Page 4 of 4

Boot Drive

Posted: Wed Jan 13, 2010 7:08 am
by MtnHermit
matt_garman wrote:Boot drive: 8 GB Compact flash via PATA-to-CF adapter
Curious about your boot drive. I tried the same thing with a SATA to CF adapter using a Kingston 133X 16GB CF. It took ~3 hours to load XP SP2 on the card and I was never successful booting from it. Any suggestions or comments?

BTW: Great article, you set a high bar. :)

Re: Boot Drive

Posted: Wed Jan 13, 2010 1:22 pm
by matt_garman
MtnHermit wrote:
matt_garman wrote:Boot drive: 8 GB Compact flash via PATA-to-CF adapter
Curious about your boot drive. I tried the same thing with a SATA to CF adapter using a Kingston 133X 16GB CF. It took ~3 hours to load XP SP2 on the card and I was never successful booting from it. Any suggestions or comments?
I've only used this PATA to CF adapter. In fact I have and used at least a couple of these. But I've never had any trouble with them---I just insert the CF card, connect the PATA cable, and provide power via the floppy connector, and everything "just works".

However, they are fairly slow. All my home computers run some open source unix (Linux or OpenBSD), and particularly with things like NAS boxes and firewall/router boxes, I go with a fairly barebones/minimal install. But the three hour install of XP doesn't surprise me too much. If you can, check the DMA settings of your adapter, and also how the OS recognizes the "drive". Going from memory, some of the cheaper Syba PATA-to-CF adapters don't support any form of DMA; likewise, I think some CF cards don't have DMA support. If both the adapter and card don't support (U)DMA, you'll be forced to use PIO mode, which is dreadfully slow.

I don't have any ideas on the boot failure. Most BIOSes that I've used show at least some information about attached drives. Same goes for my CF "drives". So, can you see the CF drive in your BIOS setup screen? Or can you see the BIOS finding it during its POST process? It seems unlikely that Windows could find it in order to do the install, but your BIOS can't find it to boot from.

You might want to use a system rescue CD or Linux boot CD to use a tool to look at the partition table (e.g. fdisk). There is a "bootable" flag that can be set on a partition. I know I sometimes used to forget to set this on "manual" install Linux distros (like Gentoo), although it's automatic on "easy" install distros (like Ubuntu) and should certainly be automatic with Windows. But it's worth checking either way.

Finally, I did kill at least a CF card, and maybe the adapter once by accidentally shorting part of the adapter on the case. I don't know what happened, but there was a quick, brief, small flash, a little bit of a smoke, and the compact flash card was barely touchable (extremely hot). The card was definitely dead; I didn't want to risk killing another card with the adapter so I tossed it. I'm not sure whose fault that was, but now I always make sure the adapter rests on a non-conductive surface.

Posted: Fri Jan 15, 2010 6:21 am
by MiKeLezZ
A C.F. is basically a PATA SSD.

A C.F. PATA adapter does only link the C.F. pins to the PATA pins and adds a power connector for powering up the C.F.

So the speed only depends on the quality of your C.F. - I have a Kingston Ultimate 266x that does 39MB/s write and 46MB/s read (clearly S.L.C. NAND Flash in here), the XP install took only some minutes.

One advice, when using these PATA adapter, you have to check if the C.F. has a non-removable-device flag since it could be a problem with Microsoft OSes (but not linux).

Posted: Fri Jan 15, 2010 6:47 am
by matt_garman
MiKeLezZ wrote:A C.F. PATA adapter does only link the C.F. pins to the PATA pins and adds a power connector for powering up the C.F.

So the speed only depends on the quality of your C.F.
Are you sure that's completely true? Honest question, not trying to be argumentative. My understanding is that both the CF and the adapter have to support (Ultra) DMA. There are different levels of DMA defined within the PATA spec---if you find an old enough motherboard, for example, it won't support the newest DMA modes. I'm guessing that the increasing levels of DMA require at least some extra degree of sophistication in wiring. One obvious example is when the cables went from 40 wires to 80 wires (although the actual number of pins stayed the same).

The Wikipedia page on CompactFlash doesn't say explicitly, but seems to support your idea. It's entirely possible I had it all wrong!

Posted: Fri Jan 15, 2010 7:08 am
by MiKeLezZ
matt_garman wrote:
MiKeLezZ wrote:A C.F. PATA adapter does only link the C.F. pins to the PATA pins and adds a power connector for powering up the C.F.

