Most 780G boards break down under stress of high-end Phenom

S

Scali2

2[H]4U
Joined
Apr 20, 2006
Messages
2,845
http://www.anandtech.com/mb/showdoc.aspx?i=3279

Apparently the motherboard manufacturers have structurally underrated the power supply on the boards, using only 3 or 4-phase (how cheap can you get? Even my old mid-end Asus P5B Deluxe has 8-phase, and I think that was actually on the market before quadcores, although it later supported them, and even supports 45 nm). This literally goes up in smoke with a 125W Phenom (or even the 6400+).
Could this be a mistake in the reference design?
At least it nullifies the idea of low-cost high-performance Phenom systems on the Spider platform.
I also wonder how smart it is to put these Phenoms into older Am2 systems... There may be many Am2 boards with insufficient power supply aswell.

Just when you thought Phenom/Spider were picking up again after the delays, TLB-fiasco and underwhelming performance...
 
agreed, this is stupid.

who is going to buy a 780G board until then have assurance that the board maker has provided capable power delivery?
 
Scali2 said:
http://www.anandtech.com/mb/showdoc.aspx?i=3279

Apparently the motherboard manufacturers have structurally underrated the power supply on the boards, using only 3 or 4-phase (how cheap can you get? Even my old mid-end Asus P5B Deluxe has 8-phase, and I think that was actually on the market before quadcores, although it later supported them, and even supports 45 nm). This literally goes up in smoke with a 125W Phenom (or even the 6400+).
Click to expand...
No. of phases alone isn't necessarily an indicator of a good PWM design (in fact it can be cheaper to build a poor 1 with more phases than a good 1 with fewer), it's quite possible to build a high quality 4-phase PWM that will handle more than 125W adequately.
Also, a lot of PWM designs that are marketed as 8-phase or 12 phase aren't truly but are 2 4-phase or 2 6-phase running together or just have more MOSFETS per phase.
 
At the same time, these motherboards are directed at HTPC use where most users would probably not have 125watt processors, but instead 65 or 90 watt at most.
 
BUFF said:
No. of phases alone isn't necessarily an indicator of a good PWM design (in fact it can be cheaper to build a poor 1 with more phases than a good 1 with fewer), it's quite possible to build a high quality 4-phase PWM that will handle more than 125W adequately.
Also, a lot of PWM designs that are marketed as 8-phase or 12 phase aren't truly but are 2 4-phase or 2 6-phase running together or just have more MOSFETS per phase.
Click to expand...

I think most of us know that. Doesn't change the fact that these boards all break, and my old P5B Deluxe can handle any quadcore easily.
 
so you agree that the problem then isn't the no. of phases but the overall quality of the PWM design?
e.g. abit's IP35 series has a 4-phase PWM design but will handle overclocked Intel quads.
 
BUFF said:
so you agree that the problem then isn't the no. of phases but the overall quality of the PWM design?
e.g. abit's IP35 series has a 4-phase PWM design but will handle overclocked Intel quads.
Click to expand...

No, I think it is a combination of low number of phases AND low quality PWM design. If they had used the same quality design and components, but used more phases, it would have been better.
Besides, some boards have a 3-phase design, which is probably slicing it too thin, no matter how good your overall design is.
4-phase might work, *if* you have a very good design. Which apparently these boards don't, because fact is that they break down.

Conclusion: with the average design quality of current motherboards, a 3 or 4-phase design isn't going to cut it. Whether they solve that by improving their quality, or by changing the design to more phases, doesn't really matter. More phases is probably the cheaper solution though (robust design with cheap components > cheap design with robust components), which would be why I would still insist that the problem is indeed in the nr of phases. QED
 
Guess that we'll have to agree to disagree then.
There's no guarantee that a design with more phases composed of cheaper components will perform any better.
 
BUFF said:
There's no guarantee that a design with more phases composed of cheaper components will perform any better.
Click to expand...

I never said that. I think your foolish pride is just keeping you from agreeing.
 
