AMD Confirms, Zen On Track For Q4 2016 Availability On High-End Desktops

[richaron]

As for the rest of the thread: I just don't understand the Intel squad intentions really, they jump into every AMD thread with negativity to the roof.

Yeah this site's "Intel super friends" are way worse than the usual OCD fans you see on the internet. The commitment to crapping on anything AMD related goes way past the usual manic keyboard warriors and fanbois grinding on their Intel logos.

That being said, it's pretty obvious their intention is to derail any substantial conversation about AMD. Question everything. Change topics and redirect focus. Strawman arguments and the like.

The goal is to ruin threads by taking the focus and momentum out of it. It only takes a few committed Intel super fans working full time to derail every AMD thread on this site.

As for keeping on topic: I came here looking for news and most of what I read is threadcrapping. I'm responding to that.


So you just add to the threadcrapping ?
Thread crapping and off-topic is not allowed here.
Markfw900

 

[mrmt]

That being said, it's pretty obvious their intention is to derail and substantial conversation about AMD. Question everything. Change topics and redirect focus. Strawman arguments and the like.

The goal is to ruin threads by taking the focus and red momentum out of it. It only takes a few committed Intel super fans working full time to derail every AMD thread on this site.

As for keeping on topic: I came here looking for news and most of what I read is threadcrapping. I'm responding to that.

Maybe the real reason is AMD itself. It's been more than three years since the last server/workstation processor and two since the last AMD mainstream processors, and they didn't exactly shatter benchmark records when they were released... So whatever is left for the fans is to speculate on whether future AMD products will be any good based on whatever info leaks from AMD and to defend on the forums that there is still value left in AMD current offers (If you overclock, run Mantle games and threaded applications you can get for $1.19 bucks less than Intel blablabla). Doesn't sound much exciting to me.

 

[ShintaiDK]

Definitely, i strongly believe they will be very competitive against Intels mainstream skus (Kabylake socket 1151) and perhaps very price and perf/watt competitive against entry Broadwell-E HEDT SKUs.

I still hope they will offer higher core count at the same or better price as Core i7 Socket 1151 models. They could do that since Construction Core ZEN will be iGPU-less and so they could actually increase core count without significantly compromise on the die size, unless they use huge L2 and L3 caches that will increase die size more than what CPU cores will.

The biggest challenge is the interconnect. And it takes up space as well.

Also we haven't seen any interlink information from AMD yet or anything else relating to something that can be used for 2P and up. Dual channel wont work there. Single socket servers isn't going to generate them any useful money and they cant compete with the SoC lines.

 

[DrMrLordX]

Let him point that out. I think, I commented on that already. You talked about a specific software, while desprado let it sound like a general claim.

Heh, fair enough. I will point out, though, that the way people react to commentary about a comparison with a specific piece of software is not always rational. The AMD fan might have read my analysis and gone on to conclude that Zen would be faster clock-per-clock than Skylake in everything. The pro-Intel skeptic might have dismissed everything I said on the basis that they thought what I said matched the conclusion of the AMD fan exactly. When in reality, all I was talking about was legacy SSE2 code.

Had I gone on to use Cinebench R11.5, I'm sure Zen would have wound up looking less impressive.

- Or new regs will be old regs.

Heh well, good luck proving that! That A9X thread had at least one recent reg that was not behaving like any of the usual AT suspects so I dunno . . .

Maybe the real reason is AMD itself. It's been more than three years since the last server/workstation processor and two since the last AMD mainstream processors, and they didn't exactly shatter benchmark records when they were released... So whatever is left for the fans is to speculate on whether future AMD products will be any good based on whatever info leaks from AMD and to defend on the forums that there is still value left in AMD current offers (If you overclock, run Mantle games and threaded applications you can get for $1.19 bucks less than Intel blablabla). Doesn't sound much exciting to me.

If it isn't that exciting, it might be better to go focus on other topics. Let the AMD fanboys salivate over the next line of AMD products. Shooting them down pre-emptively doesn't really help anyone.

