AMD - Piledriver chitchat

[ik9000]

Is it just me or does pushing 'free' SoCs to OEMs, and the 'subsidies' mentioned, sound legally iffy? Isn't that the sort of thing that they were successfully sued by AMD for?

it may well be, but don't know enough to say for certain...

[DanceswithUnix]

Is it just me or does pushing 'free' SoCs to OEMs, and the 'subsidies' mentioned, sound legally iffy? Isn't that the sort of thing that they were successfully sued by AMD for?

ISTR Intel settled out of court, paying AMD a sizable lump of money but not close to the long term damage they inflicted. In other words, strategically Intel won.

They did get some marketing subsidy flak in later anti trust cases which Intel could be considered as "losing", but again the damage was done by then so meh.

Now that the old company leaders are retiring, would be nice if the new guard cleaned up their act. If IBM managed it, Intel can.

Edit post: I think what they got pulled for last time was the clause that factories only got the marketing subsidy if they only sold Intel based kit. Bundling things is fine and happens all the time (cpu+chipset for example).

[watercooled]

I was thinking more because of the free chips - it's something smaller companies cannot do. I could be wrong though, it just seems dubious to me.

[DanceswithUnix]

I was thinking more because of the free chips - it's something smaller companies cannot do. I could be wrong though, it just seems dubious to me.

They could fall foul of anti dumping regulations perhaps, if they can be shown to be trying to damage competitors by simply selling stuff below cost price.

They always muddy things though. Sell you a CPU, then pay your marketing costs for the device it goes into. That isn't a free CPU on the books so they can claim all this revenue, but in the end the costs balance out so it might as well be.

Buy one get one free is a fairly simple marketing tactic though, so there are ways to give stuff away that people will accept.

I wish Intel were as good at instruction set architectures as they are at playing these sorts of games

[DanceswithUnix]

If they were in the net-top style PC box i think i would consider one just to add to my collection of pointless PCs that i will one day find a use for... but thats a bit bigger than i would like.

I wish they would release micro ITX board based ones for us to use in our own cases

Perhaps Zotac heard you

http://www.fudzilla.com/home/item/33...-kabini-nettop

[CAT-THE-FIFTH]

http://techreport.com/news/25702/zot...owered-mini-pc

I wonder if we'll start to see Kabini products on the electronic shelf any time soon?

Perhaps Zotac heard you

http://www.fudzilla.com/home/item/33...-kabini-nettop

I know that sounded familar!!

[watercooled]

If the articles are accurate, then it speaks volumes as to how competitive the silicon alone is if Intel have having to literally give them to OEMs to get them into products.

I think it was Semiwiki where I recently read, Atom gaining market in the mobile space is despite x86, not because of it. Regardless of what the fanboys think. Of course there are still the delusional few who swear anything non-Atom in the mobile space is so slow it's unusable, which by itself is utter rubbish of course, but also ignoring how massively underpowered the GPU was in the 32nm Atom SoCs vs competition.

That said, competition is good provided the companies play fair, and as DanceswithUnix said we can hope the new management sort things out in that regard, and concentrate on making good products rather than bullying competition out of the market and stagnating progress. They're up against some seriously well-heeled players this time though e.g. Qualcomm, Samsung, Apple, Allwinner (in the Chinese market at least), Broadcom, plus Mediatek, TI, Rockchip, Nvidia, AMD, and I've probably missed a few others.

However, I wonder how having two (or more) ISA in one market would work out? It's something we haven't really seen on the desktop in recent years - as I said it could be good competition-wise for the ISAs but a potentially headache for devs.

[CAT-THE-FIFTH]

The only reason I want a Windows netbook or tablet is so I can play some games on them,and run MS Office.

The thing is though,the IGP in BT is much worse than even the cut down Jaguar equivalents and even with Office,there are enough free alternatives now available for the odd bit of word processing. I wish Intel had reduced CPU performance and jacked up IGP performance instead TBH.

The thing is X86 is not essential for the market anymore.

[watercooled]

On the subject of office, I noticed some office apps were included with the KitKat update on my Nexus 4...

[DanceswithUnix]

They're up against some seriously well-heeled players this time though e.g. Qualcomm, Samsung, Apple, Allwinner (in the Chinese market at least), Broadcom, plus Mediatek, TI, Rockchip, Nvidia, AMD, and I've probably missed a few others.