Gigantic advances in mosfet design (faster switching, lower heat) has allowed manufacturers to reduce part counts. And since mosfets are a far greater expense than controllers they try to cut corners. Throwing tons of phases at the problem isn't necessarily the perfect or most balanced solution either. While more phases might lower ripple and increase transient response, reduce load and heat, it also lowers efficiencies -- that doesn't cut it as an HTPC.

Perhaps changing out the caps with higher rated ones is the cheapest solution. This would help to even out the power. Or perhaps just adding some efficient heatsinks to the mosfets, to dissipate the extra heat. OR maybe switch out with better rated mosfets with lower rds(on) for higher-end boards. Someone handy with a soldering station could do all these things. Just about anyone could add more/better heatsinks to the mosfets.

Although, unless Gigabyte, Asus, and ECS specifically mentioned these systems running with 125W CPUs, this story is a non-issue to me. Where was the furor about counterfeit leaky capacitors scandal, which for several years involved almost every motherboard OEM under the sun. If you have a 780G, stick to anything but 125W CPUs, and wait for 45nm parts. Problem solved. OR get out the wallet and buy a 790F/X. Either you get that IGP or a 125W, but you currently can't have both.
 
You actually *can* have both, because there are some 780G boards with 5-phase power, like one from Asus (but not all from Asus). They are just the exception to the rule...
Question is why, since the 780G chipset is part of the Spider platform, which was supposed to be for all Phenom CPUs. In fact, even the slower Phenoms use 95W, so a motherboard burning out at 125W sounds like the headroom of the design is too close for comfort. Running a board slightly out of spec is one thing, if it gets unstable or shuts down, okay... but having physically damaged components is unacceptable in my opinion.
 
vista_blista said:
Although, unless Gigabyte, Asus, and ECS specifically mentioned these systems running with 125W CPUs, this story is a non-issue to me. Where was the furor about counterfeit leaky capacitors scandal, which for several years involved almost every motherboard OEM under the sun. If you have a 780G, stick to anything but 125W CPUs, and wait for 45nm parts. Problem solved. OR get out the wallet and buy a 790F/X. Either you get that IGP or a 125W, but you currently can't have both.
Click to expand...
rubbish.

it's not like we are asking these motherboards to run the equivalent of a $1000 Intel Extreme Edition CPU.

we are asking it to run AMD's standard phenom processors, and even then we the processors in question are no faster than Intel's slowest quad core.

there is nothing that should suggest to the customer that maybe a 9750 won't work whereas a 9650 will. this is craptastic to the max.
 
Scali2 said:
You actually *can* have both, because there are some 780G boards with 5-phase power, like one from Asus (but not all from Asus). They are just the exception to the rule...
Question is why, since the 780G chipset is part of the Spider platform, which was supposed to be for all Phenom CPUs. In fact, even the slower Phenoms use 95W, so a motherboard burning out at 125W sounds like the headroom of the design is too close for comfort. Running a board slightly out of spec is one thing, if it gets unstable or shuts down, okay... but having physically damaged components is unacceptable in my opinion.
Click to expand...


They're satisfying the lowest end of the market first. ASUS and Gigabyte have ramped up 5-phase powered boards, which will likely cost 20 percent more. That's why I have learned to be patient, and never an early adopter. You never get the actual product you want without a wait.

BTW, I have bought quite a few of motherboards that falied prematurely or clearly didn't live up to their promise, in lots of different areas. Yes, one board with leaky faulty caps even melted down and could have caused a fire. To only notice now and start a rant is disingenuous. Someone needs to read off the part numbers of these boards. Because we only have half the story. And the rest of it has more drama than actual documentation. Where are the quotes and specific language that says these boards work with those parts? We need more information. Who knows what's happening here, or how many assumptions are being made.

Again, common sense dictates that cheaper motherboards have cheaper lesser performing parts. If I had my way, every motherboard would have Panasonic capacitors and top-flight IRF mosfets. But that's not reality. Maybe this story brings about a revision, and more accountability toward specific compatibility. That certainly would be a good thing. But let's get the story right then, and direct this at the entire industry and its practices.
 
