Zen delayed to 2017 (DigiTimes)

[slackerinabox]

Mmm... I am nearly sure that this is FinFet's doing.

Intel can't deliver ungodly chips like SB or IV or even Haswell. Their Broadwell E can't OC more than 4.5 Ghz and Skylake can't go past 4.9 Ghz while SB can go 5.2 Ghz on air and IB and even HW can reach 5.0 Ghz...

nVIDIA can't scale down (so no sub GP 104) for a time and their graphic chips despite are decen't OCers can't past 2.5 Ghz

AMD chips are being delayed and Polaris 10 only manages 390X levels at best and seems that can't OC more.

As for VIA, they might come at 2017, so is time to wait there.
And as for ARM... they are still good, but definately needs more improvements.

In few words FinFet ended to be a sidegrade instead a massive upgrade. A lackluster improvement that needs more fixing than other things.
So definately according to this thread:
http://forums.anandtech.com/showthread.php?t=2474487

The true upgrade will be EUV.

But I think this misses the larger point - the majority of consumer are fine with 5 year old cpu technology on the x86 end, while arm still seems to be doubling potential every two years. Hell, a core 2 duo imac with an ssd works just fine after all these years, and arm already surpasses that. Look at cyclone and the snapdragon 820- monster chips but imagine 2 more moore's law cycles? I think innovation isn't happening because the demand isn't there, and it's really as simple as pure economics. Though there will always be an enthusiast market, it's just doesn't have the demand to require the supply to bring down prices and increase innovation thru competition. I'm thinking strap a huge battery to an overclocked cyclone and put that in a touchscreen tablet/laptop hybrid- that'll suit most people just fine, I think. Of course video editors and bankers plus content creators will need desktops, but it's a small slice and not consumer facing.

http://wccftech.com/apple-a9xipad-pro-benchmarks/

I don't think just AMD is doomed (unless they adapt), but Intel will lose interest in the consumer market too and court the industrial and server market.

 

[NostaSeronx]

Just in time confirmed...

- All AMD64 architectures have been scrapped. Scrapped being non-literal, they will launch but not have any successors.
- New ISA -> AMD64/P (Think Y86-64, Power ISA v3.0, and AMD64 fused//Semi-RISC/Semi-CISC ISA w/ high register count of Power, while all compatible with AMD64)
- Redesigned+From Scratch CMT Architecture -> Tunnelborer (IoT to HPC - Ultra Wide AVFS) // SEC28FDS prototyping (two version Tunnelborer-A = Dual-core, Faster TTM; Tunnelborer-B = Quad-core, Slower TTM)
- AMD to be Wholly Owned Subsidiary of IBM by 2022. (Read the 10-Qs from 2003-2010, oops)

 

[myocardia]

...

 

[coercitiv]

In fact can you even use fast ram in a non-overclocking mother board?

The answer is both Yes and No: official specs say faster than 2133 memory can only be used with Z170 on the mainstream platform, but some manufacturers allow overclocking through manual settings. See this post and maybe even take a look at the entire thread, as it discusses matters related to this (off)topic.

 

[mohit9206]

That's why once 6700K dropped in price over time, I didn't think it was worth waiting for Zen/KBL anymore. KBL seems completely meh, maybe 200mhz higher clocks, but I doubt it will overclock much better. Zen will probably be a very solid workstation/productivity CPU but considering today I'd probably pick 6700K over 6800K, it's highly doubtful Zen's single threaded performance will meet my expectations. MicroCenter now has 6700K for $289.99, which is a stellar deal. Looks like I am good to go at least until Icelake, or even until 2020-2021 (next true Intel gen after Icelake).

All of this of course doesn't paint a positive picture long-term. If there is a lot of unsold Haswell/SKL inventory, the only way for Intel to keep increasing revenue is to keep raising prices.

Icelake sounds really cool (lol), should be a nice upgrade from Sandy Bridge to Icelake.

 

[DrMrLordX]

The difference between Intel and AMD inventory isn't super relevant. Retailers have plenty of stock of processors, full stop, and there won't be much demand for them if they were to launch in October. Not to mention the fact that you all seem to be ignoring the suggestion that AMD and Intel are both facing supplier shortages.

So you really think that there's no demand for Zen/Summit Ridge? After about four years of Piledriver?

wtf do OEMs have in stock that has anything to do with whether or not AMD customers (or would-be AMD customers) will be looking at Summit Ridge? The entire idea that it's better for AMD - or OEMs, or anyone else - to bumble along with stale inventory from years ago for another three months is completely insane. They would be better off giving the crap away to the homeless, recycling what they could and dumping the rest in a landfill, or anything else.

Or are you suggesting that OEMS are looking to clear out inventory of Intel products before they're willing to sell an AMD product?

Just in time confirmed...

