*** AMD "Zen" thread (inc AM4/APU discussion) ***

[AthlonXP1800]

Looks to me that Zen has 24 lanes. Look closer at the diagram

http://www.kitguru.net/wp-content/uploads/2015/05/amd_promontory_features_capabilities.jpg

Look closer on note 1 and 3.

Zen FX cant use it for GPU with 4 lanes reserved for chipset and other 4 lanes reserved for NVMe so it limited to 16 lanes.

Skylake CPU has 20 lanes and also 15 lanes on skylake chipset that can be use for GPU. It will be interesting to see how it can handle quad crossfireX and SLi compared to Ivy Bridge-E and Haswell-E.

 

[Gazb]

Look closer on note 1 and 3.

Zen FX cant use it for GPU with 4 lanes reserved for chipset and other 4 lanes reserved for NVMe so it limited to 16 lanes.

Skylake CPU has 20 lanes and also 15 lanes on skylake chipset that can be use for GPU. It will be interesting to see how it can handle quad crossfireX and SLi compared to Ivy Bridge-E and Haswell-E.

I'm doubting this will be an issue. Haswell has managed fine up until now with limited pcie lanes, and only in extreme setups has Haswell-E proven to be faster.
It could be that adding many more pcie lanes to a CPU pushes up the diesize and cost considerably for minimal gains (outside of server environments).

I still quite surprised AMD's 2016 APU's won't have zen cores. If software finally catches up to offload FP tasks to a CPU's iGPU then combined with hUMA they'd really be in a strong position.

 

[cheesyboy]

Look closer on note 1 and 3.

Zen FX cant use it for GPU with 4 lanes reserved for chipset and other 4 lanes reserved for NVMe so it limited to 16 lanes..

Did you not see the post I was responding to?

 

[Orangey]

I would actually like a replacement bus instead of sticking with PCIe.

The data moving across a HBM interface will be 1TB/s perhaps next year. Meanwhile to go a few extra cm's to the CPU it drops down to 32GB/s best case.

The whole ATX thing needs a fresh set of eyes.

 

[CAT-THE-FIFTH]

Has anyone not considered that the chipset might offer more lanes? It could be that the on CPU functionality is shared with the APUs which don't need as much I/O.

 

[Sylver123]

Please ignore this post I have made here.

 

[Boomstick777]

Interesting write up on Seeking Alpha regarding Zen and AMD's renewed foucs on the high end space.

(Some of the comments are funny to)

----------------------------------------------------------------------------

AMD: This Intel Ankle-Biter Is On Its Way To Substantial Profits

Summary

AMD’s new focus could lead to huge price gains by the end of 2017.

AMD’s management team is concentrating on just a few profitable areas.

The AMD-Intel performance gap is about to shrink substantially.

AMD predicts $.50 earnings in 2017. I predict at least a $1.00.

Overview:

On May 6, Advanced Micro Devices (NASDAQ:AMD) had an investors' day conference where it outlined its future direction for products and sales. During this presentation, it made several points that I think will lead to a bright and profitable future. In recent weeks, AMD's stock price has dropped to about $2.30/share giving it a market cap of only $1.8 billion. That's barely 1% of Intel's (NASDAQ:INTC) $155 billion market cap.

In this article, I will review what was presented and why it will result in a whole new AMD in the next 2-3 years. I will also explain how this new concentrated focus should lead to substantial profits.

point is that AMD does not need to beat Intel with Zen, it only needs to narrow the performance gap and it will certainly do that.

So if we assume Intel x86 sales go from $47 billion to $55 billion in 2017 and AMD only picks up a very modest 4% share from Intel, we are looking at $2.2 billion in incremental sales. For a small cap $5 billion revenue business like AMD that's a very large sales increase and a hardly noticeable sales loss for a behemoth like Intel. As the much smaller player in the x86 duopoly, AMD stands to benefit enormously by an enhanced competitive product like Zen.

Risks:

AMD's decade-long problem has been execution and that is the biggest risk. It just does not deliver the products as described on time.

Because of the distorted market shares the little guy can greatly improve the outcome by picking up just a small piece of the big guy's market. Zen is a product that can do that.

http://seekingalpha.com/article/317...le-biter-is-on-its-way-to-substantial-profits

 

[Freddie1980]

I'll wait until the benchmarks come out before I buy any shares.

 

[ALXAndy]

I was reading the other day that Zen won't quite make the same per thread performance as the Intel, but, will be far closer than before.

So for me it's all about core count. If they can knock out a 8c 16t CPU with 80% per core of the 5960x for half the price (say around £450) then they're onto a bloody winner.

Thing is I just fitted my 8c 16t 2ghz Ivy Xeon and my FPS are amazing. So now I am wondering if I will even need Zen. With games going the way they are I doubt I will, but I will remain open minded.

 

[Boomstick777]

I'll wait until the benchmarks come out before I buy any shares.

I think the point is to buy shares now (Cheap) watch the value grow as Zen re takes some market share (VS Risk).

I was reading the other day that Zen won't quite make the same per thread performance as the Intel, but, will be far closer than before.

So for me it's all about core count. If they can knock out a 8c 16t CPU with 80% per core of the 5960x for half the price (say around £450) then they're onto a bloody winner.

Thing is I just fitted my 8c 16t 2ghz Ivy Xeon and my FPS are amazing. So now I am wondering if I will even need Zen. With games going the way they are I doubt I will, but I will remain open minded.

