle, the range was from 20W to 121W.

But that CPU that's running 1000 MHz faster than the one you're looking for? Besides, it's running 4 instances of Prime95 and Furmark,
and it's impossible to get such power usage with real world. You said you won't be needing the graphics, which will lower the power draw.

I think you should have a look at which programs you're using that supports more than two cores, as it should affect the choice of CPU..

The most important factor is probably to get an Intel motherboard which will have the lowest energy utilization of any other mainstream 1155 motherboard.

I'm not sure Intels models have enough SATA ports.

[MnM]

Indeed, I need four SATA ports for RAIDed data drives, and another internal SATA port for OS. Only the Zotac can handle this, and apparently the non-deluxe Asus ITX board - well I checked that one out and it has no eSATA, which I consider an essential for speedy ad-hoc backups/restores for disks usually stored offsite. So Zotac it is - annoyingly, they tend to have higher power consumption, according to everybody.

I shan't be using graphics (dunno to what extent graphics are actually used when remoting in, but I will never use it for hardcore graphics or TV). But, I decided I want an i5, with the full feature set, and I will fully utilise four cores if doing batch transcodes with dbpoweramp, for instance.

I will look into using a 2300, as apparently a new BIOS update for the Zotac gives more flexibility, but the bus/core ratio is supposed to be locked with the H67 chipset, so I am not sure how much flexibility you get - tbc.

Right - I challenge you on your source for saying that the 2500T is just a stepped down 2300. It is my belief that the low power processors are stable under lower voltages than the non-low power processors. So, in that sense, they are higher-spec parts, like the 2500 is higher spec than the 2400 - it is certified to work with higher frequencies for much the same voltage. Also, the idle speed of the graphics in the 2500T is lower than for non-T processors. My question to you is - are you sure that they are just stepped down 2300s, or do Intel perform any selection process on their yield to identify processors suitable for the T suffix? The answer to this, and the flexibility of the Zotac BIOS update, and availability, and price, will impact whether I get a 2400, 2400S, or 2500T (2300 misses a few features, and that is simply annoying - I need those AES instructions )

[Mats]

Indeed, I need four SATA ports for RAIDed data drives, and another internal SATA port for OS. Only the Zotac can handle this, and apparently the non-deluxe Asus ITX board - well I checked that one out and it has no eSATA, which I consider an essential for speedy ad-hoc backups/restores for disks usually stored offsite. So Zotac it is - annoyingly, they tend to have higher power consumption, according to everybody.

Your only alternative to eSATA is USB 3, but I understand if you don't want to buy new USB 3 HDD enclosures if you already have eSATA.

I will look into using a 2300, as apparently a new BIOS update for the Zotac gives more flexibility, but the bus/core ratio is supposed to be locked with the H67 chipset, so I am not sure how much flexibility you get - tbc.

Since SpeedStep lowers the multiplier it's obviously not locked. It can be difficult to find a board that has the right settings though.
But you can always use the power options in Windows, set the maximum CPU speed to 80 % or whatever you feel like and you're done. Or, use some third party program.

Right - I challenge you on your source for saying that the 2500T is just a stepped down 2300.

SPCR have proven before that there's no binning needed at least for the 9550s.
http://www.silentpcreview.com/article954-page5.html
It's only a TDP difference of 30 W, but on the other hand, both CPU's runs at the same clock.

There's no magic going on and there's no need to since the 2500T is so much lower clocked than the regular models.

[CA_Steve]

My question to you is - are you sure that they are just stepped down 2300s, or do Intel perform any selection process on their yield to identify processors suitable for the T suffix? The answer to this, and the flexibility of the Zotac BIOS update, and availability, and price, will impact whether I get a 2400, 2400S, or 2500T (2300 misses a few features, and that is simply annoying - I need those AES instructions )

Intel will maximize their margin $'s per wafer by proliferating part numbers/specs to sell as many dice off the wafer as possible. This leads to a variety of binning methods during testing.

My guess: When they test, they'll see which parts meet what speed grade. After that, they can sub-bin for TDP to get more $ for a specific speed grade.

Other factor: demand/backlog. So, while a part might meet a higher speed grade, it might get marked and sold as a lower grade to meet demand. This is why some parts can be overclocked better than others.

So, can you underclock/undervolt a higher TDP part and get similar results to a same specc'ed/higher price part? Probably.