Scali2 said:
I never said that.
Click to expand...
"More phases is probably the cheaper solution though (robust design with cheap components > cheap design with robust components), which would be why I would still insist that the problem is indeed in the nr of phases"?
sure sounds like you are advocating a design with cheap components & more phases & that isn't guaranteed to perform any better (you've already admitted that no. of phases alone isn't an indicator of a good PWM design).
I think your foolish pride is just keeping you from agreeing.
Click to expand...
methinks that thou art the 1 with ruffled feathers ... ;)
 
Better quality parts trump more phases a majority of the time. There's a point where you reach diminishing returns. Or at least, distributing the load with a greater number of lower performing parts isn't the most acceptable answer. Manufacturers have the increasing capability to design and control tolerances within very small power increments, in order to maintain a certain margin of profit. Especially if a distributor cuts you a deal on some parts, more phases has greater allowances in design.
 
BUFF said:
"More phases is probably the cheaper solution though (robust design with cheap components > cheap design with robust components), which would be why I would still insist that the problem is indeed in the nr of phases"?
sure sounds like you are advocating a design with cheap components & more phases & that isn't guaranteed to perform any better (you've already admitted that no. of phases alone isn't an indicator of a good PWM design).
Click to expand...

I think it is obvious that they are *already* using cheap components. That is the assumption under which I made that statement anyway. Hence I never hinted at using *cheaper* components when I used the words 'cheap components', as you claimed ("There's no guarantee that a design with more phases composed of cheaper components will perform any better.")
So, current situation is:
- Cheap components
- Cheap design

You can improve either of the two to make it perform better, either is guaranteed to perform better, ceteris paribus.
Now, I say it's better to change the design so you use more phases, in other words, more of the same cheap components, rather than taking the current design, and try to make it more stable by using more expensive components.
This because the higher grade components required for such a design usually go up in price exponentially (remember, we have to bridge a gap from 95W to 125W, which is about 32% extra that these components will have to handle, probably inflating the price per component with more than 32%... Going from 3 to 4 phases will increase the component cost by 33%, going from 4 to 5 phases will only be 25%).

I don't see how you could possibly disagree with that.
 
does anyone have any info/links on these newer 5-phase 780G motherboards?
 
Swingline(Red) said:
At the same time, these motherboards are directed at HTPC use where most users would probably not have 125watt processors, but instead 65 or 90 watt at most.
Click to expand...

+1

AMD should be careful marketing this chipset as part of the Spider platform. These boards are for HTPCs and pair absolutely perfectly with a 45W BE (not black edition) processor.
 
vista_blista said:
BTW, I have bought quite a few of motherboards that falied prematurely or clearly didn't live up to their promise, in lots of different areas. Yes, one board with leaky faulty caps even melted down and could have caused a fire. To only notice now and start a rant is disingenuous.
Click to expand...

I've been building PCs for over 15 years myself, and I've seen my share of motherboards, but I've never seen anything so structurally wrong with an entire line of products with so many OEMs at a time.
To this day I've only had two motherboards that were really faulty. One was an Abit i815 board with a Celeron, I think it had the bad cap problem. It died after about 2-3 years of intensive use.
The other was an Abit Athlon board, which had a similar problem to these boards: it was not designed to handle the power drawn by my Athlon 1400, and it was not clearly advertised by the manufacturer. It didn't actually die though, I just couldn't get it to run stable. I then exchanged it for an MSI board with virtually identical specs (same chipset and everything), and that one still works today. Just a better overall design.

vista_blista said:
Someone needs to read off the part numbers of these boards. Because we only have half the story. And the rest of it has more drama than actual documentation. Where are the quotes and specific language that says these boards work with those parts? We need more information. Who knows what's happening here, or how many assumptions are being made.
Click to expand...

Thing is that the evidence has been deleted mostly. Many OEMs have now edited the specs on their website to indicate that 9750 and higher Phenoms will not work. This information was not present prior to Anandtech publishing the article. Which is why Anandtech is making such a fuss, and rightly so.

vista_blista said:
Again, common sense dictates that cheaper motherboards have cheaper lesser performing parts.
Click to expand...

