NVIDIA Shows Off ION-based Devices, Including Pegatron IPP7A-CP

Rob Williams

Editor-in-Chief
Staff member
Moderator
From our front-page news:
By now, you should know what NVIDIA's ION platform is all about. Simply put, it's a platform that utilizes Intel's Atom processor, but pairs it with NVIDIA's ION GPU for better graphics support, which would improve HD playback and very light gaming. As you'd expect, between ION and Tegra, NVIDIA had a lot to show off, and I have to say, I like what I saw, but I'll talk about Tegra later.

What impressed me the most ION-wise was the IPP7A-CP from Pegatron, as it was easily the smallest ION solution on display, and as a result, the most impressive. Just two months ago, Acer announced their ION-based AspireRevo, and when I first saw it, I was impressed. After all, it was a small build, yet offered so much functionality.

After taking a look at the Pegatron offering, the AspireRevo looks a little bit embarrassing, as you can see in the below photo. In a rough estimation, the IPP7A-CP is about 2.5x less the volume of the AspireRevo, but by taking a look in the back, you'll still see a wide-array of connectivity options.


I'm not sure if this couldn't be done with the straight Atom-based platform (using Intel's on-board graphics rather than NVIDIA's), but the fact that we have a rather powerful little machine in a form-factor this small is truly incredible. With the included GPU, you could actually take advantage of HD playback... just imagine this thing next to the TV! No word on release date, but hopefully we won't have to wait too long.
 

2Tired2Tango

Tech Monkey
After taking a look at the Pegatron offering, the AspireRevo looks a little bit embarrassing, as you can see in the below photo. In a rough estimation, the IPP7A-CP is about 2.5x less the volume of the AspireRevo, but by taking a look in the back, you'll still see a wide-array of connectivity options.

Impressive. Although even with ultra low power components, I do wonder about heat dissipation... I'm guessing the whole thing runs toasty warm, as is common for anything that small... Still with HD playback, headless operation, popular OSs... the VSFF (Very Small Form Factor) market has come a long way. In a couple of years we'll be looking back and wondering why "old style" computers ever had to be so big.
 

Rob Williams

Editor-in-Chief
Staff member
Moderator
I don't think "big" computers will <em>ever</em> go away. They'll always be cutting-edge and have their place. Not everyone wants a small computer that fails to handle the latest games at killer resolutions. There's certainly a market for both, and I don't see that ever changing.

For regular PCs, smaller IS better. It's just when you want to get hardcore work done, or want to game, or want to choose your own hardware, that a bigger PC will be necessary.
 

2Tired2Tango

Tech Monkey
I don't think "big" computers will <em>ever</em> go away. They'll always be cutting-edge and have their place. Not everyone wants a small computer that fails to handle the latest games at killer resolutions. There's certainly a market for both, and I don't see that ever changing.

For regular PCs, smaller IS better. It's just when you want to get hardcore work done, or want to game, or want to choose your own hardware, that a bigger PC will be necessary.

Hi Rob,
I do agree that larger computers will be with us for quite some time. I do not, however, think there is any overarching technical need for it.

I'm thinking Micro-ATX will soon overtake ATX as the standard form factor. How many ATX mobos are there out there with 5 buss connectors, all completely empty? 2 pci and 1 video connector are more than enough in 90% of cases and the reduction is size is signficant. It only makes sense that cases should get smaller as well. Most of the systems under my care are already Micro-ATX and have been for quite some time.

Yep, there are still some applications that reportedly won't run on mini-itx worth a darn. However, I think that speaks more to the immaturity of the form factor than to any inherrent loss of functionality that can be coupled to size.

What needs to happen before the downsizing will be possible without a concomittant loss of functionality are three things:

  1. Those engineering the Mini-ITX and other Very Small Form Factor (VSFF) setups will need to rethink the current buss structures. They should be pushing for a VSFF buss structure that is presented as a single .1" pin header on the motherboard. This connector would in turn connect to an optional standard riser card or daughter board that offers 2 or more connectors for VSFF modules. This will overcome the grumbles about expandability and still provide for very small cases where extra slots are not needed.
  2. Once a VSFF buss is decided, we will need peripheral manufacturers to get on board creating modules that will plug into the buss card. It's likely possible, right now, to create modules no bigger than laptop memory that will replace most PCI cards currently on the market. By moving external connectors from the modules to pin-headers and ribbon cables, we further gain new flexibility in case design.
  3. From everything I've read it's the total reliance upon integrated video that's messing with the success of the platform. There is a noticeable performance hit with the current system of having the IGP and CPU share memory. The behavior of both devices is affected. My thought is to outright ditch the IGP. Instead provide a properly buffered interface to VSFF video cards. Not only would this bring the price of motherboards down, it would greatly reduce heat and power consumption in the chipsets. Of course this relies upon #2 actually happening.

I've been around electronics a long time and by and large the move has been that every 10 years or so we see a significant reduction in size. Sadly this is sometimes accompanied with a more than significant reduction in quality but this need not be the case with the new VSFF machines... The capability exists already to shrink most computers to half size or less. So I'm going to suggest the industry (and I) get on with it.
 

