AMD - Piledriver chitchat

[watercooled]

Another major reason was because a lot of tasks would only use 3 or 4 of the 5 ALUs available in VLIW5 so there were a ton of shaders taking up die space, sitting idle a lot of the time. Moving to VLIW4 meant the available resources were generally utilised more efficiently, but at the cost of increased power draw given the same core count/clock.

Edit: It does mention that in the Anandtech article, presumably where I read it originally.

[CAT-THE-FIFTH]

The thing is though the Intel IGPs are actually not that energy efficient,but this is masked somewhat by them moving over to more efficient nodes much quicker than AMD can. This is the problem AMD faces,and even using the second most advanced fab technology it does put them at a disadvantage!!

[Terbinator]

The thing is though the Intel IGPs are actually not that energy efficient,but this is masked somewhat by them moving over to more efficient nodes much quicker than AMD can. This is the problem AMD faces,and even using the second most advanced fab technology it does put them at a disadvantage!!

I guess it depends what you mean by energy efficient.

For games - and at the lower res/settings they're attuned to - the Intel APU is propped up by a significantly faster CPU.

[CAT-THE-FIFTH]

I guess it depends what you mean by energy efficient.

For games - and at the lower res/settings they're attuned to - the Intel APU is propped up by a significantly faster CPU.

IIRC,I remember looking at some reviews testing gaming battery life with SB against Llano. During gaming the AMD system had better performance/watt:

http://www.tomshardware.com/reviews/...u,2959-22.html

http://techreport.com/review/21099/a...m-fusion-apu/7

This is the HD3000 IGP against the HD7660G:

http://www.tomshardware.com/reviews/...r,3202-17.html

Both are on 32NM.

I have not seen any comparisons between the HD4000 and the HD7660G though,regarding power consumption during gaming.

It does make me think whether how competitive the HD4000 is in theory despite being on 22NM.

[CAT-THE-FIFTH]

Actually,Hexus does look at power draw under gaming for the desktop CPUs:

http://hexus.net/tech/reviews/cpu/47...0-5700/?page=8







Under a heavy graphics workload the Intel IGP,has worse performance/watt and this is being masked a bit by the process node. The CPU is also consuming less power too,so it just shows you how inefficient the IGP is in reality.

[CAT-THE-FIFTH]

HC reviews an AMD thin and light laptop:

http://www.hardwarecanucks.com/forum...u-arrives.html

CPU performance is OKish,but it seems once you use the IGP too,it compensates somewhat. The laptop is under £400:

http://www.pcworld.co.uk/gbuk/hp-env...aprap&istBid=t

http://www.pcworld.co.uk/gbuk/hp-env...iwptr&istBid=t

Edit!!

Interestingly it seems the AMD 17W TDP A6 has better graphics performance than the 22NM i5-3317U which has an HD4000 IGP:

http://www.expertreviews.co.uk/lapto...vy-6-sleekbook

[scaryjim]

... Interestingly it seems the AMD 17W TDP A6 has better graphics performance than the 22NM i5-3317U which has an HD4000 IGP ...

Not surprising, tbh - it's been fairly well know that the IGP in Intel's low TDP chips is massively thermally limited and can barely get out of its idle clocks most of the time. Given the AMD chip has decent performance, I think it's reasonable to assume that the IGP in Trinity is pretty power efficient and can muster good clock speeds within a low thermal envelope. That'd fit with the earlier speculation about Richland being a 3.2GHz 35W chip by virtue of the IGP sipping power and letting the CPU breathe.

[CAT-THE-FIFTH]

Extensive review of the whole Trinity desktop range:

http://translate.google.com/translat...gleichstest%2F

http://translate.googleusercontent.c...eVFV_XFG6rmErQ

Both the A4-5300 and A6-5400K are included. They test IGP performance and GPU accelerated applications too.

The A6-5400K IGP is faster than the HD4000 in a Core i7 3770K. It appears 1866MHZ does help the CPU performance of the Trinity CPUs too.

It seems the A4-5300 and A6-5400K are around Athlon II X2 250 to Athlon II X2 265 level performance. They hit 4.7GHZ with their A6-5400K sample.

However,they consider the single module CPUs to be too slow for normal usage under Windows??

I find it a bit strange as I used an E4300 and E3300 with 4GB of DDR2 and a 9300 ITX and Windows 8 was fine.

Even the G530 is just about Athlon II X2 250 level:

http://www.anandtech.com/show/5005/h...buyers-guide/2

So in that sense a A4-5300 is not far off a G530.

[DanceswithUnix]

Anandtech has a Trinity buyers guide.

http://www.anandtech.com/show/6449/a...-buyers-guide/

No benchmarks just a bunch of observations and recommendations.

[scaryjim]

... they consider the single module CPUs to be too slow for normal usage under Windows??

I find it a bit strange as I used an E4300 and E3300 with 4GB of DDR2 and a 9300 ITX and Windows 8 was fine.

My primary system at home is an Acer 3810TZ with a 1.3GHz U4100, X4500 graphics, 4GB DDR3 and a 120GB HyperX SSD. It flies for normal usage. These people obviously have no idea what "normal" usage consists of.

[CAT-THE-FIFTH]

The A6-5400K has a better IGP and probably better single thread performance than our A6-3670K CPUs. Even the A4-5300 has an IGP only slightly behind an HD4000 it seems when paired with a Core i7 3770k. You compare that to the G530 and G540 thought and the IGP is miles ahead. The CPU section is not bad if it is around Athlon II X2 and the IGP also can help out.

[CAT-THE-FIFTH]

Interesting rumour:

http://www.neowin.net/news/rumors-of...d-specs-appear

[Biscuit]

Interesting rumour:

http://www.neowin.net/news/rumors-of...d-specs-appear

Interesting indeed, wouldn't mind an amd powered one!

[CAT-THE-FIFTH]

Looks like IBM is preparing test chips made on a 14NM process:

http://www.tweaktown.com/pressreleas...hip/index.html

[CAT-THE-FIFTH]

It looks like no AM3+ SR until 2014:

http://www.cpu-world.com/news_2012/2..._and_2014.html

I hope they move to Excavator instead though. Maybe it is a good thing Intel might be still using 22NM in 2014 for its socketed CPUs,as AMD is indicating they will be using 28NM then.

I really hope,that the APUs are updated before then.

[watercooled]

I think that's kind of what we've been expecting, but AFAIK the desktop AM3+ parts generally appear before the server equivalents so we might see them early in the year, if not late 2013 if we're lucky.