The Zen hype might be getting to be a bit much, but look at it this way: if AMD can't deliver with Zen, the party's probably gonna be over in 2-3 years, at least for their CPU division.

 

[maddie]

Maybe the real reason is AMD itself. It's been more than three years since the last server/workstation processor and two since the last AMD mainstream processors, and they didn't exactly shatter benchmark records when they were released... So whatever is left for the fans is to speculate on whether future AMD products will be any good based on whatever info leaks from AMD and to defend on the forums that there is still value left in AMD current offers (If you overclock, run Mantle games and threaded applications blablabla). Doesn't sound much exciting to me.

THEN WHY ARE YOU HERE?

Is this an subconscious admission that you're here to create chaos?

Several days ago you made the claim of the superior electrical properties of TSMC 16nm finfet process vs Samsung 14nm.
I asked you and another party for some data as I was interested in this for an estimate I wanted to roughly calculate.

http://forums.anandtech.com/showpost.php?p=37942013&postcount=432

Quote:
Originally Posted by ShintaiDK
I see you changed focus again. You forgot a node is not only about area, but also electrics. Hence why TSMCs 16FF+ is vastly superior to 14LPP. And why A10 will be TSMC 16FF+(FFC perhaps) only.

And I can see the Pentium-M part broke your dreams along with the "up to" moniker. So your next step was personal.

In terms of IPC, x86 stagnated with the PPro. Yet you expect miracles from a company run into the ground by incompetence.

I will be enjoying my "slow" quadcore for a long, long time

Quote:
Originally Posted by mrmt
The last official information from AMD is that they are obliged to manufacture all their MPUs with Globalfoundries. You are not shy to baseless speculation about AMD being free of the WSA, but you fight people who speculate (with more evidence than you) that the TSMC process is better than Samsung's. No wonder that even new posters are finding you terribly biased.

Quote:
Myself----
I am interested in these facts, can you post links to the info or post it here directly.

Nothing, no response, not even an acknowledgement of the request. You pretended I never asked, and continued with your misleading statements. You sir, are fabricating "data" to achieve your ulterior goal, and in the process destroying the experience for many members.



As a thought exercise. How many of us would buy and at what price, an 8 core Zen with hyperthreading giving a 60% greater IPC over a Vishera 8370E.

 

[DrMrLordX]

As a thought exercise. How many of us would buy and at what price, an 8 core Zen with hyperthreading giving a 60% greater IPC over a Vishera 8370E.

That would depend on attainable clockspeeds. If Summit Ridge is going to be 3.33 GHz base with 4.0 GHz turbo as has been speculated, then somewhere in the $500-$600 would be a good price range for AMD, especially if the OC headroom is not there. Though they will want $999, I'm sure. If the production run in Q4 is small enough, expect the $999 price.

 

[Dresdenboy]

Fixed it for you.
Curiously the drama queens don't complain when people criticize Intel for their less than stellar mobile chips, process delays, iGPUs, etc.

Dear A.E., have you checked IDC's last postings before fixing this? The second last posting was about defending AMD's choice of creating Brazos against Shintai. In my eyes his views were neutral/realistic, which means, depending on the quality of a posted opinion by others, he went with either side. AMD's management and BOD has been blamed by him, which stands in contrast to his view, that AMD's past efforts helped shaping the 64b x86 world and also the mobile world somewhat.

Back to the emotional fan groups here: It would really be nicer to concentrate more on the interesting technical stuff. Each biased posting (often repeating the same old negative points about some past, or just showing strong scepticism vs. "the other side") just triggers emotions on the other side (insult, perceived unfairness, etc.), keeping the fight going on. I'm sure, strongly staying with known facts and discussing speculations as what they are, would make the threads too boring for some. That pattern is known from sports, politics, extremism, etc. If it is for fun, OK, but for many it's a serious thing. So why would educated people on technical forums want to stay at such an immature level?

