Intel says move to 10-nanometer chips still on track for 2017

[Shawn Knight]

Intel posted a job listing on its website last month in which it mentioned that mass production of products using its 10-nanometer manufacturing technology would begin approximately two years from the date the listing was posted (January 21, 2016).

This caught the attention of The Motley Fool's Ashraf Eassa who wrote a column on the matter. It was of particular interested because in mid-2015, Intel admitted that difficulties in the move to 10-nanometer had pushed the first round of consumer products based on the advanced manufacturing process back to the second half of 2017.

Initially, 10-nanometer products were to arrive in the marketplace this year.

If the job listing was indeed accurate, that meant Intel was pretty far behind in its move to 10-nanometer. As it turns out, however, the job listing wasn't accurate.

Intel's public relations team reached out to the publication and said the job listing contained "errors" and that it would soon be taken down. Sure enough, the listing in question has since been pulled. The PR team clarified that its first 10-nanometer products were still on track to arrive sometime in the second half of 2017.

Moving to a smaller manufacturing process has numerous benefits including (but not limited to) lower power consumption (which leads to better battery life) and improved performance as more transistors can fit on a single chip.

Permalink to story.

https://www.techspot.com/news/63869-intel-move-10-nanometer-chips-track-2017.html

 

[cliffordcooley]

A 10nm or smaller platform will be when I think about purchasing again. Until then I'm happy with the 32nm Sandy Bridge.

 

[Evernessince]

A 10nm or smaller platform will be when I think about purchasing again. Until then I'm happy with the 32nm Sandy Bridge.

You may actually want to think twice. Intel announced at CES this year that many of the technologies they are introducing into their processors are targeted at improving power efficiency and you should not expect performance improvements. They even went as far as saying that it's possible that performance will decrease as a result of the power savings.

 

[cliffordcooley]

Which begs the question of how they intend on increasing power efficiency of [6x & 9x]W CPU's without increasing performance.

 

[dividebyzero]

Which begs the question of how they intend on increasing power efficiency of [6x & 9x]W CPU's without increasing performance.

Kind of depends whether the increase is some narrow single canned benchmark or real world. The former is easy, the latter not so much.
An example of the former might be use of a new instruction set. Linpack for example can use AVX2 as well as AVX. The Xeon E5 v3 (Haswell) and E5 v2 (Ivy Bridge) are both 22nm products with similar architecture, yet 20 cores of dual E5-2687v3* yields 788GFLOPS (39.4GF/core) while the 24 cores of dual E5-2695v2 yields 441GFLOPS (18.4GF/core).

[Source]
Core speeds are roughly equal ( seven of the v3's 10 cores turbo to 3.2GHz,the remaining ones 3.3, 3.5, 3.5. Eight of the v2's cores turbo to 2.8GHz, the remaining ones, 2.9, 3.0, 3.1, 3.2) and the v3 consumes a little more power - the actual TDP headroom difference is 39%, while the v3 offers a 20% faster QPI link. * The graph has a typo. The E5 2697v3 should be E5 2687v3
A more dynamic cache/core structure could also increase perf/watt ( increasing core count and using a lower base/turbo would also work), as would decreasing branch misprediction penalties, and of course decreasing the process node and its attendant decrease in transistor switching voltage........but as with any marketing claims regardless of vendor, the numbers are always going to be best case scenario.

 

[DaveBG]

LOL, I currently have exactly i7 920 (with slight overclock though) and still have no issues or at least no issues that would push me to upgrade, BUT once they finally move to 10nm I will jump head first.

 

[dividebyzero]

LOL, I currently have exactly i7 920 (with slight overclock though) and still have no issues or at least no issues that would push me to upgrade, BUT once they finally move to 10nm I will jump head first.

I wouldn't let that chart scare you. It represents best/worst case scenario for throughput. Linpack is core/thread optimized and makes use of the increased operations per clock that AVX/AVX2 brings (The i7 920 is limited to SSE 4.2). Virtually all consumer software will show a smaller spread of results. Bear in mind also that the 920 has 4 cores/8 threads, while some of that competition have considerably more available.

 

[captaincranky]

LOL, I currently have exactly i7 920 (with slight overclock though) and still have no issues or at least no issues that would push me to upgrade, BUT once they finally move to 10nm I will jump head first.

Keep in mind the Xeon at the top of the heap has 14 cores and 28 threads. Plus there are TWO of them in the board they tested.

The CPU has a streetporice of $2900,00.....er EACH!

Here's the Intel ARK page on it: http://ark.intel.com/products/81059/Intel-Xeon-Processor-E5-2697-v3-35M-Cache-2_60-GHz

 

[dividebyzero]

Keep in mind the Xeon at the top of the heap has 14 cores and 28 threads. Plus there are TWO of them in the board they tested.

The CPU has a streetprice of $2900,00.....er EACH!

Here's the Intel ARK page on it: http://ark.intel.com/products/81059/Intel-Xeon-Processor-E5-2697-v3-35M-Cache-2_60-GHz

2697 was actually a typo on the graph. It was actually the 2687v3 that was being tested (10core/20thread) - a measly $1885 each compared to the 18 core $7000 E5-4669v3 and E7-8890v3 which really need to be installed into a 4 socket (and up to 8 for the latter) board respectively. The new Broadwell (v4) parts that supercede these will be 24 core CPUs.

