Overclock.net › Forums › AMD › AMD Motherboards › MSI A88X-G45 GAMING - FM2+ goodness
New Posts  All Forums:Forum Nav:

MSI A88X-G45 GAMING - FM2+ goodness - Page 7

post #61 of 99
Quote:
Originally Posted by WackyWRZ 
Still seeing a 0.1v swing at load, which is better than the 0.2v swing I was seeing at idle.

Sounds like your board is defective.
Quote:
Originally Posted by WackyWRZ 
bouncing between 3.5/3.7 GHz. Is this what is meant by "throttling"?

Yes your getting throttled.
Quote:
Originally Posted by Formula350 
Ah, yes, the two at the top under the memory phases. I would suspect that'd not be the actual MOSFETS but still no doubt would've been ideal/beneficial to be cooled. We'll see how badly my phases heat up without airflow... Idling in the BIOS my CU is only 34C (on default-failed OC settings)

They are mosfets, they are the missing 2 chips from the photos of the mosfets on the top side, there are 22 of them in the photo's of the board, leaving just 3chips per NB phase, Overclock your NB a bit and bump her voltage then give um another feel test. It is possible MSI felt that the mosfets where overbuilt for the NB bridge, but I don't find that to be a good excuse.
post #62 of 99
I am in Utah, got my board same day from Newegg as well. Not sure where it actually shipped from.
post #63 of 99
Quote:
Originally Posted by WackyWRZ View Post

EDIT: Forgot to mention that while running P95 I also was running CPUz and HWMonitor. CPU temp was around 62 degrees and I was clocked at 4.3 GHz. I think I did have turbo and CnQ turned on, and I did notice that after about 10 seconds I would see the CPU drop down and start bouncing between 3.5/3.7 GHz. Is this what is meant by "throttling"? (Sorry if this is a dumb question - I'm still trying to learn smile.gif )

If you're using the IGP, then yea that'd be throttling. If you have AIDA64 (one ot the latest Betas, since that's what I work with), run the Stability test and see what it throttles down to. For me, with the iGPU enabled and having it selected in the Stability boxes in upper left, it'd throttle down to 3000MHz no matter what frequency I'm running at. GPU doesn't ever throttle, just the CPU clocks, even if the Stability test is ONLY running on the GPU! This is the cTDP crap, and for me does it regardless of whether CnQ or CPB is disabled or enabled.