@maddie:
Not answering is an argumentum ad ignorantiam. But stay calm, I'm preparing some collected data of the foundry FinFET processes. You might also check some slides about 14nm, 16nm FinFET processes and results (e.g. ARM, Broadcom) from different sources by Ohara-san:
http://news.mynavi.jp/special/2016/trend/
http://news.mynavi.jp/articles/2015/12/28/processing/


BTW, one thought I had about the "Zen Hypetrain": Our whole financial system is a hypetrain.

 

[maddie]

That would depend on attainable clockspeeds. If Summit Ridge is going to be 3.33 GHz base with 4.0 GHz turbo as has been speculated, then somewhere in the $500-$600 would be a good price range for AMD, especially if the OC headroom is not there. Though they will want $999, I'm sure. If the production run in Q4 is small enough, expect the $999 price.

Taking into account AMD's reputation at present, don't you think it would be a better decision to gradually move to an Intel pricing level? The worst thing is to release at $999 and within months, drop to $500-600. This destroys brand image.

Surely with the apparent small size of Zen, they can sell at $500-600 and substantially improve their margins.

 

[fourdegrees11]

Zen - "Black Edition" Chips:
$75 - 2c/4t
$150 - 4c/8t
$250 - 6c/12t
$350 - 8c/16t

If AMD cant compete at these price points they dont have an enthusiast CPU business. There's just no way anyone should expect AMD to put equal thread counts against Intel when they cant possibly equal the IPC and clock speed.

 

[USER8000]

Zen - "Black Edition" Chips:
$75 - 2c/4t
$150 - 4c/8t
$250 - 6c/12t
$350 - 8c/16t

If AMD cant compete at these price points they dont have an enthusiast CPU business. There's just no way anyone should expect AMD to put equal thread counts against Intel when they cant possibly equal the IPC and clock speed.

No,I don't agree - AMD like Intel is not a charity and unless the Zen chips are having very small dies,to be selling a largish CPU at those prices would make them hardly any money,and they might as well not bother.

Anyway,I cannot say for the US,but in the UK the Core i5 6600K is like £200 with 20% VAT and the Core i7 6700K is around £300+ and the Core i7 6700 is £270+ too.

The cheapest Core i5 chips are around £145 to £155. The Core i3 6300 is nearly £120!!

On Newegg the Core i5 6600K is $270 and a Core i5 6400 is like $190 ,so no I don't think AMD needs to sell a 4C/8T at $150,which is Core i3 level pricing. $150 is the price of a Core i3 6300 which is a dual core with HT.

So expecting AMD to price a 4C/8T chip for the same as a 2C/4T chip from Intel seems a bit wishful.

4C/8T pricing at a similar level to Intel 4C with no HT pricing would be the minimum I expect.

 

[frozentundra123456]

Well, anyone who feels a post is thread crapping is certainly free to report it and let the mods decide. Otherwise, i was not aware that individual posters were allowed to determine what is posted in a thread.

 

[maddie]

The initial Zen products won't have an IGP and therefore should be competing in the HEDT segment. I hope the many pricing mistakes from the construction core era is not repeated.

 

[USER8000]

Even if we take the 40% IPC claims over Excavator (which would 50% to 60% over Piledriver) as being even closer to 30% to 40% over Piledriver,its still about Ivy Bridge level,which is respectable.

People are forgetting that the cores in the Bulldozer line have probably only got to around K10 level with Excavator,and they were betting the boat on their CMT-like design and using shared resources to try and cut down the size of them,and then having high clockspeeds to compensate. Plus their cache designs were still worse than what Intel has. Even if you look at their "Cat" cores,they managed to get similar IPC to their later Bulldozer derived cores and the former were a fraction of the size.

They appear to be going for a far more conventional design this time,so starting from a relatively low starting point,I don't think a 30% to 40% IPC increase is as big a bridge to cross as people think.

Intel has started with relatively high IPC designs,so OFC the improvements per generation are not as big,but they started at a much higher level!!

