Intel's Future Chips: News, Rumours & Reviews

[fudgecakes99]

Dam getting my hopes up. Would love to see that performance on an air set up or even water cooled

 

[Cazalan]

2.032 vcore on water is not bloody likely.
The numbers on air are nice though.

 

[Turb0Yoda]

Hopefully that is not some cherry picked unit. Probably very likely though...

 

[juanrga]

Performance/cost is a misleading metric for CPUs, because 2x more transistors doesn't give 2x more ST performance. Also The A10 is made on a mainstream (cheap) 28nm planar process node, whereas the i7 is made on a leading (expensive) 14nm FinFET process node. The node advantage affects to parameters like efficiency, which is not taken into account in a performance/cost metric.

HBM on APUs is not expected before 2017.

What you just said makes no sense, Juan... The customers don't care if there are donkeys running in circles trying to get a correct result or if there are quantum mechanics used inside the innards of the CPU. They care about price and how it performs. You can decompose "price" into instant expense and future expense (energy bill, basically) if you want, but the overall idea is the same. That metric is the most solid of them all for the average consumer. If you want performance only, it makes even less sense, since well, performance is the metric you want to look at.

If you go server, it's performance/watt usually, because "future expenses" costs are always big enough to affect how the thermal properties of the chips affect their budgets in the long run. You can also add performance/application and something along those lines. And if they don't see perf/<something>, they use perf/$.

I have never ever seen any serious person talking about "performance/process node" when discussing a purchase; in the pro level and in the consumer level. And yes, process node affect efficiency, but efficiency is perf/watt.

You misunderstood my post completely. It was not about what certain customers care or not care. It was an explanation of why fabricating a CPU with 2x higher single thread performance costs about 4x to 5x more, not 2x more. Of course some people can chose the CPU that gives 80% of the performance by 60% of the cost, whereas others prefer maximum performance possible at any cost. My points is that CPU pricing is, in last instance, a consequence of the laws of physics (the IPC--complexity relationship is highly nonlinear), not because Intel is an evil company whereas AMD is a kind of non-profit ONG.

 

[-Fran-]

You misunderstood my post completely. It was not about what certain customers care or not care. It was an explanation of why fabricating a CPU with 2x higher single thread performance costs about 4x to 5x more, not 2x more. Of course some people can chose the CPU that gives 80% of the performance by 60% of the cost, whereas others prefer maximum performance possible at any cost. My points is that CPU pricing is, in last instance, a consequence of the laws of physics (the IPC--complexity relationship is highly nonlinear), not because Intel is an evil company whereas AMD is a kind of non-profit ONG.

Thanks for explaining that then. I agree with you in that front.

I would only question if it's always *that much more*. Since, for instance, AMD I guess pays for a full wafer, and after that they do binning, the cost per CPU is the same in terms of fabrication, isn't it? Then, you just apply economy of scales: if you have a *lot* of great quality CPUs, you might not need to sell them at such a premium and vice versa. If yields are bad, then you might. Price is a function of not only how expensive the node is, but how good or bad yields are and market demand and blah blah. What does history say about it? I really don't even have manufacturing prices from Intel or AMD on any given period of time to compare to actual "street" price.

Cheers!

EDIT: Corrected idea.

 

[fudgecakes99]

So for clarification, skylake unlocked processors are dropping with the first wave of cpus? Does that mean they're won't be k processors or are the k sku's just re named to s's now? Also any word on the extreme lineup xeons etc? I'm tryn to find some articles giving me some inclination of what's being dropped but i'm not finding anything.

 

[-Fran-]

And talking about efficiency: http://www.phoronix.com/scan.php?page=article&item=intel-5775c-netburst&num=2

Do you have friends or family still on old CPUs? Maybe this will compel you guys to make them buy a new computer, haha.

Cheers!

 

[de5_Roy]

FanlessTech: EXCLUSIVE: The Skylake boost
http://www.fanlesstech.com/2015/07/exclusive-skylake-boost.html
/edit

And talking about efficiency: http://www.phoronix.com/scan.php?page=article&item=intel-5775c-netburst&num=2

Do you have friends or family still on old CPUs? Maybe this will compel you guys to make them buy a new computer, haha.

Cheers!

LAWL
not the i7...may be an i5 with h170 board.
an i7 might make them implode from high performance delta.

 

[Isaiah4110]

And talking about efficiency: http://www.phoronix.com/scan.php?page=article&item=intel-5775c-netburst&num=2

Do you have friends or family still on old CPUs? Maybe this will compel you guys to make them buy a new computer, haha.

Cheers!

Lol! That's awesome. I gave my mother-in-law my old Pentium D (925 or 930, I can't remember which) last year, which isn't much newer than the CPUs they compared there... That's a crazy performance difference!

 

[Cazalan]

Intel could skip 10nm to keep the lead over TSMC/Samsung.

https://www.semiwiki.com/forum/content/4834-intel-skip-10nm-stay-ahead-tsmc-samsung.html

 

[juanrga]

Intel could skip 10nm to keep the lead over TSMC/Samsung.

https://www.semiwiki.com/forum/content/4834-intel-skip-10nm-stay-ahead-tsmc-samsung.html

What he suggests doesn't make sense to me. If Intel want to maintain the lead it would accelerate the development of the 10nm node and get it ready before other foundries. But lack of money and technical difficulties obligated Intel to delay the node to 2017. The 7nm node is more difficult to get than the 10nm node. Therefore, the 7nm node will not be ready before 2017, and TSMC schedules 7nm for 2017. Where is the lead?

