ASUS M4A89GTD Pro - A Core Unlocking 890GX for $140
by Rajinder Gill on March 10, 2010 12:00 AM EST- Posted in
- Motherboards
Although we gave AMD’s 890GX/SB850 a lukewarm reception last week, there’s no denying that a sub $150 price point for a feature laden board such as the ASUS M4A89GTD Pro is deserving of closer inspection.
While most of the boards within a certain price bracket follow a common component formula, ASUS and ASRock have managed to segregate themselves by breaking rank and offering hardware level core unlocking for budget dual/tri-core AMD processors. That means that you can buy a sub $100 CPU and unlock it at your leisure to enjoy 4 core number crunching. The feature itself is not new per se, it’s just that AMD decided to drop support for ACC on series 8 chipsets, leaving vendors to do things on their own accord. ASUS’ workaround is almost certain to have upset AMD, but we think they made the right choice - if moving to the 8-series chipsets meant giving up core unlocking, many enthusiasts wouldn't.
If you take a look over at the blue corner (Intel), you’ll see that sub $150 expenditure almost exclusively limits you to the micro-ATX and mini-ITX form factors on the H55/H57 chipsets. While such motherboards are perfect in the context of small HTPC builds, there are times when users need increased levels of plug-in flexibility and future-friendly upgrade paths, and that’s where a full sized ATX motherboards start to make sense.
We’ve got a couple of AMD board reviews lined up for you this month, kicking-off today with the M4A89GTD Pro, as subjected to our standard test suite. Do note that we’re still in the process of bringing you a meaningful SATA 6G performance comparison – we should have a dedicated article up on this shortly.
Summary/Overview
One of the things that came to light during our first look at the AMD 890GX platform was an issue with our Sharkoon USB QuickPort. After a lot of back and forth, it’s been discovered that the problem lies with Sharkoon’s QuickPort firmware (the controller is made by Lucidport) and not ASUS’ M4A89GTD Pro.
ASUS have since acquired six Sharkoon QuickPort units and have discovered three different firmware releases have been used by Lucidport/Sharkoon at random, while our lab unit contains a fourth. Sadly, none of these firmware versions sit well with the AMD chipsets due to a ‘USB hand-off issue’. ASUS have reported to us that USB 3 drives from BYTECC and KINGWIN work fine with the AMD chipsets, although our advice would be to hold off for a few weeks while we find out which controllers these units contain and if Sharkoon’s problem can be fixed via a firmware update.
That leaves us with a few things on the M4A89GTD Pro that need attention, the first relates to S3 sleep resume. We used the Corsair TX 950 for our testing and found a hit and miss scenario because there are instances where pressing the power button to instigate resume does not work after the board has been in S3 state for 2~3 hours; the only way to get the board to post again is to switch off at the mains, let residual power drain from the PSU and then turn back on. It’s not something that manifests all the time, so tracing it down could prove a little tricky, although we’re told ASUS is working around-the-clock to find the problem.
The other area of operation that needs some tuning relates to how the M4A89GTD Pro handles Elpida Hyper based memory modules for overclocking. When used in conjunction with ASUS’s ‘Core Unlocker’ feature; a failed overclock can result in the need to clear CMOS before the board will post again with 4 cores active. In fact, we found that the Hyper based modules are more stable at CAS 6, than they were at CAS 7 or 8 on this motherboard. We were sent a beta BIOS late last week that improves memory compatibility and although things are a little better, the Elpida Hyper issues are persistent - there’s clearly a need for more work.
Once you work out how to circumvent some of these quirks, the M4A89GTD Pro does overclock quite well – get it dialed in properly and the board is comfortable holding 8GB of memory stable whilst maintaining high memory controller frequencies in tandem. We managed to squeeze a perfectly respectable 4GHz from our Phenom X2 555 processor, unlocked to 4 cores with 8GB of memory running DDR-3 1600MHz 6-7-6-18 timings at a 1T Command Rate using our Corsair Dominator GT modules.
ASUS have also put a lot of work into promoting their automated overclocking routines to us over the past few months so we put this feature to the test. ASUS offers two ways to instigate automated overclocking, the first method is via BIOS, and the second is by using the supplied Turbo V overclocking utility within the OS.
The BIOS level utility was a little hit and miss for us depending upon which memory modules we were using. With Elpida Hyper modules the board has a tendency to select CAS 8 based timing sets which does not sit well with the current BIOS releases, resulting in an unstable overclock. Using more conventional and affordable memory like the Corsair 1600MHz Dominator kits, we obtained an easy 3.6GHz overclock with CAS 9-9-9-27 timings at a 2T Command Rate.
It’s the OS level Turbo-V utility where ASUS have really nailed automated overclocking, there are a variety of tuning options which can result in overclocks right on the hilt and beyond where we’d setup our components manually. ASUS have done a great job in simplifying what can be a complex process for many users.