Rob Williams

Editor-in-Chief
Staff member
Moderator
When I made that comment, I was thinking more along the lines of the Mini-ITX and smaller form-factor. I do agree that for most people, mATX would suit their needs just fine. The biggest limiting factor to me are the GPUs. The next-gen from NVIDIA are looking to top-out at 300W... and that's just ridiculous. It's hard to cram something like that into a small chassis and not expect it to overheat.

As for integrated graphics... it is a huge limiting factor. In order to have a graphics capability worthy of being proud about, you need a discrete card (unless you are talking HD video). I don't think we'll always have this limitation, but with companies pushing out higher TDP graphics cards, it seems like this issue doesn't have a huge focus.

I'm not against the change though. I'd love to have an mATX machine next to this desk. I'd love to look at what's there now in ten years and think, "Holy crap, that's ugly!", haha.
 

Kougar

Techgage Staff
Staff member
2Tired2Tango, I think that'd be cool and I'd be all for it, but I'd like to mention a few things.

1) Finding a case designed for Mini-ITX and other related VSFF's isn't easy, I think there would need to be many more options here for it to take off. Yet adding connectors for optional daugher boards would again pose new form-factor issues, because it wouldn't adhere to the original form-factor or fit many cases that would fit the original card. There would need to be an official form-factor specification and then industry support behind it.

Shuttle has been making small form factor PC's for years now, yet many of the motherboards they use are actualy custom designs, built to order motherboards that fit the specific case design and follow pretty much zero specifications on form factor. Just look at this very crazy thing they ordered that's an upcoming P55 board (obviously manufactured by Gigabyte): Link Of course, who needs standards... I thought this design was nice. :D

Seriously though, along your line of thought there is this: Pico-ITXe if you've not seen this already. This odd board allows multiple expansion boards to be plugged into it like a lego set (every Geek's dream?), and at least up to three boards if I recall one image I've seen but can't seem to find again. Here's one optional add-on board. Link

Some better images are Here and here, which show the base board and the I/O module board. Details on those are over here.

Anyway, the people behind the Pico-ITXe form-facter are trying to solidify it as a standard along with their stackable connectors (ties into your #2 comment) and other designs, will see how that goes: http://www.sff-sig.org/

With #3, the biggest problem is finding a chipset that is actually built on current fabrication technology and is specifically designed for just what is needed. That by far would have the largest impact on board complexity, power consumption, and cooling. Atom might be good, but it uses a desktop chipset that was designed to do everything from multiple serial ports, parallel port, 6 or more PCI slots, full PCIe 16x bandwidth, IDE controller, floppy controller, 8+ USB ports, etc and a few million other things it could or would never use in an ITX formfactor. Intel also builds them using old fabrication technology, for 945 it's still at a whopping 130nm size.

Just look at this image of Intel's idea of a mini-ITX board. Not only is the Dual-core Atom roughly the same size as the 130nm 945GMCH silicon, but the chipset actually runs hotter than the Dual-core Atom processor and requires active cooling. And we have not even reached the ICH7 southbridge chip yet, which is built on the same 130nm fab size and rated for about 4 watts consumption. The ICH8 was 90nm, current model is the ICH10 but I'm not sure what fab size that one uses (heard it was 90nm still). G35 chipsets were 90nm, G45 chipssets were 65nm... that fabrication node would shrink the G945 chipset silicon by significantly more than half it's size!

Honestly, if Intel would build just one chipset on it's latest processing node then there could be some impressive gains, even better if they scrapped the ICH outright and built a custom IOH to handle whatever the processor couldn't. Thankfully they are at least doing the latter with Pine Trail... Not only will it be a single chip IOH, but it will allow current Atom motherboard PCB designs to be simplified from 6-layers down to just four.
 
Last edited:

2Tired2Tango

Tech Monkey
2Tired2Tango, I think that'd be cool and I'd be all for it, but I'd like to mention a few things.

1) Finding a case designed for Mini-ITX and other related VSFF's isn't easy, I think there would need to be many more options here for it to take off.

I agree. We seem to be just at the beginning of this, despite the form factor having been around for a while for use in embedded systems. One application really got my attention...The Sienna ham radio transceiver kit with a mini-itx system embedded.

Thing is I don't think VIA thought they'd bust out into the end user market as they are just now doing, so certain limiatations were accepted in the design. This is forcing a rethink, especially of casing and storage capabilities.

Perhaps one of the better cases available is the Foxconn chassis. This guy isn't as small as some but it does allow the use of standard optical and hard disks, rather than the less durable laptop ones. And it will take the Zotac AMD boards with a 1u cooler.

MSI has a nice mini-itx case as well, but it's part of their Barebones WindTop and I don't think it can be had without the motherboard.

Most of the other cases currently around (in Canada, anyway) are diskless, intended to be used in cop cars and electronic signs, which as you point out puts a serious crimp in our style.

Yet adding connectors for optional daugher boards would again pose new form-factor issues, because it wouldn't adhere to the original form-factor or fit many cases that would fit the original card. There would need to be an official form-factor specification and then industry support behind it.

