ATX vs mATX

[JackCaver]

Hi,

I'm considering building a fast mATX system as I don't really need many PCI slots except maybe 1 for a sound card, and I'm not really bothered about overclocking.
I just wanted to know what are the main differences between ATX and mATX? Is it just overclocking, expandability and SLI/Xfire? Or do ATX boards generally run faster than mATX boards?
I'm considering a Foxconn G9657MA with an E6600 and 7900GS.

Thanks

 

[tcsenter]

Or do ATX boards generally run faster than mATX boards?

There is no fundamental difference between ATX and mATX at all, except footprint or real estate.

The market has dictated that mATX boards are generally not desirable or preferred as enthusiast or performance segment boards. If ATX boards on average are 'faster' than mATX boards, its because vastly more ATX boards are designed and marketed for the enthusiast or performance segment.

As such, its an unusual mATX board that offers much in the way of overclocking features. Many offer very limited or incomplete overclocking features (e.g. frequency adjustment with no voltage adjustment), but not the comprehensive adjustment and tweaking that can be found enthusiast boards (virtually all of which are ATX).

Personally, I prefer mATX because I prefer smaller cases. I can't remember the last time I used more than two 5.25" and two 3.5" external bays, or more than two PCI slots for that matter.

 

[PClark99]

currently if you want core2duo and matx I think the best board is the Asus P5B-VM.

Thats what I run in my matx rig.

If you can wait another couple weeks Abit is coming out with a couple boards that might be better.

So far there is no real overclocking matx core2 board.

If you want matx and overclock you either need socket 939 AMD, DFI RS482 or Biostar TForce, or AM2, Abit Nf2 nView.

 

[Zap]

There is no difference beyond what is artificially created by manufacturers. Certain chipsets are almost always found on mATX boards while other ones are almost never found on mATX boards.

All else being equal, ATX and mATX are identical except mATX is smaller and thus lacks the PCB room for a few extra slots. ATX can have up to 7 (just count the possible PCI/PCI-E slot openings on the back of your case) and mATX can have up to 4. That is the whole difference between the form factors.

Once manufacturers get done with it, the market has mATX boards with onboard video and almost no overclocking (sans a few special boards). Does it have to be that way? No. Will it continue to be that way? Probably.

All hope is not lost however, because Abit just announced a new mATX "home theater and gaming" board for Core 2 Duo. It uses the new ATI chipset with integrated video (supposedly x700 level). No word on overclocking, but Abit is definately showing that they aren't ignoring this niche.

 

[MrNeutrino]

Originally posted by: tcsenter
If ATX boards on average are 'faster' than mATX boards, its because vastly more ATX boards are designed and marketed for the enthusiast or performance segment.

Anyone have links to actual benchmarks between mATX and ATX boards for the same chipset family for non-gaming tasks?

E.g. Core 2 Duo + P965 / ATX board like Asus P5B-E vs. Core 2 Duo + G965 / mATX board like Asus P5B-VM.

Not all of us care about OC or gaming, yet there are many other intensive tasks where people want to squeeze out as much performance as possible. E.g. encoding, image manipulation, rendering, compression, and other such long running tasks.

I for one would like to see for example, how sharing the memory bandwidth with the integrated graphics' shared VGA memory (= main memory), affects performance for such non-gaming tasks. That of course, assumes you have the integrated graphics enabled (something many HTPC folks may want to do).

Unless I'm missing something, I have yet to see any such review comparing the non-gaming performances of the mATX (G965) vs. ATX (P965 etc.) C2D MB reviews here at AT! Simply adding an mATX MB in the mix during the many C2D ATX MB reviews published over the past half a year or more would have done the job...

Hopefully we'll see more mATX MB reviews here at AT, soon!

 

[Gary Key]

We have neglected the mATX market, not on purpose, but there always seemed to be the latest and greatest performance board coming out. We have close to a dozen mATX boards in testing now.

In answer to your concern, exclude any overclocking aspects of the boards, the difference between the G965 and P965 is about 2% on average without a discrete video card solution installed in the G965. The difference with one installed is negligible if all other settings are equal. The issue is that most G965 boards do not offer extensive memory controller options and so the ability to tweak the memory is not available and the memory subtimings are usually run a lot looser.
Overall, I have not seen more than a 5% difference in performance between a G965 board at 5-5-5-15 and a P965 board at 3-4-3-9 at DDR2-800 in media applications. We attribute most of the difference to the fact that you are reducing the available memory pool for the OS to utilize with the integrated graphics solution.
We are finding under Vista that memory timings on the IGP solutions are very important, especially on the G965 boards when using the on-board video and the memory reduction footprint is even more critical when considering Vista's memory requirements.