 

[ern88]

Upgrading my i5 2500K to the i5 6600k with 32 gigs ddr4 ram. Asus maximus VIII HERO board.

 

[Alpha3031]

@juanrga Intel 10nm is delayed to 2017 for mass production (Cannonlake). TSMC is running 10 nm risk production (basically, the final "testing", with full size wafers from actual consumers) for 2017. If Intel can ramp and get products to market early-2018, it'll still lead, by the barest of margins.

 

[Cazalan]

Looks like Intel found something to do with their excess fab capacity.

3D XPoint memory

http://fortune.com/2015/07/28/intel-micron-memory/

 

[gamerk316]

Looks like Intel found something to do with their excess fab capacity.

3D XPoint memory

http://fortune.com/2015/07/28/intel-micron-memory/

Shot across the bow of Samsung, perhaps? Memory manufacturing is Samsung's wheelhouse after all...

 

[Cazalan]

Looks like Intel found something to do with their excess fab capacity.

3D XPoint memory

http://fortune.com/2015/07/28/intel-micron-memory/

Shot across the bow of Samsung, perhaps? Memory manufacturing is Samsung's wheelhouse after all...

Yes, the analyst who predicted Intel would get serious about memory was spot on.

Between the custom HMC (MCDRAM) and this that's quite an investment in memory tech.

 

[Cazalan]

I'm sure this was linked before but with the recent announcement that gives new perspective on this leak.

http://wccftech.com/massive-intel-xeon-e5-xeon-e7-skylake-purley-biggest-advancement-nehalem/

3D XPoint will be initially implemented on a DDR4 module (Apache Pass) as a drop in replacement for DDR4. Offering 4x the capacity. 6PB vs 1.5PB

 

[fudgecakes99]

So i should wait until the xeon chips drop instead of the consumer grade skylake chips. Then again i don't know the relevance for me getting server grade equipment in a gaming rig?

 

[Cazalan]

So i should wait until the xeon chips drop instead of the consumer grade skylake chips. Then again i don't know the relevance for me getting server grade equipment in a gaming rig?

This is largely just for HPC for the next 5y or so.

 

[fudgecakes99]

So i should wait until the xeon chips drop instead of the consumer grade skylake chips. Then again i don't know the relevance for me getting server grade equipment in a gaming rig?

This is largely just for HPC for the next 5y or so.

Servers.

 

[-Fran-]

http://www.phoronix.com/scan.php?page=news_item&px=VIA-Announces-Telecom-Sale

Interesting.

Do you guys think Intel might have gotten that chunk?

Cheers!

 

[gamerk316]

Almost certainly. Either Intel, or Qualcomm. MAYBE NVIDIA, but I think that's reaching a bit.

 

[fudgecakes99]

So it seems that august 8th is when the skylake news will drop at gamescom. That said, i own a hyper 212+ at the moment and plan on doing an open cooled water loop in the very near future with the new chip. My question is do you guys think the new socket lga 1151 i believe. Would i be able to use my current after market heatsink on it or would i have to buy a new one completely which i don't mind as their fairly cheap. Also whats the speculation on intel releasing an "enthusiast" version i.e. one without an included heat sink you think it'll decrease costs by 20-40 bucks?

 

[Supernova1138]

So it seems that august 8th is when the skylake news will drop at gamescom. That said, i own a hyper 212+ at the moment and plan on doing an open cooled water loop in the very near future with the new chip. My question is do you guys think the new socket lga 1151 i believe. Would i be able to use my current after market heatsink on it or would i have to buy a new one completely which i don't mind as their fairly cheap. Also whats the speculation on intel releasing an "enthusiast" version i.e. one without an included heat sink you think it'll decrease costs by 20-40 bucks?

I imagine the mounting brackets will remain the same as they have been since LGA 1156 for the mainstream Intel sockets, as the pin count isn't changing significantly, so the socket should still be the same physical size.

As for a chip being sold without a heat sink, you won't see a retail box on any mainstream CPU that doesn't have one. If you're looking to save a bit of money, you'll have to look for an OEM tray CPU, which doesn't include the cooler, but also has reduced warranty support.

 

[fudgecakes99]

So it seems that august 8th is when the skylake news will drop at gamescom. That said, i own a hyper 212+ at the moment and plan on doing an open cooled water loop in the very near future with the new chip. My question is do you guys think the new socket lga 1151 i believe. Would i be able to use my current after market heatsink on it or would i have to buy a new one completely which i don't mind as their fairly cheap. Also whats the speculation on intel releasing an "enthusiast" version i.e. one without an included heat sink you think it'll decrease costs by 20-40 bucks?

I imagine the mounting brackets will remain the same as they have been since LGA 1156 for the mainstream Intel sockets, as the pin count isn't changing significantly, so the socket should still be the same physical size.

As for a chip being sold without a heat sink, you won't see a retail box on any mainstream CPU that doesn't have one. If you're looking to save a bit of money, you'll have to look for an OEM tray CPU, which doesn't include the cooler, but also has reduced warranty support.

What do you speculate the prices will be? for the 6700k 400-300? or 250-350?