I think the goal is to design a whole new form factor then standardize that. I'm certainly in no position to cause that, on my own, but there are ideas wandering around in the ole noggin that I'd just love to see happen ;)

Shuttle has been making small form factor PC's for years now, yet many of the motherboards they use are actualy custom designs, built to order motherboards that fit the specific case design and follow pretty much zero specifications on form factor.

I like Shuttle's concept. It's not tiny by any means, but it's certainly better than some garbage can sized box sitting on the corner of your desk.

Seriously though, along your line of thought there is this: Pico-ITXe if you've not seen this already.

I'd not seen that, thank you.

My concept is similar, but more vertical than theirs. What I saw in my mind's eye was a vertical card that plugged into the motherboard and housed perhaps 4 SODIMM-like sockets that held the expansion cards on an angle above the motherboard... in much the same way laptop memory sits on the mainboard... overlapped or angled to save as much space as possible, with ribbon cables going over to the I/O panel.

A pci riser card with sockets for a couple of half-height pci cards will give about the same result without using up a ton of space.

Anyway, the people behind the Pico-ITXe form-facter are trying to solidify it as a standard along with their stackable connectors (ties into your #2 comment) and other designs, will see how that goes: http://www.sff-sig.org/

And yet another Google Alert hit's the road :rolleyes: This will bear watching, at least until I can see if it flys or flops.

With #3, the biggest problem is finding a chipset that is actually built on current fabrication technology and is specifically designed for just what is needed.

Exactly... the current trend seems to be to stuff as much into the I/O chipset as possible and here I am suggesting the exact opposite... Strip it down to half a dozen USBs, a Network connector, PS2, and drive interfaces. Do the graphics through a separate (hopefully small) board that might even have a separate power connector to allow for smaller tracelines on the mainboard.

Honestly, if Intel would build just one chipset on it's latest processing node then there could be some impressive gains, even better if they scrapped the ICH outright and built a custom IOH to handle whatever the processor couldn't. Thankfully they are at least doing the latter with Pine Trail... Not only will it be a single chip IOH, but it will allow current Atom motherboard PCB designs to be simplified from 6-layers down to just four.

The Atom-itx implementation strikes me as something of a technical curiosity. Given that Intel (of all people) could have produced an ion-like single chip solution on the same process level as the Atom itself... one has to wonder why they didn't.

Thanks for your comments and the new information... great stuff.
 
Last edited:

2Tired2Tango

Tech Monkey
When I made that comment, I was thinking more along the lines of the Mini-ITX and smaller form-factor. I do agree that for most people, mATX would suit their needs just fine.

Certainly that's been my experience. I've only had one customer I couldn't switch to Micro-ATX and that's because I just can't convince him there's no performance hit for doing it... What you lose is a couple of expansion slots that most people don't use anyway.


The biggest limiting factor to me are the GPUs. The next-gen from NVIDIA are looking to top-out at 300W... and that's just ridiculous. It's hard to cram something like that into a small chassis and not expect it to overheat.

Yes, it is ridiculous. Aside from extreme gaming, I can't think of anyplace such a card might be necessary. I use the ASUS Silent en8400 cards quite often and so far nobody's complained... Even in HTPC environments that card acquits itself nicely.

As for integrated graphics... it is a huge limiting factor. In order to have a graphics capability worthy of being proud about, you need a discrete card (unless you are talking HD video). I don't think we'll always have this limitation, but with companies pushing out higher TDP graphics cards, it seems like this issue doesn't have a huge focus.

I tend to agree. The IGP is there mostly for people who don't need performance or simply don't know any better. It's also a good recourse in a system that won't boot... pull the video card and see if it will run it's own. But what you got is what you're stuck with in that case... I would much prefer they go plug in video with the VSFF setups rather than relying exclusively on chipset video. Of course the performance hit could be neutralized quite nicely if they would provide "side memory", a video memory connector that is not shared with the CPU... a graphics card with 2 gigs of memory, sweet!

I'm not against the change though. I'd love to have an mATX machine next to this desk. I'd love to look at what's there now in ten years and think, "Holy crap, that's ugly!", haha.

My main computer is micro atx, IN-WIN case, Asus M2N-MX-SE, 2 gigs ram, 500gig hd, dvd optical, EN8400 video, and an ancient sound blaster PCI-64 card for midi... It's never disappointed me yet.
 

Kougar

Techgage Staff
Staff member
Rob Williams said:
The biggest limiting factor to me are the GPUs. The next-gen from NVIDIA are looking to top-out at 300W... and that's just ridiculous. It's hard to cram something like that into a small chassis and not expect it to overheat.

I gotta disagree. Current GPUs already use these envelopes, especially the dual-GPU models. In all honestly, I can measure a 200W difference between my GTX 260 being idle and under load. For the sake of argument lets give it 50W of idle consumtpion... so that's 250W of power I'm measuring from the wall due to the GPU, and this is the basic, lowest class GTX 260 with a smidgen of an overclock. :D

The Atom-itx implementation strikes me as something of a technical curiosity. Given that Intel (of all people) could have produced an ion-like single chip solution on the same process level as the Atom itself... one has to wonder why they didn't.