So the speed only depends on the quality of your C.F.
Are you sure that's completely true?
A PATA C.F. Adapter is a PASSIVE adapter - think it as a 2'5" IDE to 3,5" IDE adapter. It only does a mechanical conversion and has not any chip for converting the signal (it would be easier to explain if you ever see one). So the speed only depends on your C.F. (and not every C.F. does support UDMA, for istance, as you saw in wikipedia, and not every memory is egual on speed).

A SATA C.F. Adapter is instead an ACTIVE adapter - think it as a SATA to PATA adapter. It does a mechanical AND a signal conversion. So the speed depends on your C.F. AND on the chip on the adapter, that is needed for converting PATA signals to SATA ones.

Posted: Mon Jan 18, 2010 3:48 pm
by zodaex
MiKeLezZ wrote:
matt_garman wrote:
MiKeLezZ wrote:A C.F. PATA adapter does only link the C.F. pins to the PATA pins and adds a power connector for powering up the C.F.

So the speed only depends on the quality of your C.F.
Are you sure that's completely true?
A PATA C.F. Adapter is a PASSIVE adapter - think it as a 2'5" IDE to 3,5" IDE adapter. It only does a mechanical conversion and has not any chip for converting the signal (it would be easier to explain if you ever see one). So the speed only depends on your C.F. (and not every C.F. does support UDMA, for istance, as you saw in wikipedia, and not every memory is egual on speed).

A SATA C.F. Adapter is instead an ACTIVE adapter - think it as a SATA to PATA adapter. It does a mechanical AND a signal conversion. So the speed depends on your C.F. AND on the chip on the adapter, that is needed for converting PATA signals to SATA ones.
I know there is a model of syba cf to ide adapter that only supports pio mode, and not DMA. So it is not solely dependent on the flash card, but also it matters for your cf to ide adapter to support at least DMA-66 speeds or it can bottleneck your card, much like if you hooked up a hard drive to a motherboard port that only supported PAE.

Posted: Mon Jan 18, 2010 4:31 pm
by MiKeLezZ
zodaex wrote:
MiKeLezZ wrote:
matt_garman wrote: Are you sure that's completely true?
A PATA C.F. Adapter is a PASSIVE adapter - think it as a 2'5" IDE to 3,5" IDE adapter. It only does a mechanical conversion and has not any chip for converting the signal (it would be easier to explain if you ever see one). So the speed only depends on your C.F. (and not every C.F. does support UDMA, for istance, as you saw in wikipedia, and not every memory is egual on speed).

A SATA C.F. Adapter is instead an ACTIVE adapter - think it as a SATA to PATA adapter. It does a mechanical AND a signal conversion. So the speed depends on your C.F. AND on the chip on the adapter, that is needed for converting PATA signals to SATA ones.
I know there is a model of syba cf to ide adapter that only supports pio mode, and not DMA. So it it not solely dependent on the flash card, but also it matters for your cf to ide adapter to support at least DMA-66 speeds or it can bottleneck your card, much like if you hooked up a hard drive to a motherboard port that only supported PAE.
no.

Posted: Mon Jan 18, 2010 9:09 pm
by zodaex
MiKeLezZ wrote:
zodaex wrote:
MiKeLezZ wrote:A PATA C.F. Adapter is a PASSIVE adapter - think it as a 2'5" IDE to 3,5" IDE adapter. It only does a mechanical conversion and has not any chip for converting the signal (it would be easier to explain if you ever see one). So the speed only depends on your C.F. (and not every C.F. does support UDMA, for istance, as you saw in wikipedia, and not every memory is egual on speed).

A SATA C.F. Adapter is instead an ACTIVE adapter - think it as a SATA to PATA adapter. It does a mechanical AND a signal conversion. So the speed depends on your C.F. AND on the chip on the adapter, that is needed for converting PATA signals to SATA ones.
I know there is a model of syba cf to ide adapter that only supports pio mode, and not DMA. So it it not solely dependent on the flash card, but also it matters for your cf to ide adapter to support at least DMA-66 speeds or it can bottleneck your card, much like if you hooked up a hard drive to a motherboard port that only supported PAE.
no.
http://omgili.com/reviews/SD@2dCF@2dIDE ... @2d-Retail

Yes, oh god, yes!

Posted: Tue Jan 19, 2010 3:39 pm
by zodaex
Lol, you got owned big time MiKeLezZ!

