Intel Starts Qualification of Ice Lake CPUs, Raises 10nm Volume Expectation for 2019 (anandtech.com) 73
During its Q1 earnings call, Intel provided an update regarding its 10 nm process technology as well as the ramp up of its Ice Lake-U processor for notebooks, which is the company's first 10 nm design that will be mass produced and broadly available. From a report: Qualification for the new processors has already started, so systems based on Ice Lake-U will be available by the holidays, as promised. Furthermore, Intel believes that it will be able to ship more 10 nm parts than it originally anticipated. Intel started production of its Ice Lake-U processors in Q1, but Intel has been building up a stockpile of them first before they are sent to PC makers for qualification. Once the CPUs are qualified -- something that Intel expects to happen in Q2 -- the manufacturer can start sales/shipments of these CPUs, which will likely happen in Q3. Considering the lead-time required to get built systems on to store shelves, Ice Lake-U-based PCs are on track to hit the market in Q4 (something Intel reaffirmed today).
Re: (Score:2)
What Intel calls a node N process is what the rest of the industry calls a node N*2^(-1/2) process. This has been true for decades.
(Once upon a time process node N mean that the minimum feature size, typically the effective gate length for FETs, was N nanometers. These days it just means the process node that comes between N*2^(1/2) and N*2^(-1/2). I've even seen specific dimensions increase when shifting to a "smaller" process.)
Re: (Score:2)
What Intel calls a node N process is what the rest of the industry calls a node N*2^(-1/2) process. This has been true for decades.
That would imply that this has been true since 1999 at the latest. You'll have a hard time trying to convince me that in 1999 Intel was using different measurements.
Re: (Score:1)
Considering the lead time they've had to fix meltdown/spectre, I'll doubt that Ice Lake CPU's are completely immune to Spectre.
Like literately to remove Spectre as a potential vulnerability entirely, Hyperthreading would need to be redesigned (since it doubles the virtual cores by splitting the pipelines ALU's.) Ironically I think AMD's solution to HT was probably the better one, just design more cores and dispense with the pretense of having fake cores. Note their Zen tech is SMT just like Intel's. The Bul
Re:hard pass (Score:5, Informative)
AMD may have introduced it, but until Intel adopted it, it wasn't going to be implemented in anything.
Actually you are wrong. Microsoft started writing the x86-64 port to AMD processors before Intel had their own offering and in fact they forced Intel to switch to x86-64. At one point Intel wanted to create their own (incompatible) x86 64-bit extension.
Intel already had x64. (Score:3)
Look up Itanium, or the Itannic, as it was called, lol.
Re: (Score:2)
Re: (Score:2)
I remember very well how all that went down; Intel already had an expensive, hardly anyone was using x64 architecture, and M$ was not going down yet another lockin by intel.
This was yet another attempt to keep AMD at bay.
I personally think M$ was really tired of intels fucked up memory mapping; remember 4k chunks in the early processors? The 'everything goes thru the accumulator' era of programming?
x64 is a lot more fun to program with. (I started with the 8088, and I still write assembler code, even now.)
A
Re: (Score:2)
IA-64 is totally different. Like it is not even CISC. It is VLIW.
Re: (Score:2)
Further to that, AMD64 was already making strong inroads to the UNIX server market before Microsoft started supporting it with Windows. Sun had Solaris running on AMD64 pretty quickly, and their AMD64 boxes had better price/performance than SPARC for anything that wasn't massively parallel. Linux on cheap AMD64 boxes was also driving adoption. Until Intel came up with a replacement for Netburst with x86-64, AMD were eating them alive in price/performance and performance per watt.
Re: NOT VEGAN (Score:2)
My dominatrix will be unhappy. (Score:3)
She loves Rubber.
Re: (Score:2)
Their CPUs are NOT vegan.
Intel does have [wikipedia.org] Vegan CPUs.
Re: (Score:1)
Re:Laptop chips only, no desktop. (Score:5, Interesting)
Charlie at SemiAccurate had an article where he believes the issue is that Intel has an issue with higher frequencies at 10nm. Laptops usually run at slower processor rates, so this lines up with his analysis.
https://semiaccurate.com/2019/04/25/leaked-roadmap-shows-intels-10nm-woes/
Re: (Score:2)
Re: (Score:3)
You think the new Intel chips will have graphics better than the Vega 10 in the 3700U?
Re: (Score:2)
The 3700Us are also based on Zen+, not Zen2.
Re: (Score:2)
Re: (Score:1)
Laptops are a bigger market - approximately 166M/year laptops vs 88.4M desktops. Also, I would hazard a guess that most new desktops sold are low-spec generic machines sold to businesses for the purposes of running Word, Excel and Outlook.
https://www.statista.com/statistics/272595/global-shipments-forecast-for-tablets-laptops-and-desktop-pcs/
Re:Laptop chips only, no desktop. (Score:4, Insightful)
Re: (Score:1)
Re: (Score:3)
Re: (Score:2)
And likely the same way the 12nm 2080Ti kills the 7nm Radeon VII, also while using less power.
Smaller process doesn't guarantee improvements in either.
Re: (Score:2)
Re: (Score:2)
The new process is supposed to be just as good or better then the TSMC 7 nm process so this is good news.
I'm not sure that's a terribly high bar to hit.
AMD's current 7nm parts haven't been impressive vs. 14nm from competitors parts in either performance or power, only price.
If the AMD hypsters are to be believed, Zen3 is going to change the world, though, so here's hoping.
Eat this, AMD! We have you beat in vaporware! (Score:2, Insightful)
I mean, how long has Intel answering to Ryzen/Threadripper with "don't buy them, we'll make something better eventually" now? And it's not like they have any plans of how to address the whole Spectre family of nightmares yet without losing even the vaporware contest. And the U.S. government is concerned about Huawei.