Thanks for your comments and the new information... great stuff.

You're welcome. I might suggest you browse over to Nordic Hardware and add them to your news feed, they seem to cover pretty much all developments SFF, ITX, and everything else related to this. They also even have covered a few mini/pico-ITX cases that I recall.

Regarding Intel, I'm not sure even they expected Atom to take off as it did. In fact I bet they didn't. I'm not sure what fabrication size the Pine Trail single-chip chipset will use, so it will be interesting if it uses the same 45nm fab size... but I wouldn't be surprised if it didn't. Intel has way too many fabs using to many different out-dated process technologies that they need to keep running somehow, and fobbing chipset production off on them is how they've done it.
 

Rob Williams

Editor-in-Chief
Staff member
Moderator
2Tired2Tango said:
The Atom-itx implementation strikes me as something of a technical curiosity. Given that Intel (of all people) could have produced an ion-like single chip solution on the same process level as the Atom itself... one has to wonder why they didn't.

One would have to assume this was due to the fact that their graphics chip is lackluster ;-)

2Tired2Tango said:
Certainly that's been my experience. I've only had one customer I couldn't switch to Micro-ATX and that's because I just can't convince him there's no performance hit for doing it... What you lose is a couple of expansion slots that most people don't use anyway.

I have an mATX board here for review (ASUS Rampage II GENE), so perhaps that article will prove that this is the case. I'd like to review more mATX boards I think, which might mean I need to get a proper chassis for it. Have recommendations?

Plus, expansion may be the reason many people look away, although certainly not all. An mATX board might support 6 S-ATA drives, but if the SFF chassis can't fit them all in, then it's kind of useless.

2Tired2Tango said:
Yes, it is ridiculous. Aside from extreme gaming, I can't think of anyplace such a card might be necessary. I use the ASUS Silent en8400 cards quite often and so far nobody's complained... Even in HTPC environments that card acquits itself nicely.

Of course, that's true. But the people you build machines for are obviously not gamers. Anyone who games is going to want something more than just a "mainstream" GPU.

Kougar said:
Current GPUs already use these envelopes, especially the dual-GPU models.

Who's using a dual-GPU in a SFF chassis? That's my point... no one wants to put something that requires a massive cooler to properly dissipate heat in a small chassis.
 

2Tired2Tango

Tech Monkey
I have an mATX board here for review (ASUS Rampage II GENE), so perhaps that article will prove that this is the case. I'd like to review more mATX boards I think, which might mean I need to get a proper chassis for it. Have recommendations?

I've always liked the IN-WIN line. They are ruggedly built, reasonably priced, no sharp edges, mostly rolled corners, lots of ventilation and easily accessible. Unlike many M-ATX cases these ones use full sized ATX power supplies. My desktop one is one of their older models from the V series, on it's second paint job, still in perfect condition and I'm getting a little reluctant to part with it.

Plus, expansion may be the reason many people look away, although certainly not all. An mATX board might support 6 S-ATA drives, but if the SFF chassis can't fit them all in, then it's kind of useless.

Well do keep in mind that usage topologies are changing, Rob. Where only a couple of years ago I was installing more drives on top of more drives (oddly right about the time UTorrent came out). Now I'm being asked for compact desktop units with networked mass storage... people are realizing they don't want all their eggs in one basket. It's quite typical for me to install systems with under 300gig hard disks on each desk with a hot swapping network server hosting several terabytes stuffed in a closet someplace.

The kind of tech-savvy person who will use SFF or VSFF systems on their desktop is also the same person who's likely to have a fully networked small office or home. By and large they don't want huge boxes with their precious data out where it can be kicked, splashed, stolen, sat on or marked as territory by the dog.

The "BKC error" types who want smaller footprints tend to either not know or not care about networking and expansive storage... these are the people I've been interested in the Atom systems for... for most a laptop would suffice, but they still want the full sized keyboard and screen.

The market segment for ATX is quite different from this. By and large the "big board" guys (well, the ones who aren't over-compensating for something else) are either into heavy gaming, only have one machine in the house or want to use them as servers...

Of course, that's true. But the people you build machines for are obviously not gamers. Anyone who games is going to want something more than just a "mainstream" GPU.

True enough. My demographic tends to be one of:
  1. "Hip young couple wanting HTPC for the living room"
  2. "Office drones who only care about getting the job done"
  3. "Retired or disabled people looking for low cost systems"

...Three markets that are sadly underserviced and a wonderful little niche for me.


Who's using a dual-GPU in a SFF chassis? That's my point... no one wants to put something that requires a massive cooler to properly dissipate heat in a small chassis.

Yep, that's a problem. Most M-ATX systems are getting much better about cooling. IN-WIN does it very well... Even then you're still not going to put a 300watt anything in a space that small without boiling the paint off the boards.

Simple test... get a 100 watt work lamp (not flourescent) put it inside an empty case with the normal compliment of fans running... measure the temperature rise over time. The lower the final temp and the longer it takes to get there... the better the case. The spec is rated as "degrees per hour"... 10 years ago you would see 25 now you typically see 10 or less.