LMAO!! :lol: :lol: :lol: :lol:

Posted: Fri Jan 22, 2010 1:10 am
by stevejones
Hi Matt,

I'm still confused with last point you've mentioned.

Can you please come again with specification.

Thanks.

Posted: Fri Jan 22, 2010 5:12 am
by matt_garman
stevejones wrote:I'm still confused with last point you've mentioned.
To which point are you referring?

Posted: Sun Jan 31, 2010 8:05 am
by MoJo
Apologies for such a late reply.
MikeC wrote:The KaW always gave lower power (W) readings compared to the Extech -- 92W vs 97.3W, 71-72W vs. 75.4W, 33-34W vs. 36.4W, 13W vs. 15W. It will be pulled from active duty in the lab.

...

The second SPA was close to the first, but slightly less consistent, with PF readings slightly lower sometimes, and power reading also slightly lower sometimes. This one will also be retired.
Any chance you could take a look inside those two, particularly the KaW? It would be helpful to know what IC they use. most of these things are basically one IC and a high power resistor.

By knowing what IC is in use we could try to identify the better ones. If you look at the colour codes on the resistor you can tell what the tolerance is. Standard ones are only about 10% accurate so that could account for all the variations you are seeing. It would also mean that the accuracy could be improved simply by replacing the resistor, or even just picking the most accurate resistor from a selection of meters.

I have a few here that I will open up and test. Unfortunately they are all the same make, but could still be useful for checking tolerances.

Posted: Sun Jan 31, 2010 9:33 am
by zodaex
MiKeLezZ has no clue what he's talking about! HAHHAHANAAHAHHAHAHHAHAHAHAHHAAAA!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
SO PWNED!!!!!!

Posted: Wed Feb 03, 2010 5:25 pm
by kaotikfunk
Has anyone looked into the Asus M3A76-CM yet? I'm considering it over the Biostar TA760G recommended here because:

- it's cheaper than the Biostar at Frys locally or Newegg online.
- Kingston already listed DDR-2 800 ECC ram compatible with this board
- Same feature set as TA760 (4 dimms, 6 SATA, Realtek 8111c)
- I trust Asus build quality over Biostar. Feel free to correct me, I've never killed an Asus board but never owned a Biostar. Newegg comments have more DOAs than I would expect.


The full build I'm considering for a virtualization/file server is:
(existing parts)
3x WD Green Power & Samsung F2 drives
Antec Earthwatts 380

(to buy)
2x 2G Kingston DDR2-800 ECC (add 2x 2G more later if needed)
Athlon II X2 240 or X4 620 undervolted
AMD 760g motherboard w/ 4 DIMMs and >= 4 SATA

Posted: Thu Feb 04, 2010 11:07 am
by matt_garman
kaotikfunk wrote:Has anyone looked into the Asus M3A76-CM yet?
I have no personal experience with it, but one comment from looking at the pictures of it on newegg: it appears it has a four (maybe five) phase power supply (VRM). This probably uses more power than the Biostar's three-phase VRM. Because of the Biostar's wimpy VRM, it only supports CPUs with a TDP of 95W or less. Some consider this a limitation; I consider it a feature as it generally means lower power consumption. On the other hand, we're probably talking at most only a few watts here. Plus, it's conceivable that the Asus's bigger VRM is actually engineered well enough to have an efficiency that matches the Biostar's smaller VRM.

So, just to be clear, this is all pure speculation on my part. :)

kaotikfunk wrote:I'm considering it over the Biostar TA760G recommended here because:

- it's cheaper than the Biostar at Frys locally or Newegg online.
- Kingston already listed DDR-2 800 ECC ram compatible with this board
- Same feature set as TA760 (4 dimms, 6 SATA, Realtek 8111c)
Can't argue with any of that! It's nice to see the not-too-bad Realtek NIC in that Asus board. Other low-end Asus boards I've looked at in the past used the (worse) Atheros NIC chips. It also trumps the one big complaint I have with the Biostar board, and that's the "unofficial" ECC support. It seems like it turns a lot of people off.
kaotikfunk wrote:- I trust Asus build quality over Biostar. Feel free to correct me, I've never killed an Asus board but never owned a Biostar. Newegg comments have more DOAs than I would expect.
In the past, I've generally used Biostar boards for cheap projects I don't care about, and a "tier 1" manufacturer (generally Asus, but Gigabyte recently) for my more important projects (where I'm willing to spend more). Since it's cheap, I treat the Biostar stuff like I stole it, while handling the expensive stuff with kid gloves. Yet I've never had a problem with Biostar, but have had some flaky Asus (and other more expensive) products.