Had the random S3 resume issue not have been present, we’d have been hard pushed not to give this board an award for catering to every type of user and providing a feature set that belies its price. What we'll do for now while ASUS polishes the M4A89GTD Pro, is encourage you to put it at the top of your shortlist as one to watch...
46 Comments
View All Comments
ReaM - Friday, March 26, 2010 - link
I don't know why you people say "I don't like AMD"Now tell me, have you ever been satisfied with intel's performance? I have never been. Since my 1400 thunderbird :P I am with AMD. Unlike the p4 3000 Northwood, it did not have fps drops in counter strike. I had to overclock the northwood to 3.4 to play it decently.
Then I had a Venice 3000+ that would run at 2.8ghz from 1.8 base. I still have it.
And still today, in games AMD beats the more expensive i5 750. How many of the market will use i5 750's benefits? Almost noone! People really need to look and see that the best bang for buck is AMD. Cheap Boards, Cheap CPUs, and in summer RAM prices are expected to drop. It will be a 200 dolla upgrade to a quadcore if you have PSU and a case already.
I have just bought i7 860 yesterday (i got it real cheap, I wanted the 920) with p55m ud2 Gigabyte and gskill 4gb 1600 cl7 ONLY because I need the render faster in Maya. I just have no choice here, because AMD does not seem to be the renderer's choice.
It is bad to have no choice, because there is not direct competition between the two, but often I do not understand why many mainstream people get i7 920 for gaming. Because they don't care how much they spend?
I am also forced to choose Quadro over FireGL, but that's another story. Radeon since 4xxx series is SICK!!! I love ATI and AMD.
You should not hate AMD, cmon
But for mainstream user, why i5? Way too expensive
butdoesitwork - Thursday, March 11, 2010 - link
Forget speed. Is it safe to be an early adopter?1. Does the Linux kernel support this chipset at all yet? Does it need any special boot parameters to work at all? e.g. Do I need to forcibly disable MSI again?
2. How _well_ does it work? If you copy files over these new-fangled SB850 SATA ports, does it get there correctly? e.g. can you do a linux "diff" to prove that it isn't broken? Are there any strange messages in /var/log/messages?
Been bitten by this too often in the past. Most recently the cheap JMicron SATA-to-USB bridges found in most external enclosures often caused bus resets and data corruption. May have just hit a similar problem with my board's JMicron eSATA port.
Taft12 - Thursday, March 11, 2010 - link
As a Linux user you should be well aware by now that it's never safe to be an early adopter.Check the changelogs of the newest kernels.
Rajinder Gill - Thursday, March 11, 2010 - link
I have not tested Linux on this platform yet - I find things need a while to settle before everything works properly (that goes for all operating systems). I might give things a whirl on our next review depending upon time constraints. For this article we tested platform compatibility with Windows 7 and found the big problem was with the Sharkoon QuickPort (LucidPort perhaps, although its still up in the air).Regards
Raja
AceKing - Wednesday, November 10, 2010 - link
So has anyone tested this with Linux yet? I'm curious to know if there has been any problems. I see the last post on this was in March, which was 7 months ago.bji - Friday, November 19, 2010 - link
I own one, and it works perfectly in Linux. Onboard video and audio are flawless and well supported in Xorg (no xorg.conf even needed) and ALSA (easy alsaconf setup). I use a Phenom II 1075T and man does it fly. I primarily do software development and 6 thread compiles are just so fast. Even faster than the much more expensive 4 core hyperthreaded Xeon that I have at work.The board does do this weird thing where if you power it off before the OS has shut it down completely, when you power it back on it thinks that you have an overclocking problem (probably detects an incomplete shutdown and assumes bad overclocking) and requires going into the BIOS and back out to continue the boot. And I don't have any overclocking enabled at all, I just manually set the timings for my memory because the defaults were not right.
I disable the JMicron SATA RAID and IDE controller though as I don't need them and I never trust anything that JMicron makes. Nor should you.
caraway - Tuesday, November 30, 2010 - link
Which kernel and distribution are you using?GeorgeH - Wednesday, March 10, 2010 - link
Measuring DC power consumption as opposed to AC is a very welcome improvement - hopefully this is Anandtech's new SOP for all reviews.The only question I think you should make clear (or clearer, with apologies if I missed it) is if you're measuring and summing the DC power across all of the rails or just on a select few; i.e. are you including HDD, ODD, GPU PCIE, and fan connections in the power consumption figures?
Rajinder Gill - Wednesday, March 10, 2010 - link
Hi,I'm not adding HDD's/DVD peripherals into the figures. The power figures shown are from all 4 rails - EPS 12V, ATX 12V, 5v and 3.3v. The CPU fan consumption is about 2w and left in. I can make things lot more granular than I have here - but I figure what we're providing at present is enough for reviews like this, unless concentrating on specific areas of consumption (which may be called for at some point).
regards
Raja
GeorgeH - Thursday, March 11, 2010 - link
I saw that in the 980X article - tell Anand you need a raise.