- All AMD64 architectures have been scrapped. Scrapped being non-literal, they will launch but not have any successors.
- New ISA -> AMD64/P (Think Y86-64, Power ISA v3.0, and AMD64 fused//Semi-RISC/Semi-CISC ISA w/ high register count of Power, while all compatible with AMD64)

Whaaaaaat

- Redesigned+From Scratch CMT Architecture -> Tunnelborer (IoT to HPC - Ultra Wide AVFS) // SEC28FDS prototyping (two version Tunnelborer-A = Dual-core, Faster TTM; Tunnelborer-B = Quad-core, Slower TTM)

Let me guess, 22FDX?

- AMD to be Wholly Owned Subsidiary of IBM by 2022. (Read the 10-Qs from 2003-2010, oops)

Whaaaaaat

 

[Qwertilot]

Or are you suggesting that OEMS are looking to clear out inventory of Intel products before they're willing to sell an AMD product?

I think that it is the suggestion, and it is surely entirely believable?

Even the most optimistic forecasts for Zen aren't expecting a performance increase over Skylake, just better performance/$ etc.

Ok, that's great if you're ordering new computers from scratch, but if you've already got and paid for a big stock of skylake based systems/chips? Really not at all relevant

You have to get rid of the Skylake systems first, discounted if need be. The same with Kaby of course. No way that'll do enough extra to usefully differentiate vs Skylake.

Whether it is actually happening to a big extent, I've no idea. I can believe it for big bits of the Kabylake product stack - consumer PC sales aren't doing well. I did think Zen was aiming at servers first and they're still selling quite well?

 

[agfkfhahddhdn]

.

 

[Sweepr]

The desktop CPUs are expected in Q4 2016 which is quite late to the game, but better late than never.

www.fudzilla.com/news/processors/40853-kaby-lake-s-brings-15-percent-more-i-o-lanes

 

[sm625]

I literally don't know what an "infield single" is.

Notice how the infield sits way back, because that is a big dude at the plate. That is what makes it easy for sluggers to get infield hits. At this point, even a play such as that would be a step up for AMD. I dont think I need to post a gif of what Bulldozer was, in baseball terms.

 

[IndyColtsFan]

Why? Assume conservatively a 25% ipc increase since SB for intel. 8 cores is only 33% more cores than 6. So that leaves maybe 10% more performance for workloads that can use all 8 cores, and a deficit in anything that uses 6 or less. And that assumes Zen will clock equally high as intel. You can already get a hex core intel for 400.00. Doesnt add up to me.

To sell for 500.00, I think they need ipc between IB and Haswell and competitve clockspeeds. There is a huge gap in price between intel hex and 8 cores where AMD can fit in at 5 to 6 hundred dollars, but they need better than SB performance to do it.

You're ignoring the higher costs of X99 to get those 6 cores. Additionally, Sandy isn't a dog in performance - it is still very competitive and is part of the reason Intel is struggling right now because people can't justify moving on to newer platforms. I've been a computer nerd since my first computer in 1981 and my SB platform is the first platform I've had that has been my main rig for at least 5 years. The main reason I'm looking to upgrade now is for platform improvements and getting that 25%-30% bump is icing on the cake.

You might be right on the pricing though. FWIW, I do believe Zen will probably be between Ivy and Haswell in terms of IPC and if it is closer to Haswell and has good OC headroom, Intel may be forced to cut prices to compete in the short term. That only helps us all and is one reason I am trying to hold off until Zen is released before I buy again, though with some of these sales on the 6700k and Haswell HEDTs, I might not make it.

 

[Minkoff]

Just in time confirmed...

- All AMD64 architectures have been scrapped. Scrapped being non-literal, they will launch but not have any successors.
- New ISA -> AMD64/P (Think Y86-64, Power ISA v3.0, and AMD64 fused//Semi-RISC/Semi-CISC ISA w/ high register count of Power, while all compatible with AMD64)
- Redesigned+From Scratch CMT Architecture -> Tunnelborer (IoT to HPC - Ultra Wide AVFS) // SEC28FDS prototyping (two version Tunnelborer-A = Dual-core, Faster TTM; Tunnelborer-B = Quad-core, Slower TTM)
- AMD to be Wholly Owned Subsidiary of IBM by 2022. (Read the 10-Qs from 2003-2010, oops)

It's a matter of time, for this, to be picked up by some tech related news site... :whiste:
Our sources say...

 

[NTMBK]

Just in time confirmed...