That's no excuse. Firstly these aren't low-performing boards. They have AMD's cutting-edge 780G chipset, which is part of the Spider platform, and the Spider platform is advertised with high-performance Phemon quadcores and CrossFire.
Secondly, they're not exactly cheap. They're mid to high-end in terms of pricing. Since AMD's Phenoms are also mid to high-end (only just barely, if you are willing to accept the Q6600 as a high-end part, which the pricetag really defies).
Now if they were supercheap Asrock boards aimed at last-gen parts and performance, then yes, you can't expect them to work properly... but in this case, 780G -> Spider -> Phenom, that's what both AMD and the OEMs are saying (or were anyway, before they edited the info).

Thing is that prior to these boards breaking down, Spider was seen as a single platform, and Phenom as a single CPU line. Now there's this vague subdivision where some Spider boards may support some Phenoms, but there are no guarantees, and you'll have to study the small print carefully... It defies the whole idea of having a platform.
A sharp contrast with Intel, where many 775-socket boards required nothing but a bios update to support quadcores, and in many cases even 1333 bus and 45 nm models.
I also wonder why even those cheap Asrock boards with 775 socket support Intel's quadcores without a problem. The first Kentsfields were also rated at 125W, and these boards start at only just over half the price of these 780G boards.

vista_blista said:
Maybe this story brings about a revision, and more accountability toward specific compatibility. That certainly would be a good thing. But let's get the story right then, and direct this at the entire industry and its practices.
Click to expand...

They will probably just continue selling the boards, but with a warning not to use CPUs rated over 95W. Some may or may not do a revision later. Sadly this is detrimental to the potential of the 780G chipset and the Spider platform as a whole.
 
jwlangs said:
+1

AMD should be careful marketing this chipset as part of the Spider platform. These boards are for HTPCs and pair absolutely perfectly with a 45W BE (not black edition) processor.
Click to expand...

It's not the chipset that is the problem. In fact, the same problem also exists with some nVidia-based Phenom boards. It appears that AMD perhaps supplied the OEMs with a bad reference design or spec sheet about the Am2+ socket or so. Most OEMs just use the reference design for regular boards like these, only more expensive boards have custom designs for overclocking etc. I can think of no other explanation why so many OEMs have problems with Am2+ boards. These same OEMs also make Intel boards, and socket 775 also needs to deliver 130W according to spec, because of the Pentium D and the extreme edition quadcores. Even the cheaper boards can handle these chips without a problem... they just would not be as suitable for overclocking as the more robust expensive boards.
So it's not like these OEMs don't know how to build boards that work with 125+W CPUs, and it's not like such boards have to be expensive.
 
I have to agree, whether AMD and the OEMs anticipated a greater share of real-world market interest to blossom amongst the HTPC enthusiasts for the 780G boards, they have all this time pushed the high performance angle ALL THE WAY up and down the line. I just built a new sys with the ASUS M3A78-EMH HDMI and I can't tell you how many times they reference the performance boosts and the HYBRID CROSSFIRE using a discrete PCI-E card + the onboard GPU... although if pushed I prolly will go through and count all those references, I mean, y'all SEEM like good folk
 
Looks like Anandtech has been working hard on this particular issue. 780G issues Part Deux:

http://www.anandtech.com/weblog/showpost.aspx?i=427

Basically the manufacturers said that if they used higher quality components to handle the additional heat generated by the uATX design, it would go out of their targeted price range for this group of boards.

I, myself, got the 780G because it had the SB700 chipset and I was wary of the stability of the 790X system. If they had an ATX version of the 780G chipset at the time I bought it, I would of definitely snapped that up.
 
I'm right there with you on that - I looked and looked for ASUS's M3A78-EH ATX board,
but could find it nowhere, and settled instead for the mATX version from ZZF for $95
 
MSI have four mATX boards listed as 125W capable, two AMD and two nVidia.

one of the nVidia 8200 boards (K9N2GM-FIH) is also listed as being "Solid Caps", and this is what i will wait for.