Is your overclock manual, or from OC Genie?

~~~~~~

So I did some additional stressing and that cTDP really prevented me from loading the MOSFETs as much as I had wanted I personally think. Even if it was technically drawing the same, those heatsinks were only warm tot he touch, which for comparison the FCH (assuming they're still calling it Fusion Controller Hub) was a little bit warmer then the FET's sinks.

And this is kind of BS IMO... I can load of LinX and run it through while monitoring the AIDA Power usage, and either AIDA is off, or something smells ... Idle (stability window open and monitoring) is ~21.28W for the whole package. Running LinX is... 21.60W for the whole package (22.95W running Prime95 Tortue Test #2) :| Just the GPU loaded and the entire package consumes... ~40-45W. 7850's max package rating is 95W, yet 45W+22W = 67W, but my CPU throttles? :\ We've figured out it isn't temps, the Gigabyte boys have figured out it isn't temps. So did AMD get a bit wrong in the CPU, or is this something that can be rectified in a Microcode update if it IS in fact not supposed to be doing this? Pre-Post Edit: Ok I'm starting to think AIDA's output isn't totally accurate now, or it doesn't output correct unless using their stability test, because LinX and Prime both are not registering one the Compute Units for any power draw. Looks like I'll have to try and do the below method <_> Yay Math!

I'll have to run this one AMD program I managed to grab (back when I found PSCheck and spread it around) that lets you monitor and change any register, to see what the CPU is actually saying (provided I can decipher it). Here's what their BIOS Dev guide says about cTDP:
2.5.2.1.1.1 TDP Limiting
TDP limiting is a mechanism for capping the power consumption of the processor through a configurable TDP
limit. APM varies the P-state limit of the processor to keep processor power consumption close to the TDP
limit within a thermally significant period. See 2.5.2.1.1 [Application Power Management (APM)]. TDP limits
can be set by BIOS or APML.
• TDP limiting is enabled if D18F4x15C[ApmMasterEn]=1 .
• TDP limits are specified in units of power credits. One TDP unit equals one power credit. TDP units can be
converted to watts by multiplying by the scaling factor MSRC001_0077[Tdp2Watt].
• APM regulates the total processor core TDP. The maximum non-core TDP is specified by
D18F4x1B8[BaseTdp]. The maximum processor TDP is specified by D18F4x1B8[ProcessorTdp]. The maximum
total processor core TDP limit can be computed as D18F4x10C[NodeTdpLimit] * (D18F3xE8[Multi-
NodeCpu] + 1).

• The minimum total processor core TDP limit is the power consumed when all cores are in
D18F3xDC[PstateMaxVal]. See 2.5.3.1.1.8 [Notification of TDP Limit Changes].
• There are two sources of TDP limits in the processor. APM uses the lower of the two limits to determine the
effective total core TDP limit for the processor.
• The maximum sum of TDP for all cores on the processor is D18F4x10C[NodeTdpLimit] *
(D18F3xE8[MultiNodeCpu] + 1).
• An APML defined percentage of the maximum processor TDP limit, MSRC001_0075[ApmlTdpLimit-
Percent].
• The highest performance P-state available to software varies with the TDP limit. See 2.5.2.1.1.2 [Notification
of TDP Limit Changes] and 2.5.2.1.1 [Application Power Management (APM)].
• If software requests the highest performance P-state available and processor power consumption remains
within the TDP limit, APM enables transitions to higher performance P-states.
• If processor power consumption exceeds the TDP limit, APM may restrict transitions to P-states that are
lower performance than the software requested P-state.
(As far as I can tell this is not occurring for me

Also, I think this is relevant to what is happening but am not entirely sure if I'll be able to sort it out (I suck at math despite how simple this may infact be lol)
2.5.2.1.1.2 Notification of TDP Limit Changes
APML or BIOS may set a TDP limit that results in limiting the number of P-states available to the OS. When
this occurs, BIOS is notified of the new TDP limit via an SMI interrupt. See D18F4x16C[ApmTdpLimitIntEn].
In the interrupt routine, BIOS determines the highest performance P-state that satisfies the D18F5xE8[ApmTdpLimit]
by performing the following algorithm on all cores of each node. The SMM BIOS should then notify
the OS of the new P-state limit via an SCI interrupt.
1. All cores on a node must do the following before any core on the node can perform step 2:
SwPstateRequest = MSRC001_0062[PstateCmd].
2. Only core 0 does the following before any core on the node can perform step 3:
BaseWatt (real number in W) = ((D18F4x1B8[BaseTdp] * D18F5xE8[Tdp2Watt[15:0]])/(2^16)).
ApmWattLimit (real number in W) = ((D18F5xE8[ApmTdpLimit] *
D18F5xE8[Tdp2Watt[15:0]])/(2^16)).
IF (0 == MSRC001_00[6B:64][PstateEn] indexed by D18F3xDC[HwPstateMaxVal]) THEN
D18F4x15C[TdpLimitPstate] = MSRC001_0061[PstateMaxVal] - 1.
ELSED
18F4x15C[TdpLimitPstate] = MSRC001_0061[PstateMaxVal].
ENDIF.
For (i=D18F4x15C[NumBoostStates]; i < D18F3xDC[HwPstateMaxVal]; i++) {
PstateWatt (real number in W) = (D18F2xF8[PwrValue] * (D18F5x84[CmpCap] + 1)
* (D18F3xE8[MultiNodeCpu] + 1)) / 10^D18F2xF8[PwrDiv].
IF ((PstateWatt
- BaseWatt) <= ApmWattLimit) THEN
D18F4x15C[TdpLimitPstate] = i - D18F4x15C[NumBoostStates]