Anyway, that's just anecdotal info on one hobbyist's non-statistically significant samples. :)

Posted: Thu Feb 04, 2010 4:35 pm
by MoJo
A little update. I have stripped down a couple of the energy meters (Kill-A-Watt clones) and sure enough they use cheap 5% resistors. I am going to get hold of some 1% ones to try.

Hopefully they will improve the accuracy quite a lot.

Posted: Wed Feb 10, 2010 5:51 am
by zif
kaotikfunk wrote:Has anyone looked into the Asus M3A76-CM yet? I'm considering it over the Biostar TA760G recommended here because:
- Kingston already listed DDR-2 800 ECC ram compatible with this board
I'm using the M3A76-CM with 4GB of Kingston DDR2-800 ECC as a file/backup server running EON 0.59.9.

Other than the fact that my Syba CF-IDE adapter (which they claim has UDMA support) and Transcend CF card (which they also claim has UDMA support) won't work in UDMA mode, everything's running perfectly.

Oh, wait, one of my 3 Samsung HD154UI 1.5TB HDDs has some errors already too. Dang it.

The board and RAM are perfect though! :-)

I haven't done any power measurements or anything like that, but if you have any other questions...

Max

Posted: Tue Apr 13, 2010 10:31 am
by kaotikfunk
kaotikfunk wrote:Has anyone looked into the Asus M3A76-CM yet? I'm considering it over the Biostar TA760G recommended here because:

- it's cheaper than the Biostar at Frys locally or Newegg online.
- Kingston already listed DDR-2 800 ECC ram compatible with this board
- Same feature set as TA760 (4 dimms, 6 SATA, Realtek 8111c)
- I trust Asus build quality over Biostar. Feel free to correct me, I've never killed an Asus board but never owned a Biostar. Newegg comments have more DOAs than I would expect.


The full build I'm considering for a virtualization/file server is:
(existing parts)
3x WD Green Power & Samsung F2 drives
Antec Earthwatts 380

(to buy)
2x 2G Kingston DDR2-800 ECC (add 2x 2G more later if needed)
Athlon II X2 240 or X4 620 undervolted
AMD 760g motherboard w/ 4 DIMMs and >= 4 SATA
Stock of 760 boards were depleting, so I decided to go with this setup:
Asus M4A785M
Athlon x3 440 (Newegg deal made it same price as 240, couldn't pass it up)
2x 2G Kingston DDR2-800 CL5 ECC
1x WD Caviar Blue 500G (boot/VHD's)
1x WD Green 1TB (passthru for WHS)
1x Samsung F2 1TB (passthru for WHS)
1x recycled PATA DVD drive
Windows Server 2008 R2 (Core Install)
VM's: Windows Home Server running 24/7, other OS's used for testing and game servers but not 24/7

Net result: 63w idle without undervolting. BIOS is up around 75w, but it drops to 63w once Windows has booted. Heavy WHS activity such as file copies or streaming videos bumps up to around 68w.

With performance tracing, I was able to verify that 2 cores stay in a parked state the majority of the time, and the remaining core is spending most time in C1e.

I haven't verified which drives are being spun down at idle. That might shave a few watts if its possible to spin down WHS storage volumes.

Re: Biostar A760G-M2+ - 30W barrier broken

Posted: Mon Jun 14, 2010 2:56 am
by mgiammarco
matt_garman wrote: Enter the Biostar A760G-M2+. This motherboard caught my eye, because it appears to address the issues I have with the Gigabyte board:
  • Three phase VRM (can't use high-power CPUs over 95W)
  • Explicit ECC options in the BIOS
  • EDIT: both PCIe slots can be used with non-video cards (see follow up below)
Are there other Biostar motheboards with the same features above?

Thanks,
Mario

No POST, No Beep, No video - DOA cpu?

Posted: Sat Jun 19, 2010 2:03 am
by soobaerodude
Wanting similar goals as Matt, I decided to purchase the TA760G and an Athlon II X2 240e.

After plugging everything in, I powered the system on, and it did not post, there was no beep, and no video.

I exchanged the motherboard for another TA760G from Fry's, and followed the Bare Bones Boot procedure. Basically, the only things connected to the board were the 240e, the case speaker, and power. No RAM, no video, no keyboard. Powered up the system, and silence. I should have heard beeps complaining about no RAM.