And finally, you aren't going to sell M-ATX systems to the kind of people who need that sort of over-the-top graphics capability... The smaller motherboards just won't support that kind of current and many cases won't support the needed powersupplies.
 

Rob Williams

Editor-in-Chief
Staff member
Moderator
2Tired2Tango said:
I've always liked the IN-WIN line. They are ruggedly built, reasonably priced, no sharp edges, mostly rolled corners, lots of ventilation and easily accessible. Unlike many M-ATX cases these ones use full sized ATX power supplies. My desktop one is one of their older models from the V series, on it's second paint job, still in perfect condition and I'm getting a little reluctant to part with it.

Ahh alright, I'll check out to see what they have. Another option is SilverStone... they seem to be the de facto for HTPC, and even the chassis shown off at Computex would be suitable. I'm doubtful many people purchase mATX boards to install into an ATX chassis, so I think getting a chassis like this in is important. Don't look too forward to benchmarking each board that way though ;-)

2Tired2Tango said:
Now I'm being asked for compact desktop units with networked mass storage... people are realizing they don't want all their eggs in one basket.

That's true, but I STILL don't think that applies to everyone. NAS boxes are expensive, and if all you need is extra storage, the least-expensive way to add it is to get a bare drive and toss it in your rig. That also avails the best possible performance as well. I'm not opposed to the idea of network storage being the ultimate storage solution, but I don't see that happening right away.

2Tired2Tango said:
Simple test... get a 100 watt work lamp (not flourescent) put it inside an empty case with the normal compliment of fans running... measure the temperature rise over time. The lower the final temp and the longer it takes to get there... the better the case. The spec is rated as "degrees per hour"... 10 years ago you would see 25 now you typically see 10 or less.

Very interesting methodology there, haha. I like it.

2Tired2Tango said:
And finally, you aren't going to sell M-ATX systems to the kind of people who need that sort of over-the-top graphics capability...

That's all I was saying, really. I don't see huge computers dying off soon. I'll be happy when they do, though. I'd love the idea having a PC so quiet, I just don't hear it.
 

Kougar

Techgage Staff
Staff member
That's all I was saying, really. I don't see huge computers dying off soon. I'll be happy when they do, though. I'd love the idea having a PC so quiet, I just don't hear it.

That's why I watercool. Every case fan is undervolted to 5-7v or is a Noctua NF-P12. Nothing like running a brutally overclocked high-end system with near silence... I could make it completely silent but due to the roar of the AC unit there's no use in going that far. As it is the case is right next to me, but I don't hear it.
 

2Tired2Tango

Tech Monkey
Ahh alright, I'll check out to see what they have. Another option is SilverStone... they seem to be the de facto for HTPC, and even the chassis shown off at Computex would be suitable. I'm doubtful many people purchase mATX boards to install into an ATX chassis, so I think getting a chassis like this in is important. Don't look too forward to benchmarking each board that way though ;-)

One thing to watch *very* closely with these M-ATX cases is the height. Some of the newer ones are only 3.5 inches tall. Unless they arrange the expansion cards sideways with a riser board, you aren't going to put a full sized Video or PCI card in them. Then, even if they do give you the riser board, often the cases only provide cut-outs for the "low profile" PCI cards... which aren't the easiest things to find.

To accept full sized buss cards you're looking at a case, minimum 5 inches tall.
(But I'm betting you already knew that... didn't you?)


That's true, but I STILL don't think that applies to everyone. NAS boxes are expensive, and if all you need is extra storage, the least-expensive way to add it is to get a bare drive and toss it in your rig.

And if your machine is stolen or there's a system killing event?

I'd rather sacrifice a bit of performance than lose all the family pictures, financial records and accumulated media files.... Worse still if there are business records involved.

Very interesting methodology there, haha. I like it.

The "heat rise" test is something of an industry standard in other areas of electronics (commercial stuff... high end audio, industrial controllers etc) where it is understood that proper thermal solutions in housings can change Mean Time Between Failure (MTBF) specs by a factor of years.

That's all I was saying, really. I don't see huge computers dying off soon. I'll be happy when they do, though. I'd love the idea having a PC so quiet, I just don't hear it.

True enough. If nothing else, stick one in your closet and put in a mess of raid drives and use it as a backup device :eek:
 

2Tired2Tango

Tech Monkey
That's why I watercool. Every case fan is undervolted to 5-7v or is a Noctua NF-P12. Nothing like running a brutally overclocked high-end system with near silence... I could make it completely silent but due to the roar of the AC unit there's no use in going that far. As it is the case is right next to me, but I don't hear it.

Actually there are near silent fans out there, but they don't exist in the world of computers.

UpFront look for PanaFlo fans. I'll explain why in a minute...


A few things to keep in mind with small fans...