- All AMD64 architectures have been scrapped. Scrapped being non-literal, they will launch but not have any successors.
- New ISA -> AMD64/P (Think Y86-64, Power ISA v3.0, and AMD64 fused//Semi-RISC/Semi-CISC ISA w/ high register count of Power, while all compatible with AMD64)
- Redesigned+From Scratch CMT Architecture -> Tunnelborer (IoT to HPC - Ultra Wide AVFS) // SEC28FDS prototyping (two version Tunnelborer-A = Dual-core, Faster TTM; Tunnelborer-B = Quad-core, Slower TTM)
- AMD to be Wholly Owned Subsidiary of IBM by 2022. (Read the 10-Qs from 2003-2010, oops)

I genuinely worry about you sometimes, Seronx

 

[sirmo]

I don't see anything pointing to a delay. AMD has always maintained end of the year, with 2017 being a full year of Zen production.

Even that german article speaks of possible delays, which is true for any new CPU design.

 

[VirtualLarry]

I genuinely worry about you sometimes, Seronx

Yeah... some of that stuff is pretty off-the-wall.

I mean, I can believe the Via x86/64-ARM combo chips, those at least make sense. But scrapping AMD64 architecture, and merging with Power? And why would IBM even want AMD? IBM used to be in the x86/PC market, and left...

 

[Minkoff]

He's joking guys, don't take it so seriously

 

[RussianSensation]

If they have Sandy Bridge or Ivy IPC (min) with 8 cores, they'd be insane to sell for $300. I think $500-$600 would be a bargain.

I have to ask a straight forward question, can the people looking to buy 6-10 cores actually name 3 programs where the extra cores will be widely used by them? Video conversion / encoding app, rendering app, encryption app, compression/decompression app, etc.?

My litmus test is if there are no 2-3 programs that I use weekly that run faster on a max overclocked 6950X over a 6700K max overclocked, then Skylake wins every time. This is similar to buying tri-SLI/Tri-Fire and having 97-98% of games never actually use the 3rd card. Same story with PCIe 3.0 16x/16x SLI over PCIe 3.0 PCI 8x/8x = useless paper spec.

Also, compare the prices of 4x4GB DDR4 3466-3600 vs. 2x8GB DDR4 3466-3600 -> more premiums. X99 board --> more premiums --> higher end cooler required for a 4.4-4.5Ghz BW-E, more premiums.

The way I look at it is once more games use 6-8 cores, I'll just sell Skylake and buy Icelake-E. I honestly think it will be 3-4 years before a 6-8 core 4.4Ghz BW-E/HW-E starts to outpace 4.8Ghz 6700K.

This is why I think even if 3.8Ghz 8C/16HT Zen ~ IBV IPC, it's not good enough. If they are going to go that route, they should figure out how this CPU can run at 4.2-4.4Ghz with 6 cores, and 4.5-4.6Ghz with 4 cores. At least the IPC deficit can be somewhat narrowed with higher clocks.

Dmitri from Hardware Canucks did tests of 3930K OC vs. 6700K OC and the former lost in almost everything, while barely winning in multi-threaded apps. Someone who legitimately uses/benefits 6-8 cores with Sandy/Ivy IPC already has such a CPU. How else am I supposed to believe that since 2012 until 2016 this user didn't need 6-8 fast cores but come 2017 when Zen shows up, all of a sudden 8 cores > 4 very fast Skylake cores? Makes no sense. You cannot have it both ways.

That's why to me unless Zen ~ Haswell IPC at minimum and overclocks at least as good as Skylake, it's not that exciting. I suppose for workstation and productivity users, 8-core Zen for $499 is going to cost less than 1/2 of 6900K. It would be a good deal but to me and many users both of those processors are not worth the $. What's better, $1070 worth 'useless 8 cores' or $499 of 'useless 8 cores'? Neither when $289.99 CPU smashes them both while using less power, and the entire platform costing less.

 

[IndyColtsFan]

The way I look at it is once more games use 6-8 cores, I'll just sell Skylake and buy Icelake-E. I honestly think it will be 3-4 years before a 6-8 core 4.4Ghz BW-E/HW-E starts to outpace 4.8Ghz 6700K.

Yeah, that's a good point and one to consider for sure - buy the cores you will use today and then upgrade whenever more cores are needed for your particular use case rather than try (in vain, if you have noticed my constant flip-flopping in these threads ) to "future proof" everything. I don't know why I have such "paralysis by analysis" over this decision - it really isn't that big of a deal, though I have a sneaking suspicion this may be my last build for a long time, if not ever.

I suspect that you're right in that it will probably be at least one more upgrade cycle for most of us (3-5 years) before 6+ core support is the norm in most cases and until that time IPC matters more.

 

[NTMBK]

I have to ask a straight forward question, can the people looking to buy 6-10 cores actually name 3 programs where the extra cores will be widely used by them? Video conversion / encoding app, rendering app, encryption app, compression/decompression app, etc.?