+1

If the performance core VS core doesn't match Skylake for example, but is decent, the key being that with AMD you get '8/16' VS '4/8' Intel for similar cost still think Zen would be a better option.

They can def find a market with higher performance CPU's. I will build with Zen for sure. Jim Keller won't let us down.

AMD not bringing them to market in time might let us down though Hopefully AMD get the execution right with Zen.

 

[FlanK3r]

I was reading the other day that Zen won't quite make the same per thread performance as the Intel, but, will be far closer than before.

So for me it's all about core count. If they can knock out a 8c 16t CPU with 80% per core of the 5960x for half the price (say around £450) then they're onto a bloody winner.

Thing is I just fitted my 8c 16t 2ghz Ivy Xeon and my FPS are amazing. So now I am wondering if I will even need Zen. With games going the way they are I doubt I will, but I will remain open minded.

no, this is differet segment story. How many people, typical John to buy i7-5960x+8x DDR4+PCIe SSD and example TitanX? Believe me, its less than 1% of customers.

Zen is segment against modern and profitable market = LGA1155/1150/1151. 95W TDP. Still most of users buy Pentium dual core or core i3. And still core i5/i7 in unlocked version are very, very good in real performance.
You can see some guys in forums with powerfull PCs, or many of them with core i5-4690K or 4790K etc, but you must to know, this is segment of enthusiast. Forum guys are enthusiast, gamers etc. And this market is more interesting than LGA2011. Intel can hold LGA2011 because company is very strong in servers and they have money for it.

 

[David Bisset]

Did you not see the post I was responding to?

You were responding to my post about Skylakes 20 lanes by telling me I was wrong because Zen has 24...

 

[ALXAndy]

no, this is differet segment story. How many people, typical John to buy i7-5960x+8x DDR4+PCIe SSD and example TitanX? Believe me, its less than 1% of customers.

Zen is segment against modern and profitable market = LGA1155/1150/1151. 95W TDP. Still most of users buy Pentium dual core or core i3. And still core i5/i7 in unlocked version are very, very good in real performance.
You can see some guys in forums with powerfull PCs, or many of them with core i5-4690K or 4790K etc, but you must to know, this is segment of enthusiast. Forum guys are enthusiast, gamers etc. And this market is more interesting than LGA2011. Intel can hold LGA2011 because company is very strong in servers and they have money for it.

Most X79 and X99 sales will be to professionals, not gamers. It's a segment AMD used to be very good in, because their Opterons were far cheaper than the Intel equivalent.

Lately though their Opterons have become really stale, and Intel now have X99 to tempt business users into buying their rebranded Xeons.

Right now X99 for a gamer is overkill. As you say, there's a very small market for it but it's still overkill and massively over priced.

If AMD can bring their flagship boards in at prices to rival Z97 and so on like you say *and* make the CPUs far cheaper then they could be onto a complete winner.

If I were AMD though? I would be strongly focused on sales to those who are simply bored of Intel's stagnation who would quite happily lay out the cash just to own something different, but something they won't get pointed at and laughed at for buying.

So as has been mentioned in a couple of articles I've read, AMD don't need to beat Intel at all. Well, not on performance at least. If they can come close and make that technology affordable and thus, available to all, then Zen will be pure win.

 

[Streetlight]

I think the point is to buy shares now (Cheap) watch the value grow as Zen re takes some market share (VS Risk).



+1

If the performance core VS core doesn't match Skylake for example, but is decent, the key being that with AMD you get '8/16' VS '4/8' Intel for similar cost still think Zen would be a better option.

They can def find a market with higher performance CPU's. I will build with Zen for sure. Jim Keller won't let us down.

AMD not bringing them to market in time might let us down though Hopefully AMD get the execution right with Zen.

I think its a great buy @2.32 very low risk imo

 

[cheesyboy]

Tech report's Zen podcast is worth checking out, if you're interested in a fairly deep analysis of what was revealed by AMD.

I'm about halfway through it and it's pretty interesting, if a little dry.

 

[=XDC=FluphyBunny]

I'll wait until the benchmarks come out before I buy any shares.

So if its successful the shares would already be up and then you would buy them. Wow you are a regular financial wizard aren't you

 

[elmuzzy66.]

So if its successful the shares would already be up and then you would buy them. Wow you are a regular financial wizard aren't you

Well he is, considering amd track record on delivery and performance in their last couple of architectures for the desktop.

 

[Orangey]

If Zen is a success and enables AMD to bounce back they can leverage it bigtime to snap up market share in various areas over the next few years. Buying shares in that inbetween period could be profitable.

 

[Freddie1980]

So if its successful the shares would already be up and then you would buy them. Wow you are a regular financial wizard aren't you

(gets his magic wand, waves it around a bit, says some funny sounding words)

Check your bank account you should find a couple of millions quid in there for you to spend how you wish.

It's worth bearing in mind just how much AMD have talked up products over the years and not followed through or undelivered, Bulldozer being the glaring example so you will forgive me if I'm cautious. However if you do think Zen is the real deal I've already said AMD shares are worth a small punt at their current price (say £500) but I'm going to invest thousands based on a few announcements. If Zen is successful yes I would buy as it would be a good indicator for long term growth rather then short gains.

 

[eddiew]

... Bulldozer being the glaring example ...

I spent a good 6 months waiting for Bulldozer - then the review sites got samples and I quickly bought an i5 the next day

I have no idea what AMD thought they were doing with all that hype, it seemed like they didn't realise people would actually run their own benchmarks... Sincerely hope Zen is not a lie. Would love to see some proper competition and someone giving Intel a reason to do better!