The motherboard is THE most important part for a stable system, i do NOT buy cheap-ass motherboards. ever.
 
popowrx said:
Looks like Anandtech has been working hard on this particular issue. 780G issues Part Deux:

http://www.anandtech.com/weblog/showpost.aspx?i=427

Basically the manufacturers said that if they used higher quality components to handle the additional heat generated by the uATX design, it would go out of their targeted price range for this group of boards.
Click to expand...

One manufacturer claimed that the stock AMD cooler doesn't have enough airflow to the components surrounding the socket. That's an interesting theory... Intel has round coolers with blades that fan out round the socket.
I guess it's just the combination of a couple of bad factors together... We have the CPUs themselves, that draw more power than expected (AMD had to bump up the TDP unexpectedly, perhaps board designs weren't modified for this)... then we have the stock coolers that aren't designed to generate airflow around the socket... and to top it all off, we have cheap components that run too hot, because they are over-stressed.
I think Anandtech or some other site should try modifying an Intel stock cooler to fit on an Am2 system (or perhaps find an aftermarket cooler that closely resembles the design, I think there are a few, don't know if any fit an Am2 socket), and then measure if the surrounding components are cooled significantly better.
 
For those who know, and it seems many here know more then me about the power sub systems in a motherboard, how is the asus m2n32 sli delux wifi board? Is it goign to handle a 9600? Or am I waiting for it to burn out?

Thanks
 
Well here is a question. I was thinking of buying a 780G motherboard and putting in an Athlon 5000+. A year down the road when Phenom's hit 45nm, I could upgrade to that without having to buy new memory and a motherboard. There are two problems.

1) Will the 45nm Phenoms still be using the same AM2+ socket?
2) The 780G chipsets may not be able to handle the power required by the current Phenoms but I would assume that the 45nm ones would use less power.

What are the chances that I successfully be able to upgrade in the future?
 
See the below about Sapphire mobos-

http://www.rage3d.com/board/showthread.php?threadid=33921257

Supposedly their 780G board will even handle the new 9850 well.

From the link:

"Of particular note is that SAPPHIRE has full support for these new CPU’s on its 780G product, PI-AM2RS780G. This is NOT the case for the majority of competing 780G designs as they have used a far less capable configuration for the power sections and will only be able to support the earlier 95W CPU’s. (9500, 9550, 9600, 9650)"

So most 780G mobos are limited to 95W CPUs. If you want an Agena for a board that is restricted to 95W CPUS that means you'll use a 9600 or lower and you will not be able to OC it without risking the health of the board.
 
Looks like this board just popped up here at Newegg. A full ATX 780g board with...whats that?....4+1 power phase. Why not just call it 5? either way just wanted to give a heads up as this should be looked if you want to run the higher end Phenoms with a 780g and you have not upgraded yet.
 
When the quad-cores go 45nm, it's possible to see them drop to the 95W CPU level. Like everything with AMD, we just have to wait.
 
Given this is about the 780G boards, i have spent a while now trying to find out anything about the Foxconn 780G board. Anyone have a source that lists what kind of power phase it has. I dont really know much about Foxconn but i think they are mostly on the Intel Camp correct? Looks like a good board and i am looking to set up a new computer hoping to get a matx that can handle the 9850. I think the Gigabyte looks to be the best bet if this foxconn cant handle it either so with a scythe zipang and enzotech mosfet coolers the Gigabyte should make it :D


Any Foxconn info would be appreciated thanks
 
In relation to the Gigabyte mATX 780g board, just installed a 9850 on it and have had no problems so far. One thing i have noticed is the hot mosfets. Using HWmonitor one of the sensors says 85C which i think is for the mosfet temp (not counting the other two detected sensor with safe temps at 34 and 38c)??? He is ordering mosfet heatsinks and will have a top-cooler with a 120mm to blow on them as well. I will post back if there encounters any problems but so far so good at 2.7 stock voltage.
 
You must log in or register to reply here.
Back
Top