[Ashex]

Anyone know what the amp rating is for the 4-pin connector on the PicoPSU is?

I'm talking about the 4-pin connector used to attach the barrel connector, see this photo.

I was going to use part of a P1 connector to plug the DA-2 into an adapter that would connect to the 4-pin connector but I wasn't very sure how well the hacked P1 would carry the current so I'm going to modify the DA-2 to use a high amperage connector (30A rating) and connector to an adapter to the 4-pin however I don't know if it will handle the 18A through it.

Edit: Found a 20 to 24 pin connector so all the wires are already pinned in, if the 18A are split across 3 wires, 18 gauge wire should be able to carry the full current, correct? I bought 16 gauge as it's rated for 22A versus 16A for 18 gauge.

Edit Edit: Remembered that run current across multiple wires splits it so I'm fine. Looks like I won't need to use the parts I bought! Just need to get the 4-pin connector which I'll be removing from a psu at work.


Also, I'm getting my 2500T through eXtremePCgear, it's pre-ordered right now and when I asked them when it would be delivered that said they should be getting it the end of this month so I'm banking on it being shipped out next week.

[dvbmaniak]

How to connect dell da-2 directly to the P4?
Topic:
http://forum.purepc.pl/Modyfikacje-f15/ ... 36262.html
diagram:

01schemat (1).jpg

How does it work?
http://www.youtube.com/watch?v=72HO8yS- ... r_embedded


Configuration

procesor: AMD Phenom II X6 1055T (TDP 125W)
płyta główna: Asus M4A88T-I Deluxe (Mini-ITX)
chłodzenie: Noctua NH-U12P SE2
RAM: 2x Kingston 2GB 1333MHz DDR3 CL9 (SODIMM)
SSD: OCZ Vertex 2 E 120GB 2.5”
napęd optyczny: SH-S223L SATA (ze starego kompa)
zasilacz: picoPSU-160-XT (160W) + Dell DA-2 (12V 220W)
obudowa: Lian Li PC-Q07B

Sorry for my English which I do not know, the links found on the web.

[Vicotnik]

Making use of the reserved cable from the DA-2, very clever.

I do it the quick and dirty way - the DA-2 triggered to be active all the time and ATX +12V active always. I use a powerstrip with a switch to turn the whole thing off sometimes. I've noticed no practical downside to have the ATX +12V powered all the time. Seems to be working..

[Ashex]

The newer PicoPSU has a connector behind the barrel plug so you can attach different connectors, so instead of removing it and wiring the DA-2 directly to the board with a plug I decided to make an adapter plug to connect it.



It looks a bit messy as my dremel died (you can see it charging) and wasn't able to clean-up the plug on the DA-2 side.

It plugs in directly and I measured ~12V coming through the adapter to figured I was good to go. I plugged the PicoPSU into my computer and then hooked up the DA-2 then hit the power button and it powered up! That's about as far as it gets as the computer does not POST. I then plugged in the DC brick I use with my other PicoPSU and again the computer powered on but no POST. Plugged in my other PicoPSU with its DC brick and still no POST.


It appears that I didn't wire something properly and killed my motherboard.

Edit: computer is working again, had to reseat the memory so it was probably an unrelated issue (motherboard is really picky about the memory). I'm going to play it safe and just take the power supply in to work tomorrow and test on a spare computer.

[Ashex]

Looks like the i5-2500T will be a hard one to find, the site I pre-ordered from hasn't gotten it in stock so I'm going to cancel my order and get the i5-2400 and see about undervolting it (I'll have to do some reading).

[ces]

Looks like the i5-2500T will be a hard one to find, the site I pre-ordered from hasn't gotten it in stock so I'm going to cancel my order and get the i5-2400 and see about undervolting it (I'll have to do some reading).

Just get a cheap 2400, then use windows power management to turn it down to 80% or even 40%.

All the Sandy Bridges idle at about 4 watts. The only difference is in the unlikely even that you run them at full load. That you can control through the windows operating system. It makes no sense to pay Intel the extra money to do it in hardware when you can do the same in software.... much more simply much more cost effectively.

[Ashex]

Just get a cheap 2400, then use windows power management to turn it down to 80% or even 40%.

Interesting, I wasn't aware I could do that in Win7. As soon as I get the pre-order refunded I'll get the 2400.