A10 6800K 'Richland' Power consumption and efficiency

[csbin]

http://techreport.com/review/24954/amd-a10-6800k-and-a10-6700-richland-apus-reviewed/3

 

[NTMBK]

The 6700 is about the same performance as a 5800k, so those efficiency gains are actually pretty impressive.

 

[frozentundra123456]

The 6700 is about the same performance as a 5800k, so those efficiency gains are actually pretty impressive.

True, but if you read the whole article, the conclusions are still pretty lukewarm about the product in general, and it did use more power than the intel chip with similar TDP.


Edit: The best advantage seems to be in highly MT productivity workloads vs the i3, and of course in igp perfomance, if anyone is interested in that. Despite the apparent demise of the i3 posited by some posters, it still offers gaming performance slightly better than Richland when used with a discrete card in the games tested in this article.

 

[Zor Prime]

Those low power usage numbers for the AMD chips are very surprising to me.

 

[NTMBK]

True, but if you read the whole article, the conclusions are still pretty lukewarm about the product in general, and it did use more power than the intel chip with similar TDP.

Oh yeah, the overall performance/W is still lagging behind Intel noticeably. But the improvement from essentially the same silicon, on the same silicon, is pretty nice. Shame they didn't ship Bulldozer like this in the first place!

 

[Enigmoid]

The power consumption numbers for the i3 seem weird. Idle seems way too high. I know my laptop i7 quad when idling at 3.2 ghz uses around 5-6 watts with a few internet tabs open according to HW monitor. Here the i3 is using about 10 watts more than the apus. Also really weird is that the peak power of the i3 (igp) is 57 watts or only 21 watts higher than idle.

 

[sm625]

It has low idle power. It would seem to be a nice chip for a system that is always on. Like a kid's system. But you know how kids just stop using a computer and then run outside and play. Problem is if they leave a game running then its not idling, and it is probably chewing up 80 watts rather than 25. And of course since it is in a game it never goes to sleep.... its not until 2 hours later that you realize they left the game running, then you have to go lecture them for wasting 3 cents worth of electricity...

 

[NTMBK]

Those low power usage numbers for the AMD chips are very surprising to me.

Trinity has very good idle power- it has a few tweaks similar to Haswell, in that the CPU's power management can tell other parts of the platform to sleep. (Similar concept, though obviously not as refined as Haswell's implementation.)

 

[BSim500]

Despite the apparent demise of the i3 posited by some posters, it still offers gaming performance slightly better than Richland when used with a discrete card in the games tested in this article.

No idea where these "the i3 is dead" fantasies come from other than impatience that Intel typically release them in Sept rather than June . New Haswell i3's include the 3.6GHz i3-4340 which will have roughly 15-20% performance boost over a current 3.3Ghz i3-3220/3225 along with another 5-8w lower idle power consumption reduction. With a slight +5% BCLK overclock to 3.78GHz, it's single-thread performance will be roughly equal to a 3.9-4.0GHz Ivy Bridge, which in turn requires a +5.0-5.5GHz AMD chip to match core-for-core in many games at which clock rates AMD's power consumption goes through the roof (see 220w 5GHz FX-9590) - far higher than on listed charts.

If anything, i3's see an even bigger jump on Haswell's than i5's due to a +200MHz stock clock speed boost (current fastest i3-3240 3.4GHz gets upped to 3.6GHz i3-4340 (on top of Haswell's +5-10% IPC improvements)) and now include AES-NI (previously allowed only on i5/i7's). Far from "dying", the new i3's are actually looking better & better, across the whole overall performance & power usage envelope.

As for discrete cards, from page 8 of same review:-

Tomb Raider:-
72fps = A10-6800K
87fps = i3-3220 @ 3.3GHz
114fps = i5-3470
95-100fps (estimated) = i3-4340 Haswell @ 3.6GHz
http://techreport.com/review/24954/amd-a10-6800k-and-a10-6700-richland-apus-reviewed/8

Given Haswell's observed IPC improvement and 200-300Mhz clock boost, an i3-4340 is going to be 25-35% faster in many games than an A10-6800K - with half the cores and a 500-800MHz clock speed penalty (and another 10% reserve headroom to go before hitting the 3.9GHz stock turbo freq of 2-cores loaded i5's).

 

[Insert_Nickname]

The power consumption numbers for the i3 seem weird. Idle seems way too high. I know my laptop i7 quad when idling at 3.2 ghz uses around 5-6 watts with a few internet tabs open according to HW monitor. Here the i3 is using about 10 watts more than the apus. Also really weird is that the peak power of the i3 (igp) is 57 watts or only 21 watts higher than idle.

That's properly because your mobile i7 run at lower vCore and can downclock to 800MHz. The i3 can only idle at 1600MHz.

The 6700 is about the same performance as a 5800k, so those efficiency gains are actually pretty impressive.

They are.

 

[richierich1212]

With a slight +5% BCLK overclock to 3.78GHz, it's single-thread performance will be roughly equal to a 3.9-4.0GHz Ivy Bridge

I thought we couldn't adjust blck at all on non-K Haswell processors?

 

[Yuriman]

I thought we couldn't adjust blck at all on non-K Haswell processors?

I think they just don't have bclk straps.

 

[Enigmoid]

That's properly because your mobile i7 run at lower vCore and can downclock to 800MHz. The i3 can only idle at 1600MHz.



They are.

Nope. Under maximum performance idles at 3.2 ghz and voltage is 1.01-1.07 according to cpu-z. According to HW monitor it uses around 5-6 watts for the package at idle.

 

[BSim500]

I thought we couldn't adjust blck at all on non-K Haswell processors?

I thought it's only the straps / gearing ratio's (1.25x, etc) you can't adjust on non-K chips? You should still be able to squeeze 103-107MHz BCLK out of it though (extra 100-250MHz) before instability sets in (may need a Z motherboard).