TR Forums

It seems like MSI also failed me, at least for now - prepare for a long story below.

I used to have a Gigabyte board with an AMD 770 chipset, which wasn't exactly doing justice to my components - Phenom X4 9550 and 4 GiB of GeiL Value memory (800 MHz, 5-5-5-15).
The system would, for example, refuse to go to sleep mode successfully (whether I overclock or not), as it would crash upon comeback (in the better case) with screen corruption, or simply not come back at all (the usual case).
I was baffled to find out that after 6 BIOS updates, the situation remained largely unchanged. Gigabyte support was unable to give me any reasonable advice other than "give more voltage to your RAM" which, in fact, didn't help.
In addition, RAM was nigh unworkable using the (optimal) Unganged mode, so I had to use legacy dual-channel and grit my teeth about it. That's not mentioning the board was a far cry from top-end "140 W support" boards, and only allowed a maximum of +0.1 V deviation from the default processor voltage (with no undervolting supported, by the way). This caused the board to have sub-par overclocking performance too, although I was satisfied - I ran the quadcore at 3 GHz stably, and the memory was chugging along at a relatively decent speed too.

Most enraged by the sleep issue however, and refusing to deal with the shoddy warranty service of the distant shop I ordered from, I decided to simply upgrade and forget the old board.
Today, here I am, using MSI's shiny new KA790GX board.

Let us form a bit of a review before I get to the point.
The board arrived in a very nice box, with a weird, rainbow-tinted object on the front side - it looked like a mergence of a subwoofer, a large fan and a daisy. Okay, said I, and opened the box to find out that the components are plentiful and neatly packed. At this point, I was excited that I'm about to get a more stable, better and much faster (SB750!) board that'll make my previously snappy system feel like it's taking off.
Here's the specification sheet: MSI KA790GX

I attach, screw and connect everything quickly to/around the vomit-coloured motherboard (an unwelcome change from the old Gigabyteblue), turn the case upright, plug in the back panel apparatuses and power up the system in anxiety.
An extremely annoying, high-pitched shriek comes out from my computer. I become unsettled a bit, and immediately power down the computer, which makes the strong beep fade in an awkward fashion. I try again, 3 or 4 more times, only to discover that the computer will still shriek at me while refusing to boot.
At this point, I'm panicking, and I'm taking apart the computer piece by piece to pinpoint the problem. Not the CPU (and it was a bitch to reattach my custom heatsink yet again), not the RAM, not the hard drives. I was afraid physical damage could occur, since the computer was crying even after I unplugged my primary suspect, the motherboard speaker. Finally I discover that I didn't plug in the PCIe power to my videocard (good ol' 8800 GT) and everything was dandy once I did - but man, I've never heard of a whistling videocard before. It was intimidating.

My system powers up, Vista 64 runs as usual, all I have to do (seemingly) is to reinstall motherboard-related drivers just to be sure. Sleep mode works. I cheer like there's no tomorrow. This time maybe I've found the holy grail.
I even tried out the jumper-based quick OC option, which overclocks the processor by 20% (reference clock is therefore 240 MHz) and it works nice and stable. There are no memory related issues or instability.

Not until I try to overclock manually.

After setting back the jumpers, I try to one-up the automatic overclock by setting the reference clock to 244 MHz. The system doesn't boot. I do my usual routine of "press Reset 5 times in a row" and the system boots, telling me that the overclock failed. I'm completely baffled. At 240 MHz the system is perfectly stable - at 244 MHz it doesn't even manage a POST. I decided to investigate further.
I could get the system to start by giving 2.1-2.2 V to the memory, but I thought that was odd (and unsafe no less). At default voltages (or "higher by one notch", as I usually set it), the system fails to boot from a manual overclock to 240 MHz - highly suspicious given that the jumpers achieve the same thing with 100% stability.
I become nervous - my old motherboard couldn't sleep, but at least it let me set the reference clock to 273 (x11 = 3003) MHz and ran completely well. This is where I begin to find strange things.

1) I cannot set the tRC timing for RAM (only option is Auto, when I can set, like, 20 other timings for each RAM channel), and according to Everest and AMD OverDrive, it's set at a dangerously unstable value by default.
This might account to overclocking instability - believe it or not, I can't boot 244 MHz even after setting the RAM to 400 MHz (with proper timings forced except tRC), as it still shows the symptoms of unworkable RAM (POST stops in place, or the BIOS freezes)
2) The components are otherwise flawless, they shouldn't require extra voltage, but even given the voltage, the system fails to give me the performance I was expecting
3) The motherboard has been out for at least 1 month, and there are no BIOS updates at all.
4) Both Everest and AMD OverDrive (v2.1.4, the latest) claim that the board is SB700-based. Impossible. It's an AMD 790GX chipset with 140 W processor support to boot - the box, the official website, everything says it has SB750. Because of this, I can't use Advanced Clock Calibration. To add insult to injury (or oil to the fire, or any custom metaphor), AMD OverDrive is incapable of setting anything on the board - as if the controls were not set/exposed in the BIOS. Voltage and clock controls remain locked, voltage readings are zero, and other strange whatchamacallit things happen (or don't happen) if I attempt to, err, overdrive the motherboard anyway.

