AMD - Piledriver chitchat

[HalloweenJack]

Would be nice to see cf results with a 6670 on the new richland

[kompukare]

Idle power doesn't look that impressive, especially as AMD managed to get FM1/FM2 to idle lower despite being more power hungry at load:

http://www.computerbase.de/artikel/p...cs-im-test/17/

For some reason AT missed that, but anyway since all these figures are for the complete system whatever the changes to the VRM were meant to achieve seem to have close to 0W difference.

Edit: Hardware.fr do measure of ATX12V (you need to select it though since the graphic defaults to 220V full system):
http://www.hardware.fr/articles/897-...ergetique.html

Well with the CPU circuitry idling down to 3W, VRM is probably all that can still be done. But on the other hand mobo choices make a huge difference:
http://www.anandtech.com/show/6964/g...77hd4-review/3
The same CPU and the same GPU but 61W vs 113W; that's 52W difference at idle.

[CAT-THE-FIFTH]

The Iris Pro Intel test system is meant to simulate a laptop as I mentioned previously,so it will be an efficient system.

[Jimbo75]

Idle power doesn't look that impressive, especially as AMD managed to get FM1/FM2 to idle lower despite being more power hungry at load:

They used a 1500W PSU so the results are going to be a bit whacked because of that.

[watercooled]

12v connector measurement isn't a great way to measure CPU power as some CPUs will use power from the 24 pin connector for the uncore.

As for reducing power, ATX power supplies and motherboards are generally the biggest culprits here. You only have to look at laptop power consumption to see the difference it can make.

ATX power supplies, even '80plus x' tend to have rubbish efficiency below 10% or so, in fact a lot of them will eat 10-20W (sometimes far more) powered up but not loaded. That alone can double the DC load. And then you have the trend of using something like a 1.6kW PSU to power a core i5 and a mid-range GPU - higher wattage PSUs tend to be even worse.

Then motherboards, while things are improving, there's still some way to go. Efficiency of components doesn't seem to be that high on the priority list for motherboard MFRs, and it seems fairly common to use a load of inefficient linear regulators to power on-board devices, often needlessly re-creating voltages already available from the ATX PSU itself. Maybe it's easier/cheaper to just copy/paste some building block schematic to a new board rather than hand-tweaking it, or maybe an LDO is easier/cheaper to use rather than routing another power trace to that section? Who knows.

But in terms of wasted idle power, even desktop CPUs seem to be doing fairly well now, starting to scrape away at the last few single-digit watts.

Don't get me wrong, I don't like to over-exaggerate small differences in power consumption (and it's important to remember sample variance/testing error margins), but it's annoying when something could be fixed by a simple change to PCB layout.

[inf64]

Source

Originally Posted by The Stilt
Had a better look at the Gigabyte sheet.

It says "AMD next generation FX".
Vishera is no next gen, it is almost two years old already

I assume Gigabyte refers to "TJ", which still should pop into AM3r2.

The sheet is for Rev 3.0 990FXA-UD7 which has some changes to the VRM.
Yet it appears to be a multiplied analog VRM.

Ah man, new leaks . What is "TJ" ? It should be AM3r2 compatible and the new board is Rev 3.0 with changes to VRM . Is "TJ" modified Piledriver core?? Why at 5Ghz and 220W tho? So many questions so little answers

[CAT-THE-FIFTH]

A picture of the Temash tablet:

http://liliputing.com/2013/06/gigaby...emash-cpu.html

Edit!!

This tablet looks better:

http://liliputing.com/2013/06/quanta...cpu-video.html

[HalloweenJack]

142w under load for an overclocked 6800K @ 49 degree`s - clocked at 4.8ghz / 1013mhz gpu.

ram support is 2133mhz


edit:

http://www.techspot.com/review/681-a...6800k-a4-4000/

another review

[kompukare]

They used a 1500W PSU so the results are going to be a bit whacked because of that.

Yes, 1500W (or 1250W which is what Anandtech list) is excessive but surely the huge difference between the various motherboards would still stand even with a smaller PSU; unless there is some kind of PFC effect in terms of the inefficiency of using such an OTT power supply and some boards make that worse?
Computerbase actually specifically mention changing their PSU for a be quiet! Straight Power E9-400W and that they now can get 40W idle even with a high-end GPU.
(Source)

As for reducing power, ATX power supplies and motherboards are generally the biggest culprits here. You only have to look at laptop power consumption to see the difference it can make.

PicoPSU (DC>DC) might be an answer but there's little choice there. Or a standard for an external GPU so that a docked laptop is not limited to whatever is built in, but ExpressCard hacks or even ThunderBolt don't really work that well.

[watercooled]

PicoPSUs can improve greatly on an ATX power supply - with a modern PC, that 30 or 40W idle could drop to around 20 with a PicoPSU and decent power brick. Out of interest I tried one on my Microserver and it dropped idle power from 28 to 21W.

It's all well and good shouting 50% load efficiency from the rooftops, but in reality, when you have even a 400W PSU, how often are you going to be pulling 200W from it? A heavy focus on half-load efficiency makes sense for servers or for those few PCs which are loaded constantly for whatever reason, but most PCs will spend a great deal of their time idle, so the less power wasted here the better. The 80Plus rating is somewhat abused IMO - MFRs can get a shiny gold badge but hush up the fact their PSU is only 70% or less efficient with a modern system idling.

[CAT-THE-FIFTH]

Looking at the Richland review,the A10-6700 looks pretty good.It could do with a price drop,but better performance than the A10-5800K and lower power consumption too.

[CAT-THE-FIFTH]

Kaveri is smaller than Richland and Trinity:

http://www.hardware.fr/news/13148/co...oto-da-mo.html

FM2+ socket detailed:

http://translate.google.com/translat...-compatibilita

[shaithis]

Hmm, hope AMD start sticking with an APU socket for a while. Seems nuts that AM3+ has been around long enough to see FM1, FM2 and FM2+

While Intel are the masters of depreciating CPU sockets, AMD are most certainly trying to take that mantle.....

[CAT-THE-FIFTH]

AMD might be also including TSX at some point too:

http://www.sisoftware.co.uk/?d=qa&f=ben_mem_hle

[watercooled]

According to Charlie, the k versions of Haswell forgo those extensions?

Ironically, it seems the 'enthusiast' parts are the most butchered.

[CAT-THE-FIFTH]

According to Charlie, the k versions of Haswell forgo those extensions?

Ironically, it seems the 'enthusiast' parts are the most butchered.

Intel product segmentation at its best.