AMD FX-9370 on 990FX Killer

WKjun wrote: Dear parsec! Dear Moderator Group! Dear users!   This is a really interesting consideration in many ways, parsec! I consider this thread a perfect fit to discuss what components are needed on a mainboard to use that CPU.   I'm totally with you, parsec, that this CPU can regrettably be plugged into any AM3+ socket, it shoudn't be used on boards with poor VRMs/cooling and I totally understand that ASRock prohibits its use on unsupported boards.   My experience is though, that if the BIOS is not supporting a CPU, thus not recognizing it, either no POST screen shows up, or wrong/lower clock rates are set, which are hard or impossible to correct manually. Therefore I wonder, why so many people were able to use it without running into configuration troubles! I can only think of a mistake by ASRock in accidently having it supported at BIOS level, while correctly not listing it in the CPU support list? If that is true, I don't take your point blaming AMD, while board vendors are unobservant and accidently include the CPU in their BIOS programming.

The answer is the BIOS simply did not block the use of an FX-9000 series processor. If that was a simple mistake or more likely an oversight, I don't know. I can only theorize about why the FX-9000 series processors were allowed to be used on boards until recently, that could not handle them long-term, or when stressed. Remember, these are only theories and guesses:

• AMD thought they would be fine to use with any half way decent board. AMD may have told mother board manufactures this. It was later that this was discovered to be false.
• AMD is not doing very well with sales, and hoped a CPU that could do 5GHz would attract attention and sell very well. They also did not want to limit sales to a small group of boards. But in this case, AMD's wide range of compatibility with their CPU sockets and chipsets caught up to them because the FX-9000 series processors cannot be used safely in a low or midrange board.
  
• Imagine if one mother board manufacture did not block the use of an FX-9000 series CPU on any of their boards, but another manufacture did on some models. Then some people would think the manufacture that did block them made bad, cheap boards, even though that manufacture was doing the right thing.
• I think that most people (outside of AMD) really knew how much power these processors would actually need, and were surprised when boards started failing. Also as you experienced yourself, if you don't use a "top down" type CPU cooler that blows air across the VRM heatsink and chips, the CPU VRMs can over heat and fail. So the users of these CPUs did not configure the cooling of their boards properly.
  

WKjun wrote: Really interesting by the way, that even review sites chose that combo, without knowing of its absent support: http://www.tweaktown.com/articles/5989/asrock-fatal1ty-990fx-killer-s-native-pcie-ssd-performance-preview/index3.html http://benchmarkreviews.com/21005/asrock-fatal1ty-990fx-killer-am3-motherboard-review/5/ So it seems certain that the FX-9000 series must have been supported at BIOS level, otherwise they would have reported the issue or used other parts and why else should there be a BIOS release to correct this?   Besides why the CPU previously worked at BIOS-level, I begin to wonder why it shouldn't work at hardware-level! Even though I do not have the same problem, it correlates to my recently posted problem (http://forum.asrock.com/forum_posts.asp?TID=660&title=970-performance-fx9590-overheating-ram-incomp)   How come that both the Fatality FX990 Killer and 970 Performance has "Digi Power  8 + 2 Power Phase Design" but only one of them supports 220W CPUs? It cannot be the phase count alright, although it is said that more phases can handle more load more stable, i.e. 3 phases cannot handle 125W CPUs, but 5 phases can. I never heard of differences between how they were realized. So what's the deal with 8+2 Phases which allegedly cannot deal with 220W? Is it the VRM heatsink? They seem quiet similar.   This knowledge could appease your customers and clear things up! Beer

I'm so glad you brought this up, it's a topic I've wanted to talk about for a long time!

My father was an electrical engineer, and I was taught enough to know something about this.

As is common with people when dealing with complex topics like computers and electronics, that takes a lot of training to truly understand, we try to make things simple so we can understand and discuss them. But making things too simple causes mistakes in our reasoning and conclusions.

For example, in general a "phase" is one chip used in the CPU VRM stage to convert +12V DC into the ~1.5V DC maximum that is used by a CPU. That is a bit different for Intel Haswell CPUs, but in general is the same.

These chips have a limit to the amount of "power" (Amps or Watts) it can produce without over heating and destroying itself. So multiple chips or "phases" are used because one chip alone cannot produce enough power for most CPUs. Then if we OC that CPU, it uses even more power, and even more chips are needed to provide the necessary power. Otherwise the VRM circuit reaches its limit for power output, or over heats trying to produce more power than the heat sink can keep all the chips/phases cool.

Then some people start analyzing the VRM circuit, and say it really does not have, for example, eight phases because of the way the circuit is designed with its drivers or doublers. IMO, that is where people make mistakes about what is "better" or what is "worse", and overlook what is really important.

Imagine I have a CPU VRM design that could produce 300 Watts of power for a CPU, and not over heat or fail. Does it matter if it has one phase or 25, as long as it does the job? Transistors which are used in the VRM chips have different power capacities, which means the chips have different power capacities. So what is better, one 100 Watt chip, or ten, 10 Watt chips? My point is counting phases without knowing what each phase is capable of, does not tell us very much.

So the 8 + 2 phases on one board may not be equal to the 8 + 2 phases on another board. One phase is not a standard of power output, it refers to one chip in the VRM circuit.

Actually, IMO the best test of a VRM design is if we could measure how much power in Watts it could produce continuously without over heating. Doing that is very difficult, and requires special equipment and knowledge of electronics and the design of the VRM circuit.

I noticed in the specs of the ASRock X99 OC Formula it states about the VRM design, "12 Power Phase design (Supports up to 1300w)". Finally an actual power output specification!

While more phases can make for a more efficient VRM design (less wasted power), IMO should I care if one board uses a doubler to drive two chips, and another doesn't, if both produce the same amount of power to a CPU?

The heat sink on the VRM chips is very important. The chips power output vs temperature is shown as a graph in the specifications for the chip, or for a single transistor. You measured how hot the chips or VRM heat sink became, and posted that in your thread about your board. They over heated, failed, and apparently damaged your CPU!

I just realized something, how many electronic devices like audio amplifiers have fuses in the circuit that open/blow when a certain amount of power is being used by the circuit, so it won't be destroyed if more power is drawn through the circuit? Most of them do. We never see that in mother boards, it might be a good idea.

Enough of this, if you read it all, thanks, I hope it made sense, feel free to comment away!!

Edited by WKjun - 19 Aug 2015 at 3:28pm

It seems that support for the FX9000 series was removed from the bios in the latest update. The only thing I know will work is to get your hands on an AM3 CPU that is supported and install it then flash back to your previous bios.

And this just happened to me, too.  And boy, am I angry!  Can't enter UEFI to try to reflash it or anything.  Aren't we just screwed?