However, I wonder how having two (or more) ISA in one market would work out? It's something we haven't really seen on the desktop in recent years - as I said it could be good competition-wise for the ISAs but a potentially headache for devs.

Well heeled is nice, but the new dimension here is that the likes of Samsung really have no need if Intel at any level. I just don't see how Intel can bully them.

Similarly with Apple, if Intel try anything on then we will get ARM based Apple laptops pretty quickly and that will be the end of that.

As for devs, it used to be a bit tricky when you need to do stuff in assembler, but frankly these days it just isn't a problem. Every job I have had bar one has involved writing code that has to run on multiple architectures, including having to run on 8 bit and 32 bit with just a recompile, big and little endian. Marketing types might baulk, but there really isn't a problem there. Perhaps games devs will not want the extra tuning, but they seem to cope with the range of Android devices quite well (ARM, Intel and a few MIPS devices).

[CountBartok]

I guess the point is they claim to have seen no reduction in performance when "disabling" certain cores of the i5. Personally, I suspect either recording error or methodology error, because claiming BF4 runs as well with a 2 thread CPU as with a 4 thread is entirely contrary to everything else that's been written and published about BF4 (AFAIK, anyway).

Ok that is along the lines of my thinking.
Just wanted to ask as the posters in this thread have a better understanding of
cpu's than me.

Thanks to every one for there replies.

[watercooled]

I don't know why I've only just spotted this, but supposedly there are a ton of different Haswell die variants.
http://wccftech.com/haswell-die-conf...idge-revealed/

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

Something no-one seems to have noticed in those images, but is touched upon in the much earlier cpu-world article, is the size of the memory controller in the various die images - the left 4 (3 of which we already knew about) have a larger IMC which probably means dual channel. The right 4 have a much shorter IMC, maybe they're the single-channel parts? Also, I wonder if these are all physically separately fabbed dies, or just virtual dies from binning?

Either way, that seems an awful lot of dies. 2+2 could potentially cover desktop i3, and 2+1 perhaps the new Pentium, but they both have the shorter IMC so does that rule them out? Four separate specially-fabbed single-channel dies seems unlikely IMO.

Even 4+1 doesn't seem to fit anywhere - the desktop quads use the '2' IGP, and the Xeons have 8MB cache vs the 6MB of that die.

Edit: Hang on what does Ivy Bridge have to do with it?

Edit2: Yeah I'm really getting carried away with this now, but looking at these images (Haswell is rightmost), and based on aspect ratio (pixel counting) it looks like it could be a '2+2' die assuming the images are all accurate - the 2+3 die should be taller, roughly 1:2.5 ratio, but here it's more like 1:2. The top die is the chipset.

Here we can see the i7-4650U, a GT3 part, and the die is about 1:2.5 as expected.

[CAT-THE-FIFTH]

Some details about the clockspeeds of the desktop Kaveri CPUs:

http://prohardver.hu/teszt/mit_tudha...eri_igp-t.html

It seems the A10 7850K has a maximum clockspeed of 4GHZ. For the 20% projected improvement in performance over Richland,that would mean a decent boost in IPC(if you compare it to the A10 6800K),since the A10 7850K has 3.7GHZ to 4GHZ clockspeeds and the A10 6800K,4.1GHZ to 4.4GHZ clockspeeds. Even if we look at the earlier AMD claims of 10% to 15% performance improvement per generation, IPC probably has improved around 20% to 30% it seems.

[CAT-THE-FIFTH]

I saw this on the HardOCP forums:

http://i.imgur.com/Rg9fKas.png



It could be a photoshop as it does not look like a proper AMD slide.

I have some doubts over its validity,as I doubt if AMD would bother with AM3+ that long,especially as their APUs will be on 28NM and probably 20NM.

Edit!!

Supposedly from this Hungarian website:

http://prohardver.hu/index.html

I don't see it mentioned.

[KinTharEl]

I personally just want to know when they plan on releasing the enthusiast (FX) Steamroller architecture so I can stop waiting and get a chip already.

[ik9000]

I personally just want to know when they plan on releasing the enthusiast (FX) Steamroller architecture so I can stop waiting and get a chip already.

+1 and I share your hesitance Cat, no mention of DDR4 even in 2015 seems to contradict previous AMD announcements IIRC