der8auer examines AMD X570 chipset power consumption

[HEXUS]

Asserts it doesn't need a fan, and consumption isn't high due to PCIe Gen 4 device utilisation.

Read more.

[Tunnah]

Why would it be an m.2 that pushes the power up. Lane wise, it is insignificant. Put in a pair of x16 gen 4 cards and rev them up. The difference between 3 and 4 is so tiny, because percentage wise, as a piece of the entire pie, the amount is minuscule

[kompukare]

Why would it be an m.2 that pushes the power up. Lane wise, it is insignificant. Put in a pair of x16 gen 4 cards and rev them up. The difference between 3 and 4 is so tiny, because percentage wise, as a piece of the entire pie, the amount is minuscule

Yes, but if an x16 graphics card only losses 1-2% of its potential speed going from PCIE 3.0 to 2.0, it follows that it will be very very hard to get a graphics card to really push PCIE 4.0.
NVMe drives on they other hand are currently using all which a PCIE 3.0 x4 slot can deliver.

[afiretruck]

Yes, but if an x16 graphics card only losses 1-2% of its potential speed going from PCIE 3.0 to 2.0, it follows that it will be very very hard to get a graphics card to really push PCIE 4.0.
NVMe drives on they other hand are currently using all which a PCIE 3.0 x4 slot can deliver.

There is a benchmark that AMD used to demonstrate the difference PCIE 4.0 makes over PCIE 3.0 with the Navi GPUs. Not sure if this benchmark actually saturates PCIE 4.0 x16 but it's a good start.

[flearider]

I think he may have forgot something ..the gen4 nvme was run on it's own .. and got a load of 8.86 .. now if you ran that with the bottom yellow line would you not have to add them up ?
or is it a case of one or the other ?

[kalniel]

There is a benchmark that AMD used to demonstrate the difference PCIE 4.0 makes over PCIE 3.0 with the Navi GPUs. Not sure if this benchmark actually saturates PCIE 4.0 x16 but it's a good start.

If you create a benchmark that just tests bandwidth then yes, you see an improvement with a PCIE 4 card on PCIE 4 over 3. That's all it does, it doesn't suggest there is any kind of improvement when bandwidth is not the limit, which it isn't in any game or other test.

[LSG501]

I think he may have forgot something ..the gen4 nvme was run on it's own .. and got a load of 8.86 .. now if you ran that with the bottom yellow line would you not have to add them up ?
or is it a case of one or the other ?

Assuming the yellow is the x570 as well, you can see the nvme as 8.2 and 8.55 versus nvme gen4 at 8.63 and 8.86. A little bit higher but not massively higher, although obviously the impact of other items isn't measured in the picture.

[Corky34]

Going on AMD's philosophy on why they still use blowers on their GPUs I'm guessing they only made the fan on the PCH a requirement because they can't know if the system it's going in has decent airflow.

[Tabbykatze]

No disrespect to der8auer but isn't it under nvme pcie 4.0 raid that is the heat generator, he is not utilising it fully so he is not getting the results defined by the board vendors at comoutex...

[huggi]

Why would it be an m.2 that pushes the power up. Lane wise, it is insignificant. Put in a pair of x16 gen 4 cards and rev them up. The difference between 3 and 4 is so tiny, because percentage wise, as a piece of the entire pie, the amount is minuscule

Because the PCIe lanes for the x16 slots come from the CPU and not the chipset, so using graphics cards won't stress the chipset at all. The only things that the chipset controls are the secondary NVMe slots (AFAIK, the primary slot is still direct to CPU), SATA, USB, Ethernet/WiFi, and other peripherals.

[afiretruck]

If you create a benchmark that just tests bandwidth then yes, you see an improvement with a PCIE 4 card on PCIE 4 over 3. That's all it does, it doesn't suggest there is any kind of improvement when bandwidth is not the limit, which it isn't in any game or other test.

Yes, I am well aware of that, thank you.

I meant that the benchmark would better test the power consumption of PCIE 4.0 compared to 3.0 (the topic of this post), as it could happily saturate the bus.

Of course that was before I remembered the GPU slots are directly wired to the CPU and so are inconsequential regarding chipset power consumption.

[Tabbykatze]

Now this is interesting: https://www.techradar.com/uk/amp/new...y-power-hungry

Might be a turn around point for Steves aggravating remarks about power consumption. If the chipset is also pulling power from the EPS lines, that could explain the on the wire differences from stated TDP versus power consumed.

[simonpreston]

All I need to know is am I going to lose much performance, sticking a 3000 in to a 370x mobo? Because, if not , I'll be happy to stick with that.

[Tabbykatze]

All I need to know is am I going to lose much performance, sticking a 3000 in to a 370x mobo? Because, if not , I'll be happy to stick with that.

The performance differences for the CPU are negligible/margin of error for the x470. Not seen many do it for x370 but likely to be the same unless you overclock.

[Corky34]

All I need to know is am I going to lose much performance, sticking a 3000 in to a 370x mobo? Because, if not , I'll be happy to stick with that.

Not currently although it's possible that could change going forward as the development focus will mainly be on the new boards, an AGESA update could change the landscape.

[simonpreston]

Cheers. I'm happy with the features my Asus Prime 370 has. Doubt I'll be buying a PCI-E 4 card any time soon. Certainly not one that'll saturate it. So I'll stick with my 370 for a few years.