 

[Zap]

Originally posted by: Gary Key
We have neglected the mATX market, not on purpose, but there always seemed to be the latest and greatest performance board coming out. We have close to a dozen mATX boards in testing now.

WOOT!!!!

Originally posted by: Gary Key
In answer to your concern, exclude any overclocking aspects of the boards, the difference between the G965 and P965 is about 2% on average without a discrete video card solution installed in the G965. The difference with one installed is negligible if all other settings are equal. The issue is that most G965 boards do not offer extensive memory controller options and so the ability to tweak the memory is not available and the memory subtimings are usually run a lot looser.

Translation: If using the same parts and same BIOS settings, no difference in performance between mATX and ATX. That should be expected because the "difference" is "make it smaller." Nothing in the mATX "standard" states "make it smaller and 5% slower."

 

[MrNeutrino]

Originally posted by: Zap

WOOT!!!!

See, Gary, here's yet another 'audience member' I was referring to.

Originally posted by: Zap
Translation: If using the same parts and same BIOS settings, no difference in performance between mATX and ATX. That should be expected because the "difference" is "make it smaller." Nothing in the mATX "standard" states "make it smaller and 5% slower."

Thanks for the English -> English 'translation', Zap. :roll:

If you read through the above comments, it doesn't sound like there is 'no difference' between mATX / ATX performance (whether intentional on the vendor's part or not), especially with Windows Vista (though even without Vista). As the >0 % figures suggest in Gary's response, there is some difference. That said, I completely agree. In real-world situations, this amounts to there being no noticeable performance difference.

BTW, I presume you were joking / being sarcastic. Of course, I fully well realize that no standards committee in their right mind would propose 'dropping' performance just because of a 2.4" shorter board in one dimension! However I don't consider that argument justification enough to 'prove' that mATX running with an IGP chipset is faster / slower than it's non-IGP counterpart - at least a slightly different chipset and system architecture.

As such, my questions above are perfectly valid architecturally and worth verifying via benchmarks - sharing even the hefty dual-channel DDR2 memory bandwidth between a latency sensitive, cache starved GPU and a dual-core, high IPC CPU (need I say latency sensitive again?) can be argued as being an architectural recipe for disaster (ok, exaggeration - performance drop at the very least) compared to discrete graphics card solution.

Lack of 'actual' benchmark backed data suggesting / proving this had been unavailable to-date. Fed up with lack of actual benchmark data pointers from such posts as these in the forums, it led me to escalate this through the comments section of the recently posted 690G review and make it visible to the editors (linked above).

In short, I feel better hearing Gary's feedback based on (I presume) actual / quantifiable benchmarks vs. valid yet simply logical deductions... was simply looking for concrete (read: benchmark) data.

Now I can rest relatively happy.

 

[gorka]

This board would be nice with an NSK3400 (the new revision with the normal ATX PSU), though you may want to change the PSU.

http://www.abit-usa.com/products/mb/products.php?categories=1&model=338

but definitely...you can NOW game and OC with mATX.

 

[Zap]

Originally posted by: MrNeutrino

Originally posted by: Zap

WOOT!!!!

See, Gary, here's yet another 'audience member' I was referring to.

I've been an "audience member" for many years. I can prove at least to 2002.

Lack of decent micro ATX motherboards

Originally posted by: MrNeutrino

Originally posted by: Zap
Translation: If using the same parts and same BIOS settings, no difference in performance between mATX and ATX. That should be expected because the "difference" is "make it smaller." Nothing in the mATX "standard" states "make it smaller and 5% slower."

If you read through the above comments, it doesn't sound like there is 'no difference' between mATX / ATX performance (whether intentional on the vendor's part or not),

BTW, I presume you were joking / being sarcastic.

As such, my questions above are perfectly valid architecturally and worth verifying via benchmarks

I absolutely agree with you that it is worth verifying via benchmarks, and yes I was being sarcastic with stating the "standards." That's because too many times I see people blindly posting that mATX is slower. Maybe it is, maybe it isn't, but it seems to be one of those computer urban legends that just won't die. Maybe Gary can fix that for us.

My "no difference" means using a similar or identical chipset with all the same parts (including discrete graphics) and all the same BIOS settings down to RAM timings. In those conditions I'd say that there's no difference. Motherboard manufacturers have made ATX boards with IGP chipsets and mATX boards with non-IGP chipsets, so this can be tested.