I already verified that the PSU and speaker were working with my old components. Since I had swapped in a new motherboard, I was starting to assume the 240e was DOA. However, I found this on page 6 of the manual:
Please update the BIOS to the latest version while using AM2+ CPUs. Due to the latest CPU transition, you may encounter the situation that the new system failed to boot while using new AM2+ CPUs. In this case, please install one standard AM2 CPU to boot your system, and update the latest BIOS from our website for AM2+ CPUs support.
Since the 240e is an AM3/AM2+ compatible CPU, does this mean I have to find an AM2 cpu, boot and update the BIOS? Has anyone else had to do this for their AM3/AM2+ cpu and this board?

Re: No POST, No Beep, No video - DOA cpu?

Posted: Sat Jun 19, 2010 6:10 am
by MikeC
soobaerodude wrote: Since I had swapped in a new motherboard, I was starting to assume the 240e was DOA. However, I found this on page 6 of the manual:
Please update the BIOS to the latest version while using AM2+ CPUs. Due to the latest CPU transition, you may encounter the situation that the new system failed to boot while using new AM2+ CPUs. In this case, please install one standard AM2 CPU to boot your system, and update the latest BIOS from our website for AM2+ CPUs support.
Since the 240e is an AM3/AM2+ compatible CPU, does this mean I have to find an AM2 cpu, boot and update the BIOS? Has anyone else had to do this for their AM3/AM2+ cpu and this board?
Sounds like it, unfortunately. We've had to do similar things with both AMD and Intel boards in the past. At least it is updateable! 775 cpus, for example, have come in many variants, many incompatible w/ boards using chipsets of the previous generations. Fry's tech support might help... or perhaps a member of this forum who lives near you?

Re: No POST, No Beep, No video - DOA cpu?

Posted: Sun Jun 20, 2010 12:42 pm
by soobaerodude
MikeC wrote:Sounds like it, unfortunately. We've had to do similar things with both AMD and Intel boards in the past. At least it is updateable! 775 cpus, for example, have come in many variants, many incompatible w/ boards using chipsets of the previous generations. Fry's tech support might help... or perhaps a member of this forum who lives near you?
Picked up a Sempron 140, and system was able to boot. Flashed to latest bios, and swapped in the 240e. Same result: this CPU is DOA. First time I've ever run across a DOA cpu, especially a retail boxed one. Thank god for 30 day return policies. Unfortunately, the retailer is out of stock, and the 240e is a little hard to come by unless you win an ebay auction.

ECC no go on TA760G

Posted: Wed Jun 23, 2010 2:17 am
by soobaerodude
I installed 2 sticks of Kingston 2GB DDR2 ECC http://www.ec.kingston.com/ecom/configu ... 5/2GI&root into the TA760G and flashed to latest BIOS.

I then setup basic ECC mode in the BIOS and ran memtest86. However, the "ECC : Detect / Correct - Killchip" line says Off for me, unlike the screenshot from this user's A760G http://hardforum.com/showthread.php?t=1459497.

Anyone else with a TA760G using ECC and have "verified" it with memtest?

Re: ECC no go on TA760G

Posted: Thu Jun 24, 2010 4:40 am
by matt_garman
soobaerodude wrote:I installed 2 sticks of Kingston 2GB DDR2 ECC http://www.ec.kingston.com/ecom/configu ... 5/2GI&root into the TA760G and flashed to latest BIOS.

I then setup basic ECC mode in the BIOS and ran memtest86. However, the "ECC : Detect / Correct - Killchip" line says Off for me, unlike the screenshot from this user's A760G http://hardforum.com/showthread.php?t=1459497.

Anyone else with a TA760G using ECC and have "verified" it with memtest?
What version of memtest86 are you using? A too-old version of memtest86 won't recognize the memory controller, and therefore won't be able to use the ECC checks.

Also: in every version of memtest86 I've used, I've had to manually turn on the ECC checking. While memtest is running, there is a button you can push to bring up a menu. I can't remember what button it is, maybe "c" for configuration? But it says so on the screen. Anyway, when you bring up the menu, there is an option to configure ECC settings.

Re: ECC no go on TA760G

Posted: Thu Jun 24, 2010 8:36 am
by soobaerodude
matt_garman wrote: What version of memtest86 are you using? A too-old version of memtest86 won't recognize the memory controller, and therefore won't be able to use the ECC checks.