My litmus test is if there are no 2-3 programs that I use weekly that run faster on a max overclocked 6950X over a 6700K max overclocked, then Skylake wins every time. This is similar to buying tri-SLI/Tri-Fire and having 97-98% of games never actually use the 3rd card. Same story with PCIe 3.0 16x/16x SLI over PCIe 3.0 PCI 8x/8x = useless paper spec.

Also, compare the prices of 4x4GB DDR4 3466-3600 vs. 2x8GB DDR4 3466-3600 -> more premiums. X99 board --> more premiums --> higher end cooler required for a 4.4-4.5Ghz BW-E, more premiums.

The way I look at it is once more games use 6-8 cores, I'll just sell Skylake and buy Icelake-E. I honestly think it will be 3-4 years before a 6-8 core 4.4Ghz BW-E/HW-E starts to outpace 4.8Ghz 6700K.

This is why I think even if 3.8Ghz 8C/16HT Zen ~ IBV IPC, it's not good enough. If they are going to go that route, they should figure out how this CPU can run at 4.2-4.4Ghz with 6 cores, and 4.5-4.6Ghz with 4 cores. At least the IPC deficit can be somewhat narrowed with higher clocks.

Dmitri from Hardware Canucks did tests of 3930K OC vs. 6700K OC and the former lost in almost everything, while barely winning in multi-threaded apps. Someone who legitimately uses/benefits 6-8 cores with Sandy/Ivy IPC already has such a CPU. How else am I supposed to believe that since 2012 until 2016 this user didn't need 6-8 fast cores but come 2017 when Zen shows up, all of a sudden 8 cores > 4 very fast Skylake cores? Makes no sense. You cannot have it both ways.

That's why to me unless Zen ~ Haswell IPC at minimum and overclocks at least as good as Skylake, it's not that exciting. I suppose for workstation and productivity users, 8-core Zen for $499 is going to cost less than 1/2 of 6900K. It would be a good deal but to me and many users both of those processors are not worth the $. What's better, $1070 worth 'useless 8 cores' or $499 of 'useless 8 cores'? Neither when $289.99 CPU smashes them both while using less power, and the entire platform costing less.

"Oh nothing scales to greater than two cores properly, I'll buy the C2D instead of the C2Q." Not everybody here buys a new computer every year, when I buy a CPU I want it to last

 

[IndyColtsFan]

"Oh nothing scales to greater than two cores properly, I'll buy the C2D instead of the C2Q." Not everybody here buys a new computer every year, when I buy a CPU I want it to last

I'm like that too, but consider this - buy a 6700k now for $300 and buy a (hypothetical) mainstream 8-core CPU in 5 years when it is really necessary for $400-$500 AND be on a modern platform, or pay $1000 for an HEDT 8-core CPU now with the more expensive platform and be saddled with that same platform for 5 years?

 

[NTMBK]

I'm like that too, but consider this - buy a 6700k now for $300 and buy a (hypothetical) mainstream 8-core CPU in 5 years when it is really necessary for $400-$500 AND be on a modern platform, or pay $1000 for an HEDT 8-core CPU now with the more expensive platform and be saddled with that same platform for 5 years?

Well, I'm really hoping that a Zen 8-core will be significantly less than $1000...

 

[swilli89]

As long as it's eight cores of Sandy IPC or better for sub $300 I can wait. I am more upset Intel hasn't increased cores counts than them hitting a physics wall on IPC. I want an eight core machine before I die and I don't want to pay workstation prices for it.

Exactly this^

 

[sirmo]

I'm like that too, but consider this - buy a 6700k now for $300 and buy a (hypothetical) mainstream 8-core CPU in 5 years when it is really necessary for $400-$500 AND be on a modern platform, or pay $1000 for an HEDT 8-core CPU now with the more expensive platform and be saddled with that same platform for 5 years?

I need 8c/16t cores today, I am currently on the haswell 4770.

Skylake isn't a big enough upgrade for me. And I want a 14nm 8 core CPU that isn't going to cost $1000+.

Also I am really tired of Intel segmenting the market and arbitrarily disabling features. For instance my 4770k doesn't support VT-D, while 4770 without the k does.

Pretty sure AMD will have a BE version with all the features enabled.

 

[Arachnotronic]

I need 8c/16t cores today, I am currently on the haswell 4770.

Skylake isn't a big enough upgrade for me. And I want a 14nm 8 core CPU that isn't going to cost $1000+.

Also I am really tired of Intel segmenting the market and arbitrarily disabling features. For instance my 4770k doesn't support VT-D, while 4770 without the k does.

Pretty sure AMD will have a BE version with all the features enabled.

We don't know what AMD plans to charge for Summit Ridge, could very well be 6900K-territory pricing.

 

[NTMBK]

We don't know what AMD plans to charge for Summit Ridge, could very well be 6900K-territory pricing.

For a dual channel platform, with far fewer PCIe lanes? And probably a lot less L3?