Re: (Score:1)
lolz, several intel chips kick the ass off of Ryzen/Threadripper in the kind of apps I care about.
Heavily threaded games do well with it, sure, but I don't play games.
Re: Eat this, AMD! We have you beat in vaporware! (Score:3)
Threadripper is a nice but relatively untested design. It isnâ€(TM)t well enough understood from a security perspective as of yet. It is entirely possible that it has just as nasty issues as Intel has ever had. Remember, it took a decade (or more) to find the class of bu
but telerobots aren't robots. (Score:2)
Aren't you the lying moron that kept trying to argue that "telepresence communication robots aren't robots" and all kinds of other retarded shit, lol Iggy you fucking moron?
Telepresence communications robots aren't robots. At best they're telerobots, but "telerobot" is not a subset of "robot". A robot implies some degree of autonomy.
I don't see how this is relevant to the topic, though.
Re: (Score:3)
Yea, 6-8 months after AMD releases a chip, Intel has some sort of answer. 2017 was Ryzen first generation vs. Intel where the 7700k was the top chip. Intel had clock speed advantage, but Ryzen was better in 6 out of 8 areas. 2018 had Ryzen second generation, and Intel EVENTUALLY got the i9s out the door. So Intel did finally get a true response.
Now, AMD is set for third generation chips on 7nm to be released, and for the desktop market, Intel won't have a response until 2021 or so. The real questi
Re: (Score:2)
Re: (Score:2)
The 14nm 9900k has a lower TDP than the 12nm Ryzen 2700x, and I'm sure I don't need to point out which of those is faster.
Are you assuming since they'll have 7nm parts that they'll somehow reach finally surpass Intel's performance-per-watt metric? (Since they certainly haven't done so yet)
Re: (Score:2)
Re: (Score:2)
Thanks
Can anyone start preparing for lawsuits? (Score:3, Interesting)
If Intel doesn't have these 10nm chips actually shipping for real, can someone PLEASE make sure that there are lawsuits set to go. Intel keeps making FALSE statements to prop up stock prices, and no one has called them on it!
Re: (Score:2)
So now we know why they bought off the press. (Score:2)
I'd hate to be Kyle or the others out there Having to prop up their continued bullshit.
It must suck to sell your soul for insurance.
At least I got banned there before it folded up. :)
Calling him a GOP Faggot gut under his skin, somehow. :D
intel's 28-core processor with the 2000W refrigerator cooler, that took 2x 1000+W power supplies to run the mobo and peripherals was the demo I needed to see, to know they really have nothing but single threaded advantage.
My 10 year old system is doing great, clocking alon
Yield issues are likely to continue (Score:4, Interesting)
We've all already seen the report that Intel expects supply problems out to at least Q3, and in spite of assurances from them that "10nm is totally here guys!" We've seen nothing. There is supposedly a 10nm i3 in a China market only Lenovo. Anandtech got their hands on one and it wasn't exactly impressive. And of course their lineup is a confusing mess right now. I can find three different generations of processors, 7th, 8th and 9th in systems for sale today. And of course their "9th" gen cpus aren't really 9th gen, just slightly warmed over 8th gen.
They had better hope that 10nm pans out because they are about to be in a world of hurt. AMD's current chips are good enough in single thread performance and superior in multi threaded. Recent industry reports show that their sales have grown a healthy margin, not just from being competitive once again but also because outside of the big 3 SI's, Dell, Lenovo and HP, everyone is having major supply chain issues on Intel chips.
For the old timers out there like me, this is sounding A LOT like a repeat of 20 years ago. In 1999 the same thing happened when Intel had yield issues with its Pentium 3's and AMD dropped the original Athlon.
Re:Yield issues are likely to continue (Score:5, Informative)
20 years ago Intel's problem was that they'd let their marketing division steer the company, not their engineering division. Processors were doubling in speed roughly every 18 months, so customers were using the MHz clock speed as a quick-and-dirty benchmark for processor performance. Consequently, Intel's marketing division thus steered the company primary goal into clocking the CPUs faster and faster.
That resulted in a train wreck around 3 GHz, when the current leakage (which IIRC goes as the square of frequency) became so excessive that their Netburst CPUs became miniature ovens. Unfortunately for Intel, because clock speed had been their holy grail, the entire architecture for the Pentium 4 was designed around leveraging a high clock speed. They had to scrap it and go to their mobile division (where reducing power consumption had been more important than clock speed), and restart from the design they were using in their mobile Pentium CPUs (based on the Pentium II). That put them behind by several years, which is what allowed AMD to (briefly) capture the lead. (The 64-bit instruction set in modern Intel CPUs is an AMD instruction set because AMD was calling the shots when the first 64-bit CPUs came out.)
So no, it's not the same thing as happened 20 years ago. The problem then wasn't yields, it was excessive heat making it impossible to hit the targeted clock speeds. This is why clock speeds have barely budged in the last 15 years (the P4 hit 3.8 GHz in 2005). Instead, the primary focus for improving processor speed has since been increasing the number of cores and speeding up the execution pipeline (via wonderful things like speculative execution).
It's also worth pointing out that the nm between different companies are not comparable. Intel's 10nm is actually smaller than Samsung and TSMC's 7nm.
Have they got their Cobalt backplane fixed? (Score:5, Interesting)
Intel has had tremendous yield problems on 10nm [battleswarmblog.com], and word is it may have something to do with their decision to use a Cobalt/Copper backplane [semiwiki.com] for lower metal layers.
Anyone know if that's true and/or has been fixed?
spectre (Score:1)