  • Many manufacturers overspin their fans. This is done because they correctly expect them to slow down substantially with age. Bearing friction, dust load, magnetic discharge and other forces play a role in reducing fan speed with age. Overspinning leads to cavitation, actual vaccums forming behind the individual blades of the impeller causing vibrations in the airflow as the air is evacuated then rushes in to equalize pressure. This causes vibration on the fan blades that can be clearly audible at full speed (ZZZZZZZZZ...)
  • Hooked blades move more air than straight blades but they also vibrate more than straight blades. Couple the flexibility at the tip of a hook shaped blade with some cavitation and it can be quite annoying. I've got a whole box of these fans I've pulled out of machines over the years and under stroboscopic analysis you can actually see the tips of the blades moving in erratic manners as the fan spins (BZBZBZBZBZBZB....)
  • Ball bearings make noise. At lowered RPM, which is very popular these days a ball bearing fan can actually make more noise than it does spinning full up. Quite simply you end up hearing the bearings rolling, metal on metal in their housings. (RRRRRRRRRR....) At higher speeds this sound moves above the range of human hearing but then you start getting cavitation problems, which produce a different kind of noise. (ZZZZZZZZZ...)
  • Brushless DC motors do not rotate smoothly. These are essentially a magnetically assisted pulse engine, that just happens to move in a circle. Carefully timed pulses of DC are applied to a coil in the stator causing magnetic repulsion in the rotor, which causes motion. At lower speeds this becomes very noticeable and some fans may not even start. (FFFFFFFFF...) Add to this that most switching power supplies react badly to sudden, brief changes in supply loading (on a millisecond by millisecond basis) and you can get a combination of uneven rotation and uneven supply voltage which only exaggerates the effect.

So what's the solution?

Oil impregnated, nylon sleeve bearings and a small fix to the way we apply power to fans.

Panafllo fans very nicely address this problem with their "Hydro Wave" bushings. The idea is that the spinning fan blade is suspended in a thin film of moving oil. As the shaft spins oil is pulled along with it, causing a pumping effect, that surrounds the shaft in a thin layer of oil that prevents any mechanical contact between the shaft and the housing, practically eliminating mechanical noise. There are other compaines beginning to use the same concepts, but Panaflo is the only one I would presently consider to be "affordable". (That will likely change as the demand for quieter fans increases.)

Now about that small fix... I'm attaching a little schematic. Anyone can do this... all you need to do is hook a Diode in series with the positive lead of the fan and a capacitor across the leads between the diode and the fan to even out the DC voltages. This will isolate the fan from it's supply getting rid of the electronic noise from uneven rotation and it will drop the fan's voltage a bit, elminating cavitation from overspin. You need to do this mod on a fan by fan basis for it to work properly. The parts will cost less than $1.00 per fan.
 

Attachments

  • PC_Fan_Silencer.jpg
    PC_Fan_Silencer.jpg
    25.5 KB · Views: 695

Kougar

Techgage Staff
Staff member
Heh, wow. Nice post!

Actually there are near silent fans out there, but they don't exist in the world of computers.

I beg to differ! Good quality 200mm sized fans are completely inaudible, or at least mine are. The three Cooler Master 200mm fans used in use on my HAF 932 are inaudible at full speed but still push a sufficient amount of air. I can hold my ear right against them and I don't hear anything beyond the gentle flow of air.

For more reasonable 120mm sizes the Noctua NF-P12 fans are completely inaudible at full speed unless you hold your ear behind one. At which point you can ever so faintly hear the fan motion, but it's inaudible beyond a few inches. I can't fathom why they included 9V and 7V undervolting wires for these fans as I don't see the need them. About 1300rpm is the cutoff point for 120mm fans, above that even if the fan is silent the airflow it pushes becomes audible, so these fans give me the maximum amount of airflow I can get without it becoming audible.

I don't pretend to be an expert on fan bearings as it sounds like you know much more than I do here, but isn't what you are describing also the same as FDB, or Fluid Dynamic Bearings? A growing number of fan manufacturers use FDB bearings, in fact most hard drive manufacturers have been using FDB bearings in their hard drives for a couple years as well. Hydro Wave bearings are not specific to Panaflo either, just look at how many cheap brands on Newegg offer them!: Link Hmm, come to look closer at it, Rexus IS Panalfo... I've never heard of Rexus before, but even if you look at the photos the fans are labeled Panaflo! Very interesting indeed...

Just reading the list at Newegg, they have fans that offer "Nanoflux Bearing (NFB)", "Magnetic Barometric", "Vapo Bearing", "Superflo", and a few others that are less exotic such as Rifle ball bearings (Better than standard ball bearings). Sony and a growing number use their own version of a fluid bearing design in their own fans, but as ya can see many manufacturers call their fluid-based bearings by different names. I don't pretend to know which are marketing words and which is actually new or improved technology, I'm sure many are not true fluid bearings at all but I know FDB is.

Then we arrive at Noctua, which is an actually reputable cooling company compared to most of those. Noctua uses an "SSO bearing" in their fans, which they describe in in some detail here. Either way, safe bet it they are at minimum as good as an FDB would be, if not better. And as you mention in your points, the blade is hooked so cavitation becomes a concern. This specific model uses notches in the blades in order to reduce the causes of cavitation in the design. Noctua has some very different blade designs that range the spectrum, and they appear to have some very credible background in airflow design to back them up. In my opinion it helps they offer a six year warranty on these fans. :)

I fully agree with your points. The problem is the bearing is not what makes the most noise on computer fans, it's the airflow. So undervolting addresses the noise created by the airflow... and then we do arrive at your main points. With that in mind I do agree the bearing type and construction of the fan play the largest role with a low RPM fan. But Panalfo fans are not as great as you might think once undervolted to address their airflow noise, and yes they are designed to be undervolted to as low as 7V.