Break.
ENDIF.
}
3. All cores on a node must do the following:
MSRC001_0062[PstateCmd] = SwPstateRequest.


I never got around to doing that today... So I'll just leave this here and try and get to that crap tomorrow lol
post #64 of 99
Quote:
Originally Posted by luminousone11 View Post

They are mosfets, they are the missing 2 chips from the photos of the mosfets on the top side, there are 22 of them in the photo's of the board, leaving just 3chips per NB phase, Overclock your NB a bit and bump her voltage then give um another feel test. It is possible MSI felt that the mosfets where overbuilt for the NB bridge, but I don't find that to be a good excuse.

I can only get 2000MHz out of my NB, no matter how much voltage I throw at it, it seems. 20x multi goes into Windows just fine, does what I need it too (pretty sure it's stable). Whereas 21x multi won't even POST frown.gif
post #65 of 99
Quote:
Originally Posted by Formula350 View Post

I can only get 2000MHz out of my NB, no matter how much voltage I throw at it, it seems. 20x multi goes into Windows just fine, does what I need it too (pretty sure it's stable). Whereas 21x multi won't even POST frown.gif

Have to change the bus clock, default is 100mhz, Do be warned, doing this overclocks everything, but it might give you more room on the NB.
post #66 of 99
Listed as cpu base frequency in the MSI bios.
post #67 of 99
Quote:
Originally Posted by Formula350 View Post

If you're using the IGP, then yea that'd be throttling. If you have AIDA64 (one ot the latest Betas, since that's what I work with), run the Stability test and see what it throttles down to. For me, with the iGPU enabled and having it selected in the Stability boxes in upper left, it'd throttle down to 3000MHz no matter what frequency I'm running at. GPU doesn't ever throttle, just the CPU clocks, even if the Stability test is ONLY running on the GPU! This is the cTDP crap, and for me does it regardless of whether CnQ or CPB is disabled or enabled.

Is your overclock manual, or from OC Genie?

I've got an Athlon 760K so there isn't an IGP on it. I've not heard of or ran AIDA, but I can download when I get home and give it a go. OC is manual, I have an MSI FM1 board with the "genie" on it that I tried before. Long story short - I don't trust it one bit.
post #68 of 99
I am RMAing mine back to Newegg. I had the voltage fluctuation thing, and did not like the hot power components on the REAR of the board. The bios did not seem to have all the options it needed either.

Now the thing is what board to replace it with? I need a good top of the line ATX motherboard for my FM2+ processor, what do you folks think?

Thanks in advance,
Dan
Edited by DanStp - 3/28/14 at 3:30am
post #69 of 99
I have been testing the Gigabyte F2A88X-UP4 board alongside the MSI, and can say the Gigabyte is a really nice board - especially when its ~$20 cheaper than the MSI. It has the IR VRMs which are known to be of very high quality and it shows. It lacks some of the "frills" of the MSI (Killer NIC, Audio Boost, etc...) but I don't think either of those are worth trading for the unstable board. I have not done a whole lot of overclocking yet, but I am currently at 4.3 on my 760K with no throttling or any other issues. The VRM heat sinks actually stay cool to the touch also.

The only other board I can think that would be really high end would be the Asus A88X-PRO board - and it carries the high price tag to match. I actually had that board before I tried these two and it was DOA, so I RMAed it to Amazon and got another problematic board so gave up with Asus after reading about their hit/miss quality and RMA issues.
post #70 of 99
The Asus board is pretty nice, the Digi VRM it uses while not as good as the Gigabyte board, is extremely close.

I will tip my hat to the asrock extreme6+ board as well, but the heatsink on its vrm is a joke, so you need to be willing to buy an after market heatsink for it. the extreme6+ also has gold premium caps and similar audio separation tricks as the audio boost on that msi board. Its a good board other then its woefully pathetic vrm heatsink(you listening asrock, your board could be perfect with one minor change, and hell with that change you could ask 20-30% higher price and easily get it as well, its things like this that could move you from low/mid tier brand to high end premium maker).

I bought the Gigabyte board btw, between the powIRstages and the solid caps it seems to be the best board atm, The only fault i will give it, is the uefi bios gui is kinda spartan looking(reminds me of the pre uefi days), and this is a relatively minor thing really, the needed settings are all their so it is completely functional.

The MSI board a88x-g45 seems to have been pulled off newegg, Hopefully they will make a revision to the board and reintroduce it to the market, with a few changes it has the makings for greatness. I don't want to be to overly critical of MSI either, every maker has an off product every once in awhile, And so many parts of that board where extremely well done, if they fix the vrm issue, it would certainly be a fit to stand among the best fm2+ boards on the market.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: AMD Motherboards
Overclock.net › Forums › AMD › AMD Motherboards › MSI A88X-G45 GAMING - FM2+ goodness