Perhaps what you really meant was a difference between IGP chipsets and non-IGP chipsets? I still maintain that there will be little to no difference if again all parts (down to the video card) and BIOS settings are identical and the chipsets are of a similar family. For instance, Geforce6100 and Nforce4, or Q965 and P965.

We shall see, maybe...

Originally posted by: gorka
This board would be nice with an NSK3400 (the new revision with the normal ATX PSU), though you may want to change the PSU.

http://www.abit-usa.com/products/mb/products.php?categories=1&model=338

but definitely...you can NOW game and OC with mATX.

Uh, you probably meant the Fatal1ty F-I90HD, not the one you linked which is a regular ATX board using the Nvidia 650i chipset with SLI (with some of the "features" accidentally copy/pasted over from the other board).

BTW, I don't know what hole you've been living in, but I've been gaming on mATX based systems for years, and even some overclocking. Here's some of the stuff that I've had or still have:

MicroFly case with SLI

mATX Tforce6100 being tested...
... and overclocked to 315MHz HTT

mATX Tforce6100-AM2...
... able to do 330MHz HTT

Two of my SFF mATX gaming rigs with a Tforce6100-939 board in the background

My regular LAN gaming rig now updated to an Athlon 64 San Diego core overclocked to 3.12GHz and with a 7900GTO, all in a 13.5"x14.5"x5.75" 15 pound package.

Even without manufacturer support (as in BIOS settings) I used to pin mod and "wire trick" boards/CPUs. These days I just wait for decent boards to show up, such as my Abit NF-M2 nView (no pics though) which I've had running at 360MHz+ HTT.

 

[Renoir]

This subject interests me also. The review of the 690G over at Bit-tech.net has some benchmarks for the 690G comparing the integrated graphics vs an x1300 in non gaming benchmarks. Turns out using 2 gig of ram and having the 690G share 256MB results in about a 0-6% performance drop vs the discrete card. Only problem is they were using XP.

Hope that helps

 

[wwswimming]

i would like to see the Asus 8 phase power processing circuitry,
or the Gigabyte 12 phase, used on a micro-ATX board.

which micro-ATX board has the best power processing
circuitry ?

 

[MrNeutrino]

Originally posted by: Zap
Maybe Gary can fix that for us.

Looks like he just did. Yep, benchmarks typically have that effect, don't they?

BTW, yes, in all my comments about mATX vs. ATX, I have been implicitly linking mATX to IGP chipsets and ATX to their non-IGP counterparts. After all, bar the few Shuttle XPC SD37P2 V2 et al type SFF MBs, most do (sadly) use IGP chipsets. Yep, should have worded that part more clearly, besides just in the sections where I make reference to chipset specific performance comparisons w.r.t. memory bandwidth.

Anyway, looking forward to a mATX build real soon!

 

[magreen]

I'm still wondering about the desktop slowdown with shared memory from using integrated graphics.

I've heard and it makes sense logically (wish I had benchmark data to rely on) that there should be ~400 MB/s lost in memory bandwidth due to shared memory. That's 1280x1024 pixels * 4 bytes/pixel * 75 Hz = 375 MB/s.

So with modern dual chan. ddr2 systems with maybe 4000 MB/s bandwidth, we're talking max 10% bandwidth lost.

But I wonder if there's a more intangible factor, the "snappiness" factor when switching between apps and the like, that you lose out on with shared memory, especially with a dual core.

Maybe I'm just biased by my home pc with a 1.8GHz P4A Northwood @ 2.0 GHz... with 100 MHz SDR... it had 400 MB/s memory bandwidth on integrated graphics (a bit like self-flagellation) that jumped to 790 MB/s with discrete graphics... I'd just really like to see something concrete regarding latencies, bandwidth, and delays when switching between/loading apps with shared mem.

 

[GEOrifle]

What about ASUS P5B-VM DO LGA 775 Intel Q965 Express Micro ATX Intel Motherboard ?