 

[captaincranky]

2697 was actually a typo on the graph. It was actually the 2687v3 that was being tested (10core/20thread) - a measly $1885 each compared to the 18 core $7000 E5-4669v3 and E7-8890v3 which really need to be installed into a 4 socket (and up to 8 for the latter) board respectively. The new Broadwell (v4) parts that supercede these will be 24 core CPUs.

So AMD's "Zen", won't be much of a threat to those parts either?

 

[fktech]

I7 4790k - sweet spot!

 

[captaincranky]

You may actually want to think twice. Intel announced at CES this year that many of the technologies they are introducing into their processors are targeted at improving power efficiency and you should not expect performance improvements. They even went as far as saying that it's possible that performance will decrease as a result of the power savings.

AFAIK, Intel is already marketing a line of low power desktop chips in Skylake (possibly i3 only ?). They do carry lower clocks than the standard chips. (They have the same IGP as the standard chips though, Intel 530)
Here's that page: http://www.intel.com/content/www/us/en/processors/core/core-i3-processor.html

 

[dividebyzero]

AFAIK, Intel is already marketing a line of low power desktop chips in Skylake (possibly i3 only ?). They do carry lower clocks than the standard chips. (They have the same IGP as the standard chips though, Intel 530)
Here's that page: http://www.intel.com/content/www/us/en/processors/core/core-i3-processor.html

Intel also has a 28W mobile i3 with Iris graphics (~ 50% more performance than the HD 530) and eDRAM, and a 15W i3-6100U ( a favourite with notebook and tablet OEMs). Intel has already well and truly tipped its hand with regards low power with the 16-core/45W Xeon-D. At the other end of the scale it also has 165W / 28 core/56 thread Skylake-EP/-EX's waiting in the wings ready to do battle with IBM's POWER9 architecture.
Intel are about mix and match. Efficiency or speed or any permutation of both depending on what is required (even tailoring models for individual customers) has been part of Intel's dialling in process for years. Most parts are predicated upon performance-per-watt, but where needs must - such as stock traders needing absolute performance for stock analysis and trend prediction - Intel will happily produce parts such as the Xeon X5698 ( still Intel's fastest processor at 4.4GHz.

 

[Phr3d]

AFAIK, Intel is already marketing a line of low power desktop chips in Skylake (possibly i3 only ?). They do carry lower clocks than the standard chips. (They have the same IGP as the standard chips though, Intel 530)
Here's that page: http://www.intel.com/content/www/us/en/processors/core/core-i3-processor.html

Intel also has a 28W mobile i3 with Iris graphics (~ 50% more performance than the HD 530) and eDRAM, and a 15W i3-6100U ( a favourite with notebook and tablet OEMs). Intel has already well and truly tipped its hand with regards low power with the 16-core/45W Xeon-D. At the other end of the scale it also has 165W / 28 core/56 thread Skylake-EP/-EX's waiting in the wings ready to do battle with IBM's POWER9 architecture.
Intel are about mix and match. Efficiency or speed or any permutation of both depending on what is required (even tailoring models for individual customers) has been part of Intel's dialling in process for years. Most parts are predicated upon performance-per-watt, but where needs must - such as stock traders needing absolute performance for stock analysis and trend prediction - Intel will happily produce parts such as the Xeon X5698 ( still Intel's fastest processor at 4.4GHz.

too untalented to minimize the quote, sorry -RE "Intel also has a 28W mobile i3 with Iris graphics"

Three
Hundred
Dollars
for a two core i3? feel like I missed a memo, lol.

 

[Technician]

We heard about this a month ago, why is it news today?

 

[cliffordcooley]

It will continue to be news until launch.

 

[Technician]

OK, I'll post it again tomorrow then.

 

[cliffordcooley]

First it was months and now it is daily. Which repetition do you want to talk about?

 

[Technician]

It was last month, not months ago...if it's news everyday, it should be reported everyday, no? I am trying to make a point about the timeliness of news being a majorly important factor. Today's news is tomorrow's bird cage liner.

 

[cliffordcooley]

Bumping threads daily is usually frowned upon, but bumping them on a monthly basis might be refreshing. Especially to those that didn't see the topic the first time around. So I ask of you, please stop this non-sense.

 

[Technician]

You first.

I was making a point, not actually saying I would re-post this news everyday, besides in my facetious comment I only said I'd re-post it tomorrow, not every day, you are adding things yourself and attributing it to me.

 

[cliffordcooley]

I was making a point

No you were trolling and I called you out on it. If you have already heard this news then move on and let those who haven't read it for the first time.

 

[dividebyzero]

too untalented to minimize the quote, sorry -RE "Intel also has a 28W mobile i3 with Iris graphics"

Three
Hundred
Dollars
for a two core i3? feel like I missed a memo, lol.

The salient point is that the part is BGA mobile, so like all mobile parts - whether embedded/MXM graphics or BGA'd processors, the pricing is reflective of the fact that it isn't a generally accepted retail part and gives the OEMs the upper hand in their sales ( discouraging upgrades or repairs means another complete mobile sale sooner rather than later) as well as preserving their warranty structure.

 

[Technician]

No you were trolling and I called you out on it. If you have already heard this news then move on and let those who haven't read it for the first time.

If you seriously thought it was a troll, why did you respond? aha!
You knew it was rhetorical. I am calling you out on that.

 

[cliffordcooley]

You knew it was rhetorical. I am calling you out on that.

Well then I can honestly say that word is not in my vocabulary. Say what you mean, mean what you say, or say nothing at all. I've made my point, good day to you.