I expect the situation to be like with the Phenom II X6 and Lynnfield Core i5 chips,except AMD probably will have paid a bit closer attention to power consumption this time.

So maybe for the price of a Core i5 6600K you get a Core i7 3770K/4770K equivalent.

Also,to clarify this only my opinion and nothing more.

 

[TheELF]

So expecting AMD to price a 4C/8T chip for the same as a 2C/4T chip from Intel seems a bit wishful.

They are selling the fx 8cores at i3 level prices right now,they will sell ZEN at what ever price level it's performance level (for the average user) will be in relation to intel,if their 4c/8t will perform like a then i3 then that's going to be their price,maybe not at introduction, as was already pointed out, but it will end up there.

 

[mrmt]

THEN WHY ARE YOU HERE?

Because from a business POV they are very interesting. I can't recall of many companies displaying that levels of incompetence for such a long timespan and still surviving, as if the management team is too incompetent to break the company for good, despite trying so hard.

 

[fourdegrees11]

No,I don't agree - AMD like Intel is not a charity and unless the Zen chips are having very small dies,to be selling a largish CPU at those prices would make them hardly any money,and they might as well not bother.

Anyway,I cannot say for the US,but in the UK the Core i5 6600K is like £200 with 20% VAT and the Core i7 6700K is around £300+ and the Core i7 6700 is £270+ too.

The cheapest Core i5 chips are around £145 to £155. The Core i3 6300 is nearly £120!!

On Newegg the Core i5 6600K is $270 and a Core i5 6400 is like $190 ,so no I don't think AMD needs to sell a 4C/8T at $150,which is Core i3 level pricing. $150 is the price of a Core i3 6300 which is a dual core with HT.

So expecting AMD to price a 4C/8T chip for the same as a 2C/4T chip from Intel seems a bit wishful.

4C/8T pricing at a similar level to Intel 4C with no HT pricing would be the minimum I expect.

- Actual MSRP of 6700K is $350, 6600K $243. Retailers are gouging short supply
- No IGP on Zen helps die size
- Current AMD situation where a large 8 core chip cant even sale vs the i3.

Yes I see AMD trying to actually gain market share by attacking Intel's ancient CPU market structure. That was their goal with FX you know, it just failed miserably. Think, what would make someone buy Zen if it's slower then Intel? "Yeah this 4c/8t AMD chip equals an i7 from 4-5 years ago, sure I'll pay the same as this much faster current i5" No, if there isnt an actual benefit for the consumer it will fail. Single thread performance that is decent (unlike now) paired with a distinct multi-threaded advantage will actually make the chips worth purchasing over Intel for a lot of people.

 

[USER8000]

- Actual MSRP of 6700K is $350, 6600K $243. Retailers are gouging short supply
- No IGP on Zen helps die size
- Current AMD situation where a large 8 core chip cant even sale vs the i3.

Yes I see AMD trying to actually gain market share by attacking Intel's ancient CPU market structure. That was their goal with FX you know, it just failed miserably. Think, what would make someone buy Zen if it's slower then Intel? "Yeah this 4c/8t AMD chip equals an i7 from 4-5 years ago, sure I'll pay the same as this much faster current i5" No, if there isnt an actual benefit for the consumer it will fail. Single thread performance that is decent (unlike now) paired with a distinct multi-threaded advantage will actually make the chips worth purchasing over Intel for a lot of people.

Intel like all companies is looking at the desktop as an easy cash cow - they will gradually increase prices as they care more about things like mobile and servers which are far more important to their future than serving tight arse "enthusiasts" who want to spend $100 or £100 on a chip and not upgrade for 10 years. You talk about "gouging" - that only happens if supply is low,which means Intel is not making enough of them and its in their own interest to sell things for as high a price as they can,and the same will go for retailers.

Expecting 4C/8T chips to be Core i3 pricing is just ridiculous,and I can see what is happening here - hype up the fact it needs to be 10X time better than what Intel has and at half the cost. Then even if does deliver on what AMD claims(closer to Haswell/Skylake) I can see all the claims of "failure!failure!failure!"