It's a bit of a coincidence, but the seven fans that came with this Cube case (I love ebay) are all 120mm Panaflo brushless fans that use the the Hydro Wave bearings you describe. I can attest that even once I mute the airflow noise by undervolting them to around 9v (I have not measured exact voltages at specific controller settings, so I'm ballparking my figures), the motors develop a growl, two of them growl loud compared to the rest while a third has a very annoying clicking but doesn't growl as much. Once I set them to ~5V they are all almost silent unless I have my ear within a foot of them in a completely silent room. I can volt most of them higher to take advantage of the increased airflow to around 6-7V without increasing the noise, but three of them I must keep just above ~5V else the "growl" becomes audible or the clicking on one of them starts to occur. At normal 12V operation none of them growl or click, but the ~85CFM airflow is definitely loud at 35.5dBA. These are the fans You mention the Panaflo hydro weave fans specifically address some of your very valid points, however these are still brushless fans.

Your small fix to apply to these brushless fans is interesting, currently I use a Sunbeam fan controller as my Vantec fan controller didn't have sufficient channels even though I consider it of better construction. Is it only the voltage regulation in the power supply that is at issue here, or are there other considerations? I can take a multimeter and check on that easily with how these fans were built. This is however why I hold Noctua in such high regard... using a fan controller increases the power draw of a computer and dumps more heat into the case, those vRegs will burn you if you touch one! Noctua fans don't need any controller to be fully silent, and even if I had ears as sensitive as some I could change out the 3pin plug for a LNA or UNLA plug to lower the fan speed. It's cheaper, there's significantly less wiring and less mess, and the computer uses less power to function. The only real issue is Noctua charges even more than Panaflo/Rexus seems to for a 120mm model... but for a six year warranty, I'll probably get a free fan upgrade once I wear these out.

I've had plenty of experience with sleeve bearing and 1 & 2 ball bearing fans in my short time in the PC world. I'll never buy a sleeve bearing fan, as while quieter than ball-bearings they wear out in no time flat. Ball bearings last longer, but I've worn out enough Antec fans (And had enough noise issues with them) that I won't buy those or Antec fans ever again either. Rifle-bearing fans are better, but still as 2Tired2Tango says there are better options. The only fans I'd buy today would be FDB or some type of fluid bearing fan, they appear to be just as quiet or quieter than sleeve bearings with better longevity than even rifle ball-bearing fans. What a huge post this became... I'd better end it here. :p
 
Last edited:

2Tired2Tango

Tech Monkey
Somehow we've changed to "Thermal Solutions"....

Heh, wow. Nice post!

Thanks! I spend a lot of time on the little things, like quality of construction, appearance, noise and such on my systems. I charge a darned good penny for what I do and my people expect nothing short of their money's worth.

I don't pretend to be an expert on fan bearings as it sounds like you know much more than I do here, but isn't what you are describing also the same as FDB, or Fluid Dynamic Bearings? A growing number of fan manufacturers use FDB bearings

It is and it isn't... It's actually a kind of hard nylon sleeve bearing with a criss-crossing spiral cut in it to channel the lubricant. Traditional sleeve bearings are cintered bronze and tend to wear out very quickly since the oil ends up in the outer reaches of the porus bearing leaving metal to metal contact at the shaft. It is a kind of fluid dynamics but it lacks the little impellers you find on the ends of HD shafts.

Hydro Wave bearings are not specific to Panaflo either, just look at how many cheap brands on Newegg offer them!
If you check the Panasonic "DC Fans" page you will find they've divested themselves of the product line and licensed the technology to several other companies. "Panflo" is now more a Type of fan than a Brandname.

I simply recommended the brand I tend to use most often (Lucky me, local supplier!)

I fully agree with your points. The problem is the bearing is not what makes the most noise on computer fans, it's the airflow.

Yes. But to be precise it's "errors in the airflow" --cavitation, blade vibration, etc-- that makes the noise. With more rigid construction, torque control, heavier impellers and more magnetic poles, these fans could be made so quiet you'd have to look to see if they were running... However the cost of such construction would be increased power consumption and a rather hefty sticker price.


So undervolting addresses the noise created by the airflow... and then we do arrive at your main points. With that in mind I do agree the bearing type and construction of the fan play the largest role with a low RPM fan. But Panalfo fans are not as great as you might think once undervolted to address their airflow noise, and yes they are designed to be undervolted to as low as 7V.

Yes they are... But as you point out the original Panasonic/Panaflo fans were not worth a crap at anything less than about 11 volts. When run close to design spec voltages they would run dead silent for 10 years at a time. I have an old sound system I still look in on from time to time. The fans in it have been running 24/7 for nearly 15 years and all we ever do is blast them with compressed air to get the dust out.