http://www.newegg.com/Product/Product.asp?Item=N82E16813131085

Supported CPU
CPU Socket Type LGA 775
CPU Type Quad-core / Core 2 Extreme / Core 2 Duo / Pentium
FSB 1066/800MHz
Supported CPU Technologies Hyper-Threading Technology
Chipsets
North Bridge Intel Q965 Express
South Bridge Intel ICH8DO
Memory
Number of Memory Slots 4×240pin
Memory Standard DDR2 800
Maximum Memory Supported 8GB
Dual Channel Supported Yes
Expansion Slots
PCI Express x16 1
PCI Express x4 1
PCI Slots 2
Storage Devices
PATA 1 x ATA100 up to 2 Devices
SATA 3Gb/s 7 x Internal SATA 3.0 Gb/s ports
1 x External SATA 3.0 Gb/s port (SATA On-the-Go)
SATA RAID 0/1/0+1/5
Onboard Video
Onboard Video Chipset Intel GMA 3000
Onboard Audio
Audio Chipset ADI AD1988
Audio Channels 8 Channels
Onboard LAN
LAN Chipset Intel 82566DM
Max LAN Speed 10/100/1000Mbps
Rear Panel Ports
PS/2 2
LPT 1
Video Ports D-Sub
USB 4 x USB 2.0
IEEE 1394 1 x IEEE 1394a
Audio Ports 6 Ports
Onboard USB
Onboard USB 3 x USB 2.0 connectors support 6 ports
Onboard 1394
Onboard 1394 1 x 1394a connector supports 1 port
Physical Spec
Form Factor Micro ATX
Dimensions 9.6" x 9.6"
Features
Power Pin 24 Pin
Features Notice: Only DDR2-800 memory supporting JEDEC approved 1.8V operation with timings of 5-5-5 or 6-6-6 is supported on Intel Desktop Boards based on Intel 965 Express Chipsets.

I'm also loocking for mATX gaming board with C2D, more C2Q support.
But thise board doesn't has any recommendations at NEWEGG, strange!!!

 

[Texun]

I have had the same question. Recently, I built a few basic rigs for myself and friends using mATX boards and they have since made me think twice about going full ATX for my next main rig. I don't need all the extra slots and the small footprint makes handling and installing them a breeze. I've used MSI, ECS, Chinatech and Gigabyte and they perform as well as my ATX (with the exception of CPU and video card variations). What I am trying to say is that if I had the same speed cpu and video in each one I wouldn't know which was which. One glaring omission would be the BIOS limitations in the budget boards I used... but then the BIOS in my ASUS isn't as feature rich as the trusty Abit that I retired a while back.

With regard to how long they will last is another question. I usually prefer boards with good power management\conditioning, but so far they have been quite stable.

 

[MrNeutrino]

Originally posted by: magreen
I'm still wondering about the desktop slowdown with shared memory from using integrated graphics.

I've heard and it makes sense logically (wish I had benchmark data to rely on) that there should be ~400 MB/s lost in memory bandwidth due to shared memory. That's 1280x1024 pixels * 4 bytes/pixel * 75 Hz = 375 MB/s.

So with modern dual chan. ddr2 systems with maybe 4000 MB/s bandwidth, we're talking max 10% bandwidth lost.

But I wonder if there's a more intangible factor, the "snappiness" factor when switching between apps and the like, that you lose out on with shared memory, especially with a dual core.

Maybe I'm just biased by my home pc with a 1.8GHz P4A Northwood @ 2.0 GHz... with 100 MHz SDR... it had 400 MB/s memory bandwidth on integrated graphics (a bit like self-flagellation) that jumped to 790 MB/s with discrete graphics... I'd just really like to see something concrete regarding latencies, bandwidth, and delays when switching between/loading apps with shared mem.

Glad to see yet another person share the same question I raise above. I don't think this was addressed in Gary's recent mATX performance update review. While we wait for his mATX roundup review where I hope to see this get addressed, here's Gary's response to my question above:

Originally posted by: Gary Key
In answer to your concern, exclude any overclocking aspects of the boards, the difference between the G965 and P965 is about 2% on average without a discrete video card solution installed in the G965. The difference with one installed is negligible if all other settings are equal.
...
Overall, I have not seen more than a 5% difference in performance between a G965 board at 5-5-5-15 and a P965 board at 3-4-3-9 at DDR2-800 in media applications.

-----------------

Originally posted by: GEOrifle
What about ASUS P5B-VM DO LGA 775 Intel Q965 Express Micro ATX Intel Motherboard ?

GEOrifle, that MB is based on the Q965 chipset vs. the G965. If I recall correctly, the former is oriented toward business / corporate environment vs. toward the budget / SFF enthusiests in the case of G965... may be way off base here but that's my recollection on the chipset version I otherwise didn't care about / care to learn much about, without doing more research right now.

I'd recommend getting the P5B-VM (G965) MB if you want a similar one... that's what most folks tend to get... hence no reviews doesn't surprise me.

 

[GEOrifle]

P5B-VM (G965) got BAD reviews too.
Then recommend better mATX, nothing more?

 

[Steeeeve]

Lets not forget that these things might vary on a per-board basis.

The F-I90HD is a pretty nice one

 

[GPett]

I sure would like a Gigabyte G 965P DS3 type of board in a microATX size. Here is hoping one of the manufacturers makes a better micro atx board for the next generation of chips coming out.