Then gloat when AMD has made another crappy quarter with low margins.

They won't make any money serving tight arses. They might as well sell less chips at a higher average ASP,and the tight arses can go buy a Pentium then.

Anyway at that sort of cost structure,they might as well withdraw from selling chips to "enthusiasts" and just focus on OEM contracts for APUs and selling whatever they can in the server area.

AMD being the budget option for CPU users,has never really helped them for a long time. Even with the Athlon,Athlon XP and Athlon 64 they still sold chips at decent prices,and they were not always better than Intel for every SKU.

 

[AtenRa]

Personally i wouldnt sell Construction Core ZEN dies with less than 6 Cores.

So i would sell the 6 Cores at Core i5 prices, and 8 Cores for Core i7 prices.

For 2C/4T and 4C/8T i would use APUs only.

 

[fourdegrees11]

Expecting 4C/8T chips to be Core i3 pricing is just ridiculous,and I can see what is happening here - hype up the fact it needs to be 10X time better than what Intel has and at half the cost. Then even if does deliver on what AMD claims(closer to Haswell/Skylake) I can see all the claims of "failure!failure!failure!".

They won't make any money serving tight arses. They might as well sell less chips at a higher cost base,and the tight arses can go buy a Pentium then.

Anyway at that sort of cost structure,they might as well withdraw from selling chips to "enthusiasts" and just focus on OEM contracts for APUs and selling whatever they can in the server area.

So are you just going to ignore what their current situation looks like? And that it is virtually impossible for AMD to deliver a clock for clock competitor? Making money means selling chips for a profit, not just above cost. If an 8c chip can sell for $130 right now how is selling a lot more chips and (hopefully) gaining a chunk of market share with a $350 8c chip not going to work? That's exactly what AMD needs to do.

How much is the 4c/8t AMD chip going to cost in your world, $250-$300? Why would anyone buy that in 2017 when it is likely going to be around the performance level of a 3770K - 4770K?

 

[USER8000]

Personally i wouldnt sell Construction Core ZEN dies with less than 6 Cores.

So i would sell the 6 Cores at Core i5 prices, and 8 Cores for Core i7 prices.

For 2C/4T and 4C/8T i would use APUs only.

The problem is those non-IGP chips are not going to be small dies - look at the socket 2011 chips for guidance,especially since APU chips probably will have far less cache onboard than server orientated chips.

Add the fact all the R and D costs need to be recouped,the extra costs of production on a new node,etc,its not viable to be selling the chips so lowly priced.

People forget that chips like the Phenom II X4 955BE and 965BE were a bit cheaper than the Intel equivalents and the same goes with chips like the Phenom II X6 which tracked Lynnfield Core i5 prices.

Only towards the very end did we see some bargain basement chips.

So expecting 3 times the number of usable threads at the same price as a Core i5 seems a bit cheap!!

Even the FX8350 and FX6300 had only 50% to double the thread counts,and they are ancient chips.

If they need to price them that cheap,then its better AMD cans Zen without an IGP,and just concentrate on trying to make the GPUs more competitive against Nvidia this year,and trying to at least have some reasonable IGP performance next year.

So are you just going to ignore what their current situation looks like? And that it is virtually impossible for AMD to deliver a clock for clock competitor? Making money means selling chips for a profit, not just above cost. If an 8c chip can sell for $130 right now how is selling a lot more chips and (hopefully) gaining a chunk of market share with a $350 8c chip not going to work? That's exactly what AMD needs to do.

How much is the 4c/8t AMD chip going to cost in your world, $250-$300? Why would anyone buy that in 2017 when it is likely going to be around the performance level of a 3770K - 4770K?

Dude,they won't making any money on the fantasy pricing structure you are proposing - they might as well can Zen now if they need to sell massive chips for the price of a Core i3 or 8 core monsters with massive caches for the price of a tiny 4T chip.

You are forgetting the first releases will be server chips like the socket 2011 ones. Huge chips with no IGP and with massive amounts of the die taken up by caches,etc especially if they are 8C/16T dies.