It's a bit of a coincidence, but the seven fans that came with this Cube case (I love ebay) are all 120mm Panaflo brushless fans that use the the Hydro Wave bearings you describe. I can attest that even once I mute the airflow noise by undervolting them to around 9v (I have not measured exact voltages at specific controller settings, so I'm ballparking my figures), the motors develop a growl, two of them growl loud compared to the rest while a third has a very annoying clicking but doesn't growl as much.

You are correct. That is a trait in the older Panasonic fans. When operated from resistor controllers, their motors started complaining at about 10 volts. It's actually current starvation not undervoltage that's causing the problem. Their rotation becomes erratic causing all kinds of blade dynamics.

Your small fix to apply to these brushless fans is interesting, currently I use a Sunbeam fan controller as my Vantec fan controller didn't have sufficient channels even though I consider it of better construction.

Ok... now to loose friends and alienate people :D .... Most fan controllers made for consumer/PC use are garbage. Usually they are nothing but variable resistors that leave the fan motor starved for current. Even the more expensive ones tend to be little more than voltage regulators that can and do act just like resistors.

The good ones, like you see in industrial settings, will operate in reverse to what you see in consumer gear... they don't start slow and build up speed... they start at full speed for a few seconds then reduce speed as the air cools. This never leaves a fan starved for startup current.

Is it only the voltage regulation in the power supply that is at issue here, or are there other considerations? I can take a multimeter and check on that easily with how these fans were built.

You would need an oscilloscope to see the power supply issue. It happens way too fast for a meter to catch. The capacitor in my little drawing stores enough charge to fill in any narrow voltage spikes and any back pulses from the fan are absorbed and stored as energy available to the fan when it needs it. In normal operation, the diode really just keeps the capacitor charged. However during start up it does not starve the fan for current the way resistor solutions do. Essentially it's a miniature power supply filter to keep switching noise off the power leads of the fan... which usually stops the motor from complaining about current starvation.

I've had plenty of experience with sleeve bearing and 1 & 2 ball bearing fans in my short time in the PC world. I'll never buy a sleeve bearing fan, as while quieter than ball-bearings they wear out in no time flat. Ball bearings last longer, but I've worn out enough Antec fans (And had enough noise issues with them) that I won't buy those or Antec fans ever again either.

You might as well foget Sunon while you're at it... I've never heard such noisey fans!
 

Rob Williams

Editor-in-Chief
Staff member
Moderator
Kougar said:
That's why I watercool. Every case fan is undervolted to 5-7v or is a Noctua NF-P12. Nothing like running a brutally overclocked high-end system with near silence... I could make it completely silent but due to the roar of the AC unit there's no use in going that far. As it is the case is right next to me, but I don't hear it.

If I ever get down there to Texas, you are going to have to prove this one to me ;-)

2Tired2Tango said:
And if your machine is stolen or there's a system killing event?

Even a NAS box isn't going to protect you in that case (why wouldn't it get stolen also?). I regularly back up all my important documents to DVD, in case both my PC and NAS box happened to go up in smoke.

2Tired2Tango said:
I'd rather sacrifice a bit of performance than lose all the family pictures, financial records and accumulated media files.... Worse still if there are business records involved.

I didn't try to suggest that network storage was a bad idea, because I rely on it myself. I keep two copies of anything important on my own PC (across two HDDs) and another on the NAS box. I said I'd rather keep things local for the sake of performance, and reliability. To each their own, really.

2Tired2Tango said:
True enough. If nothing else, stick one in your closet and put in a mess of raid drives and use it as a backup device

I think ideally, it'd be even cooler to keep these PCs in a completely different part of the house, like in the basement, where it's cool and out of the way. EVGA's PCoIP products have piqued my interest, but their cost is far, far too high at this point in the game. It'd be cool to sit down in front of your ultra-high-end gaming machine and all there be is a monitor, peripherals, and some small box connected to your router.

As for the rest of the posts here, just wow. You guys are providing some invaluable information here. I just need to learn how to retain it all...

/me throws his Delta 3700RPM/141.96CFM fans in the closet.
 

2Tired2Tango

Tech Monkey
Even a NAS box isn't going to protect you in that case (why wouldn't it get stolen also?).

Because it's sitting in a locked closet where nobody even knows you have it?
I know... What's to stop them from breaking into the closet... Well, it's better than nothng.

As long as you have a backup strategy that promises not to fail you, you're miles ahead of most people who just can't envision their 6 year old computer breaking down....


As for the rest of the posts here, just wow. You guys are providing some invaluable information here. I just need to learn how to retain it all...

/me throws his Delta 3700RPM/141.96CFM fans in the closet.

:D :D :D Just another day on the farm.
 

Rob Williams

Editor-in-Chief
Staff member
Moderator
2Tired2Tango said:
As long as you have a backup strategy that promises not to fail you, you're miles ahead of most people who just can't envision their 6 year old computer breaking down....

It's sad because that's true. There are <em>many</em> who don't even consider backing up, and I'm confident everyone is going to run into a bad issue where they highly regret not taking it more seriously. Of course, it took me about five data losses before I really began to take it seriously. Now I'm super-mega-ultra-hardcore backuper guy.
 
Top