AMD's 'Zen' to arrive with 'Summit Ridge' CPUs in Q3 2016

[watercooled]

One theory I've heard is that Zen will be available for the higher-end market while Excavator on possibly cheaper 28nm will still be used for cheaper parts. (Will 28nm still be cost-competitive for APUs by then??)

It still doesn't seem to fit though, and a lot of the 'news' available is extrapolated from tiny snippets or paraphrased from undisclosed sources. Apparently Godavari is already the codename for the Mullins GPU.

Also, I'm not sure I understand the logic of still using 28nm for large APUs in late 2016 if 14nm and Zen are in production surely AMD would want to get as many people on their new core as possible and would be hoping to shift as many of the Construction core products from inventory as possible by then rather than making another one? And they already have some solid low-power cores for that market. Unless Excavator is far better than we're all expecting?

[TRUTHnTRINITY]

This is exciting news and if they bust their tails get it out Q1 2016 we're talking showdown of the 14nm CPUs and possibly a consumer friendly market.

[canopus72]

According to previous reports, the new zen cpu was supposed to be released on the new AM4 platform (and this was going to be called the FX line), but according to this review, the desktop zen cpu is going to be released on the updated APU platform (FM3). What the hell happened to the AM4 platform?. What platform will the next gen zen-opteron be released on?. There was also talk that the flagship X86 cpu would have 12 cpu cores (and 24 threads), so what happened to this one? (as a related hexus article I read 5 minutes ago is stating that the flagship zen cpu for X86 desktop will have 8 cores).

[DanceswithUnix]

I think what you were reading before is referred to as "guesswork".

If they can squeeze all the CPU lines onto one socket, then I don't care much what they call it, that would be good.

[shaithis]

Considering how far away the release is, I wouldn't be surprised if details (especially socket) change from what we are being told now.

[phoe]

I'm going to hate myself for this...

I wonder what the Serva/lan performance will be like - hopefully it will be the Liberator for AMD's market share.

At least I'm not a Slave to The System.

Enough of that, I'm off to listen to Travis in my Vila...

(I wish I could find a caption competition image I saw a while ago, there was an alien spider of some sort on one of the characters in the sick bay, saying "I was young, I needed the money")

[crossy]

According to previous reports, the new zen cpu was supposed to be released on the new AM4 platform (and this was going to be called the FX line), but according to this review, the desktop zen cpu is going to be released on the updated APU platform (FM3). What the hell happened to the AM4 platform?. What platform will the next gen zen-opteron be released on?. There was also talk that the flagship X86 cpu would have 12 cpu cores (and 24 threads), so what happened to this one? (as a related hexus article I read 5 minutes ago is stating that the flagship zen cpu for X86 desktop will have 8 cores).

Hmm, mixed feelings about this. Consolidation onto FM3 - if it's technically competent enough - makes a whole lot of sense to me. Then at least you can offer the tantalizing prospect of starting off with a "mainstream" APU and then moving onto a "proper" CPU when circumstances dictate .. an A-Z approach, APU-Zen.

On the other hand, if standardising on FM3 means that the Zen is going to be constrained - memory bandwidth, PCIe lanes, USB ports, etc - then it's a bad idea, in which case "AM4" is probably desirable. But, (conjecture system on 11), what would be ideal in that situation would not be "AM4" but rather "FM4" ... this being designed to fully-exploit Zen, but also supporting the lesser APU parts too.

Whatever gets eventually delivered (end-2016 - seriously AMD?!) needs to be a barnstormer - "APU+" will not cut it and AMD has been second-fiddle to Intel for too long.

I'm going to hate myself for this...
I wonder what the Serva/lan performance will be like - hopefully it will be the Liberator for AMD's market share.
At least I'm not a Slave to The System.
Enough of that, I'm off to listen to Travis in my Vila...
(I wish I could find a caption competition image I saw a while ago, there was an alien spider of some sort on one of the characters in the sick bay, saying "I was young, I needed the money")

Can we get some kind of award given for the most uses of Blake7isms in a single post?

[CAT-THE-FIFTH]

AMD needs to get this out sooner rather than later. Products are not the problem but execution is with AMD. 32NM Bulldozer,Piledriver,Llano and Kaveri were delayed as much as six to twelve months. If they had launched on time things would have turned out better for AMD.

[CAT-THE-FIFTH]

Repeat post.

[DanceswithUnix]

On the other hand, if standardising on FM3 means that the Zen is going to be constrained - memory bandwidth, PCIe lanes, USB ports, etc - then it's a bad idea, in which case "AM4" is probably desirable. But, (conjecture system on 11), what would be ideal in that situation would not be "AM4" but rather "FM4" ... this being designed to fully-exploit Zen, but also supporting the lesser APU parts too.

But remember, AM3+ was constrained. AMD said at Bulldozer launch time that AM3+ was a compromise because motherboard vendors were unwilling to create new motherboards for them. Die shots show what appears to be a PCIe interface on the die.

Personally I think using a small amount of die to give basic integrated graphics in a CPU die is fine. I don't want a plug in graphics card in a small server, having had a graphics card croak on me in my home rig a few years ago having basic integrated graphics (sideport ram! lol) made it easy to diagnose the fault and order a new card online. I just don't want 60% of the siliocon taken up with it.

So yeah, I'm good with FM3 as a way forward. That is what I would do.

Can we get some kind of award given for the most uses of Blake7isms in a single post?

Confirmed (plink!).

[ZaO]

Hmm, mixed feelings about this. Consolidation onto FM3 - if it's technically competent enough - makes a whole lot of sense to me. Then at least you can offer the tantalizing prospect of starting off with a "mainstream" APU and then moving onto a "proper" CPU when circumstances dictate .. an A-Z approach, APU-Zen.

On the other hand, if standardising on FM3 means that the Zen is going to be constrained - memory bandwidth, PCIe lanes, USB ports, etc - then it's a bad idea, in which case "AM4" is probably desirable. But, (conjecture system on 11), what would be ideal in that situation would not be "AM4" but rather "FM4" ... this being designed to fully-exploit Zen, but also supporting the lesser APU parts too.

Whatever gets eventually delivered (end-2016 - seriously AMD?!) needs to be a barnstormer - "APU+" will not cut it and AMD has been second-fiddle to Intel for too long.

I agree. Amd has always liked to keep our motherboard sockets going as long as they can, rather than making us upgrade them with every/every other new cpu. You might not get the same performance on an older socket, but it's still nice to be able to upgrade that at a later date and just get yourself a new cpu first. They care about their customers like that If FM3 can do the job properly, then I don't see the problem with it. AM4 would be cool too. It's whatever really..

AMD needs to get this out sooner rather than later. Products are not the problem but execution is with AMD. 32NM Bulldozer,Piledriver,Llano and Kaveri were delayed as much as six to twelve months. If they had launched on time things would have turned out better for AMD.

True. They get their cpu's out too late sometimes. But maybe it's because it's cheaper for them to produce them later? They don't have the same budget as Intel..

[shaithis]

True. They get their cpu's out too late sometimes. But maybe it's because it's cheaper for them to produce them later? They don't have the same budget as Intel..

.....and that's half their problem probably. With that type of setup your going to find it hard to get into massive OEM deals and the server market, where Intel make so much money.

Hell, even when they had better CPUs than Intel, they were finding it hard to penetrate those markets, since the pendulum swung they have been almost non-existent there.