They might as well,give up now since from a financials point of view it will be worse than Bulldozer. They are a far smaller company now than years ago and if they cannot sell Zen at decent margins it will be the end.

They have been selling at lowish margins for years.

They might as well totally exit the "enthusiast" desktop area. Even if they sell the chips "cheaply" in commercial systems,it will be probably at better margins than selling a huge 8C/16T die for £120.

I would rather AMD at least survives making GPUs,than having their CPU business destroy the whole company.

 

[fourdegrees11]

Just because Intel has milked profits out of SMT doesnt mean AMD wont do their normal thing and offer up all of their features/performance across the board (it's how they do things). Thinking they're going to mimic Intel's structure with chips that will likely perform like 4-5 year old Intel chips isnt very realistic.

 

[Dresdenboy]

Zen - "Black Edition" Chips:
$75 - 2c/4t
$150 - 4c/8t
$250 - 6c/12t
$350 - 8c/16t

If AMD cant compete at these price points they dont have an enthusiast CPU business. There's just no way anyone should expect AMD to put equal thread counts against Intel when they cant possibly equal the IPC and clock speed.

As I wrote in another forum:
The basic Zen compute unit has at least 4 physical cores. So Zen based CPU models would scale in steps of this size.

I found this line in a family 17h related patch:

+ core_complex_id = (c->apicid & ((1 << bits) - 1)) [b]>> 3[/b];

The highlighted part suggests a "core complex" size of 8. As it works directly on an APIC ID masked by a core bit mask with no sibling bit correction, this might even mean 8 physical cores per "core complex". The minimum number (as seen on leaks) would be 4 physical cores with "SMT-2" per core complex.

AMD might disable some cores or SMT capabilities based on that given granularity, similar to cat core compute units (w/o SMT of course).

 

[jhu]

Because from a business POV they are very interesting. I can't recall of many companies displaying that levels of incompetence for such a long timespan and still surviving, as if the management team is too incompetent to break the company for good, despite trying so hard.

Microsoft under Ballmer's helm. If it wasn't for most people's lockin to Windows, they'd already be dead.

 

[AtenRa]

OK, my prediction for an 8 Core ZEN die is not more than 200mm2, i believe more likely closer to 160mm2 is possible.

Also, a 4C 8T APU die will not be more than 200mm2, could be closer to 160mm2 as well. That is if they will not go ballistic with the iGPU size and keep it at the same size as the 28nm Carrizo.

So now they could sell the 6 Core ZEN at Core i5 and 8 Core at Core i7 prices.
Also, the APU could go all the way up to $300 with some HBM2 SKUs, but also lower at sub $150 to compete against Core i3s.

But in order to get those prices they need to have the necessary volume.

 

[fourdegrees11]

Dude,they won't making any money on the fantasy pricing structure you are proposing - they might as well can Zen now if they need to sell massive chips for the price of a Core i3 or 8 core monsters with massive caches for the price of a tiny 4T chip.

You are forgetting the first releases will be server chips like the socket 2011 ones. Huge chips with no IGP and with massive amounts of the die taken up by caches,etc especially if they are 8C/16T dies.

They might as well,give up now since from a financials point of view it will be worse than Bulldozer. They are a far smaller company now than years ago and if they cannot sell Zen at decent margins it will be the end.

They have been selling at lowish margins for years.

They might as well totally exit the "enthusiast" desktop area. Even if they sell the chips "cheaply" in commercial systems,it will be probably at better margins than selling a huge 8C/16T die for £120.

I would rather AMD at least survives making GPUs,than having their CPU business destroy the whole company.

So a $150 quad core is a massive chip for an unrealistic price (Intel sells quads for a few bucks more FYI). Yet AMD cant even sell a $130 8 core right now.

You keep repeating the same thing, ok you disagree. I'm still waiting to hear your scenario on how they can sell chips that perform like 4-5 year old Intel chips, cause this is mine.