At this point I'm deeply disappointed and I filed a technical inquiry at MSI's website, explaining the shortcomings and letting them choose between providing a BIOS update or a refund.
Feel free to flame me, discuss, or comment otherwise. Do you think the board needs nothing but a new BIOS, or do you think it's faulty and requires replacement?

Personally I think you are pissing and moaning about something you have no right to complain about. MSI doesn't guarantee that any mother board will over clock. If it will run at stock speed MSI has held up their end of the deal.

Now as far as you actual problem goes one thing springs to mind that might fix the overclocking problem. Go find the PCIe bus speed and manually set it to 100mhz. If for some reason the PCIe speed is going up when you turn the HTT speed up your going to have issues.

Thanks for the input.
I'd just like to add that I've never seen a board in my life that's such a bad performer. Much more "bargain-bin" boards were able to push the HT clock to 260 MHz and beyond for me in the past.
MSI should not allow their boards to be this shoddy.

You have not addressed the generally buggy nature of the board though - RAM timings that are not safe, voltage switches inaccessible under AMD's own software, multiple programs reporting SB700 instead of 750 - something's not right with this board, and I really hope that the solution is just a BIOS refresh away.

I got bit by the MSI is a great overclocker story a few years ago. It was the first and last MSI motherboard that I will own.

Like I said before MSI guarantees that the motherboard will run at default settings. If you can build a system with the motherboard and it runs at the correct speed for the processor they have held up their end of the deal. Which programs state SB700 ? CPU-z is the only one I have ever used much. I haven't kept up on AMD chipsets for awhile but if the SB750 is an updated version of the SB700 that might be why it shows up wrong. If thats the case a newer version of cpu-z or whatever might show it as actually being the SB750. As far as AMD's in windows tweak program goes I have never seen one of them yet that worked well. It being broken sounds normal for every in windows tweak program I have ever seen.

AMD OverDrive is their component overclocking and forensic application, it's not the "windows tweaking" one. That one's something else.
It worked brilliantly with my previous AMD 770 board from Gigabyte. Voltage switches were available and almost all of them worked fine. Clock frequency sliders worked flawlessly. There wasn't a problem.
With the new 790GX board, nigh all the sliders are locked and unavailable. Voltage readings are scarce and inaccurate or simply nonexistent. The ACC dialogue (for boosting Phenoms with SB750) is unavailable due to the apparent lack of an SB750 chip, which is impossible.

I also use Everest Ultimate Edition (very robust program, CPU-z is "notepad.exe" in comparison) for benchmarking and monitoring, it also says I have SB700, which must be some sort of interesting joke on MSI's behalf.
One way or another, I'll get to the bottom of this, anyway.

I fired up CPU-Z as well just for you darling, and this is what I get yet again:
http://img530.imageshack.us/img530/8403/tempxx7.jpg

My best bet is that MSI screwed up extremely big-time and are manufacturing their brand new boards with a 5-month-old 780G BIOS, which in turn makes things screech and crackle.

Quite funny, because their website has a different number order for the date, meaning "September 5", which would seem good, but the data CPU-Z (and other programs) gathered shows that it's actually 9th May, which is a sacrilege of all that's computer enthusiasm.

Meadows: I know this is a pretty old thread, so excuse me for ressurecting it.... Most members of TR forums have become so anti-AMD that I don't get on here much anymore because I can't get any useful info....

I really have no idea why you buy cheap parts to try and overclock Phenom systems, other than the fact that you like to complain...
In reality you could have bought the MSI DKA790GX Plat. and an Unlocked Phenom (9850/9950) for ~$70 more than what you paid for the Non-Plat/9550(locked) processor. I tried to warn you about the problems with high HTref speeds, Phenom and GBT boards months ago, before you even bought your first parts, but you didn't want to hear any of it... But I digress...

This is a 100% stable 24/7 clock at reasonable V's on an MSI DKA790GX Plat. using a 9950:
CLICKABLE:



This is a MAX benchable clock that will pass almost everything except 4+ hours of stress:
CLICKABLE:


If you notice, both those SS's show my board as having an SB700 Southbridge when it's obiviously is an SB750. That is a fluke of CPU-z and has nothing to do with AMD or MSI...

