Caporegime
TDP isn't power draw though.
Although they'd be drawing more power than they're giving off.
Although they'd be drawing more power than they're giving off.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
AMD not adhering to 125W spec with FX8350?
- Thread starter mmj_uk
- Start date
The more power which is drawn the more heat needs to be dissipated, it's common sense really.
FX8350 is drawing more power than they should 'officially' and hence thermals/TDP are above what they should be, so vendors making value conscious motherboards are finding that there boards are not stable under load due to VRM's overheating/being overdriven and are implementing throttling mechanisms to protect them.
If this was a case of 'rubbish MSI' as you say then why is the FX8320 (also rated 125W TDP) apparently not affected?
Last edited:
Soldato
As Martini and Humbug said above. You need to look at the difference between TDP and power draw. AMD are stating that you need to cool 125W of dissipated heat, not that it draws 125W electrical power.
When I was listing my Saber on ebay at the weekend, I noticed that Asus state that they support 140W CPUs. As mentioned by P0X_.
When I was listing my Saber on ebay at the weekend, I noticed that Asus state that they support 140W CPUs. As mentioned by P0X_.
Caporegime
However, for it to be 125W TDP, it's going to be drawing more power than 125W.
And the boards can't cope.
And the boards can't cope.
Soldato
Exactly.
All the Asus 990X and 990FX boards are listed as having support for 140W CPUs.
All the Asus 990X and 990FX boards are listed as having support for 140W CPUs.
Caporegime
That's because Asus are nails hard when it comes to VRM's 
Exactly
CPU's turn nearly all power input into heat.
You would need a little more than 125w input for 125w TDP though as it is not quite 100%.
It is a close enough correlation though, that 125w TDP or 125w power draw can generally be considered the same thing.
Anyhow take the core voltage above stock and that TDP rises fast. very fast.
125w can easily be 300w on an overclocked chip.
Even more for the WC crowd trying to get stable at 5Ghz with 1.5v+
Soldato
My guess would be around 90% but it depends on the efficiency of the chip, IHS design and how efficient it is at getting the heat out through the IHS. My experience with the 8320/8350 was that it dissipates quite a bit of heat through the socket, or it creates more heat in the socket because it draws more power through the socket than a comparable Intel.
I admit to being lazy on this and not doing my own research.
But!
How the Hell is this power (Current!) supplied by the MOBO??
This has got to be over a Hundred Amps! (Possibly Hundreds of Amps on an OC rig!)
Now, In My world a Hundred Amps is a cable the size of my Pinkie Finger! Where/How is this electrical current created, controlled and channelled on the MOBO? I don't see any humongous circuit tracks anywhere?
Take a look at a pin out of a CPU and see how many pins are marked as VCC
MB's are multi layer so you cant see the power planes.
A 125w chip running at 1.35v vcore is drawing 92.6amps from the VRM's.
That is basically stock on an 8350.
VRM's have to be able to deliver upwards of 200 amps on enthusiast level boards.
It's no wonder MB makers put throttling mechanisms in place rather than spend the money on strong VRM circuitry.
ASUS are one of the few that do give you decent VRM's that have headroom for overclocking AMD BD/PD chips.
MB's are multi layer so you cant see the power planes.
A 125w chip running at 1.35v vcore is drawing 92.6amps from the VRM's.
That is basically stock on an 8350.
VRM's have to be able to deliver upwards of 200 amps on enthusiast level boards.
It's no wonder MB makers put throttling mechanisms in place rather than spend the money on strong VRM circuitry.
ASUS are one of the few that do give you decent VRM's that have headroom for overclocking AMD BD/PD chips.
Caporegime
I admit to being lazy on this and not doing my own research.
But!
How the Hell is this power (Current!) supplied by the MOBO??
This has got to be over a Hundred Amps! (Possibly Hundreds of Amps on an OC rig!)
Now, In My world a Hundred Amps is a cable the size of my Pinkie Finger! Where/How is this electrical current created, controlled and channelled on the MOBO? I don't see any humongous circuit tracks anywhere?
1 Amp x1 Volt = 12 Watts
So for an example if you look at a PSU's 12v rail and it says 60A its 60 x 12 = 720 Watts
The Motherboard converts Amps and Volts up and down and back up and back down again where ever what is needed, the reason why the circuitry which looks and is so delicate apparently carries what are enormous currents is because the currents aren't actually that high until it gets to its final destination where voltage is converted into Amps, like the GPU chip which may pull 50 or 60 Amps and get hot.
Soldato
Soldato
Probably meant 1 amp x 12 volt = 12 watts, but you are right.
AFAIK it is not possible to measure power draw at the socket, I would like to be able to do so, maybe mobo manufacturers can. Therefore I can only do assumptions based on assumptions on power draw at a wall socket.
Efficiency of the PSU, power draw (losses) from motherboard components, power taken by the gfx card in 2D mode and by other components, drives, memory, fans, cables etc.
I kind of arrived at an empirical assessment that the 8350 overclocked at 4.6GHz used about 20-25w per core.
For my machine running everyday 2D office type tasks using one or two cores, the total base load is 135w at the wall. I have assumed that the nett base load excluding the processor is probably about 80-90w.
Full bore, priming, the total base load can be 270w at the wall. This is kind of borne out by adding a concurrent heaven bench on my HD6950 which adds a further 90w for the 3D graphics to 360w. An expected workload for a moderate gaming PC with single gfx.
From that we will have reduction for the PSU efficiency. I would expect any eight core processor at similar frequency to be drawing 200w or thereabouts.
The unknown unknowns are the internal cpu voltages which would give amperage values.
This is not TDP but system power draw, maybe of slight academic interest.
All these results were 60w more on average than a 1055T hex core at 3.8GHz with otherwise identical components.
Last edited:
Soldato
lol, Humbug you muppet!
All the reviews that measure power draw at the plug do my head in, it's kind of useful as a comparison but ultimately lazy IMO that a 'tech' site can't set themselves up to measure at the EATX12v socket.
All the reviews that measure power draw at the plug do my head in, it's kind of useful as a comparison but ultimately lazy IMO that a 'tech' site can't set themselves up to measure at the EATX12v socket.
Soldato
Any gaming machine, Intel or AMD is going to pull a similar amount of power due to the GPU and probably OC on the CPU. A friends 2700K with a gtx570 is very similar to my 8350 in that respect. Otherwise why the need for PSU's larger than 200W? If games only use 2-4 cores as frequently stated, you will not get the extreme draw I have shown through benchmarking prime and heaven.
Anyway this is off topic and does not relate to MSI and VRM's.
Caporegime
He has a point, the FX-8350 draws a lot of power because it has 8 cores, yes in day to day use it never uses 8 cores
MSI didn't help themselves in their first email to that guy by referencing TDP and power draw like they were the same thing, I guess that was Chinese whispers by the customer support guy relaying info from the engineer team. However the question remains why are AMD shipping 140w TDP CPU's marketed as 125w TDP? is this an issue with production where a large number of CPU's are ending up out of spec or is it supposed to be a 140w CPU and the spec is just wrong?
Soldato