Also: in every version of memtest86 I've used, I've had to manually turn on the ECC checking. While memtest is running, there is a button you can push to bring up a menu. I can't remember what button it is, maybe "c" for configuration? But it says so on the screen. Anyway, when you bring up the menu, there is an option to configure ECC settings.
After digging around, I realized I misinterpreted what memtest86+ v4.10 (latest) was saying. It's saying that ECC detection and correction are on, but not chipkill. It looks like the the TA760G supports chipkill with x4 chips, but my modules use x8 chips . If I turn off ECC in the bios, then memtest86+ tells me that ECC is disabled.

Posted: Fri Jun 25, 2010 6:46 pm
by km2
I recently installed WHS onto my A760G board with AHCI drivers provided on the disk. On the WHS Console, all of my drives have a trailing zero at the end of their model numbers.

Anyone here who have encountered this problem?

Re: ECC no go on TA760G

Posted: Thu Jul 01, 2010 3:12 pm
by soobaerodude
soobaerodude wrote:After digging around, I realized I misinterpreted what memtest86+ v4.10 (latest) was saying. It's saying that ECC detection and correction are on, but not chipkill. It looks like the the TA760G supports chipkill with x4 chips, but my modules use x8 chips . If I turn off ECC in the bios, then memtest86+ tells me that ECC is disabled.
Well, I picked up a plain A760G, put the same ECC modules in, and ran memtest86+. This time, it said chipkill was on. Now I'm confused :?

From what I've been reading, the Kingston modules I have do not support chipkill. My guess is that the A760Gs bios is returning some hardcoded value with regards to chipkill whenever ECC is enabled. Which could lead to the bios returning hardcoded values for all ECC properties/registers whenever it is enabled, without providing ECC functionality.

Posted: Mon Jul 19, 2010 2:03 pm
by soobaerodude
For those running this board with an AM3 cpu and 2.0GHz HyperTransport, I found 7W-8W savings by downclocking HyperTransport to 1GHz.

This is what I'm running
Athlon II X2 250u 1.6GHz, 25W TDP
BioStar TA760G
4GB DDR2 ECC
160GB 2.5" SATA laptop drive
250W Sparkle 80+ PSU

After following many of the recommendations in this thread (powernow enabled, lowering Vcore to .9125V, disable unused ports, downclock graphics to 150MHz) I was still idling around 42W and was very disappointed.

Then, I found this post in German, where they downclocked an AM3 CPU's Hypertransport to 1GHz AM2 levels and noticed a difference in idle power consumption: http://translate.google.com/translate?j ... en&act=url

Anyone else using these 760G Biostar boards with an AM3 processor and running into the same thing?

Posted: Thu Jul 22, 2010 7:23 am
by Brian
I ordered an open box Biostar TA760G M2+ from NewEgg for cheap.

26W at the wall! That's with a PicoPSU, OCZ Vertex SSD, a dual-core Athlon64 X2 @ 1.0GHz, 0.95V, at idle.

I haven't dropped the graphics card in yet, and I think I'm going to wait a while before I do. Maybe I'll fire up some old games for now, because my GPU will probably put me just under 40W at idle.

I also have some work to do to calm Windows 7 down. There's a few programs in the background, sucking down CPU cycles all day, and occasionally making my CPU jump to a higher P-state and 40W system power consumption.

Re: ECC no go on TA760G

Posted: Sun Jul 25, 2010 10:29 pm
by truckman
soobaerodude wrote: After digging around, I realized I misinterpreted what memtest86+ v4.10 (latest) was saying. It's saying that ECC detection and correction are on, but not chipkill. It looks like the the TA760G supports chipkill with x4 chips, but my modules use x8 chips . If I turn off ECC in the bios, then memtest86+ tells me that ECC is disabled.
I just put together a machine with a Gigabyte GA-770T-USB3 motherboard, an Athlon X3 425 CPU, and a Kingston KVR13333D3E9SK2/4G 4GB unbuffered memory kit (two DIMMs). The memory datasheet says that it uses x8 chips, and I wasn't able to get chipkill working unless I changed the BIOS setting for DCTs to "ganged". Apparently this CPU has two independent memory controllers and chipkill will only work with x8 memory chips if both memory controllers are forced to operate in parallel. The downside is that the memory bandwidth reported by memtest86+ drops by about 6% in this configuration. I don't know how much this will affect real world performance, but I'm not too concerned because this machine now has CPU performance to spare.