"I really have no idea why you buy cheap parts to try and overclock Phenom systems, other than the fact that you like to complain"

Simple. I've had tremendous good experience with them - it's just that I've had a bad luck streak in the past six months or so.
Anyway, the problem may be solved; the motherboard has been sent off to RMA'ness, and shortly thereafter my 600 W Atlas (Super Flower as of today - they are NOT noname) power supply blew the fuse of the entire house, so that's going off for warranty as well (luckily, nothing got damaged in the computer).

I'll see this Friday if I have any luck. If the replacement components are decent, all my problems will disappear. In my country, you can't buy/order a 790FX board yet with SB750, and quite frankly, I generally buy the cheaper versions because I simply don't need all the features of the big brother (and I'm more cash-strapped than you seem to think I am).

About SB700/750 - let's assume you're right and CPU-Z is wrong. Indeed, the latest EVEREST Beta has reported a fuzzy combination of both 700 and 750 numbers for my SB, which was confusing - it "knew" it's an SB750 board, but reading the hardware generated a result of SB700. This notwithstanding, the voltage controls, frequency controls and ACC were all unavailable in the latest AMD OverDrive (v2.1.4), and ACC should've been available since 2 older versions or something (can't bother remembering). This clearly shows that there's something up with it. Voltages were reported as zero, PCI-express frequency reported as zero, multipliers locked both ways for whatever reason, and something was wrong with the southbridge anyhow because PCI devices produced IRQ conflicts (or bad sharing) which was NOT present on my Gigabyte AMD770 board - namely, when I'd type on my (otherwise lovely) PS/2 keyboard, my Audigy would crackle.
I took it back to the store and asked for a replacement, and if I'm lucky, that one will function with OverDrive and won't suck. If I'm even more lucky, I'll get to choose another brand altogether, courtesy of the supplier.

All in all, it turned out to be quite defective (I didn't get the functionality I had paid for) and I'm entitled to a working replacement. I don't know why you poke me with more expensive components, I'm not the guy who buys a new 1000 dollar PC refresh package every 4 months.

Daveburt714 wrote:

Most members of TR forums have become so anti-AMD that I don't get on here much anymore because I can't get any useful info....

That I have to disagree. Falling out of favour because the price/performance not satisfying most members does not necessarily mean everybody suddenly hates AMD. There are situations we recommend AMD systems and the System Guide consistently includes AMD alternatives too. Also if you check the news comment most people still want AMD to be a viable alternative. They are always the underdog and as much Intel recommendations are given out these days, I bet a lot of them are still closet AMD supporters. So I really have no idea why you think most of us here are anti-AMD.

I know I've said it to you before, but if they let you pick a new brand, go for the BioStar 790GX board. Its fabulous! And its inexpensive, at least in the States. I bought it retail @ MicroCenter for $100US.

Corrado wrote:

I know I've said it to you before, but if they let you pick a new brand, go for the BioStar 790GX board. Its fabulous! And its inexpensive, at least in the States. I bought it retail @ MicroCenter for $100US.

I'm sorry but I'm not in the states, and I don't think I've ever seen that brand at our local retailers and e-tailers. I'd be happy with even MSI or Gigabyte really, if only they worked this time.
However, if they would indeed let me switch brands, I'd probably see if I can get something from ASUS for the next round.

Meadows wrote:

Corrado wrote:

I know I've said it to you before, but if they let you pick a new brand, go for the BioStar 790GX board. Its fabulous! And its inexpensive, at least in the States. I bought it retail @ MicroCenter for $100US.

I'm sorry but I'm not in the states, and I don't think I've ever seen that brand at our local retailers and e-tailers. I'd be happy with even MSI or Gigabyte really, if only they worked this time.
However, if they would indeed let me switch brands, I'd probably see if I can get something from ASUS for the next round.

http://www.biostar.com.tw/

http://www.biostar-usa.com/app/en-us/t- ... p?S_ID=376


I don't know where you're at though, so obviously I'm not sure that it will even help you.

The Asus 790GX board is apparently quite nice. I'd considered buying one before I saw the M3A79-T Deluxe (790FX + SB750). I haven't tried using overdrive with it yet, though, currently running an Athlon X2 5400+ black edition and waiting on a replacement power supply before I can start playing with it more.

I just wanted to throw some random info out there regarding the identity of the southbridge. On my old HTPC motherboard (an MSI RS482M4-ILD) I knew that it used an RS482 northbridge and SB450 southbridge. However various applications such as CPU-Z identified them as an RS480 and SB400. The idea I have is this: perhaps since the SB750 is an "update" to SB700 and not a completely new southbridge some sort of identifying microcode (or whatever) is similar enough that those applications will call it the older version until they get information specific enough to differentiate. That's just a guess, of course, but I think it's safe to say that there's definitely an SB750 on there and that the identification fault lies in the applications or, as you suggested, maybe something in the BIOS isn't up to date.

Also, I like AMD... always have.