News Intel Earnings: 10nm+ Tiger Lake Arrives Mid-Year, Company Withdraws Guidance

watzupken

Respectable
Mar 16, 2020
773
365
2,270
1
I feel the surge in data centers and laptop revenue is largely due to the sudden need to spruce up work from home capabilities due to COVID 19. At least its clear for laptops where people are snapping up the affordable ones so that its easier for 2 or more to work/ study from home. I suspect the demand will weaken for the rest of the year.
 
Reactions: bit_user

InvalidError

Titan
Moderator
I feel the surge in data centers and laptop revenue is largely due to the sudden need to spruce up work from home capabilities due to COVID 19.
While there is no doubt that covid-19 will have bumped sales from work-from-home needs, you also need to keep in mind that Intel was already on back-orders going into "covid season" so Intel's sales would likely have been even better without the supply chain disruptions.
 
Reactions: bit_user

Deicidium369

Permanantly banned.
BANNED
Mar 4, 2020
390
59
290
14
Last edited:

Deicidium369

Permanantly banned.
BANNED
Mar 4, 2020
390
59
290
14
While there is no doubt that covid-19 will have bumped sales from work-from-home needs, you also need to keep in mind that Intel was already on back-orders going into "covid season" so Intel's sales would likely have been even better without the supply chain disruptions.
Agreed there, probably a little soft, would expect Q2
I feel the surge in data centers and laptop revenue is largely due to the sudden need to spruce up work from home capabilities due to COVID 19. At least its clear for laptops where people are snapping up the affordable ones so that its easier for 2 or more to work/ study from home. I suspect the demand will weaken for the rest of the year.
Upgrading data centers is not something that happens in fits and spurts - it is a continual process - whether ppl work from home or at the office - they are still using the same data center equip - MAYBE the VPNs boxes were upgraded or increased licenses.

I have 65 people now working from home, rather than at the office - we had started to order the Palo Alto Networks end points, got the hardware really really quickly - took almost 2 weeks to get the licenses to allow them to start being provisioned and deployed.
 

everettfsargent

Reputable
Oct 13, 2017
111
22
4,585
0
So, since I'm the stupidest person in the world, I'll ask, once again, my stupidest question ever. When will Intel release desktop (or higher HEDT, workstation or server) silicon that has TPW north of 60W and at any process node that is less then 14nm? AFAIK, it would appear to be sometime this decade, century or millennia.

Is there some sort of bookie accepting bets for a given year in this decade? I'm thinking February 22, 2022.
 
So, since I'm the stupidest person in the world, I'll ask, once again, my stupidest question ever. When will Intel release desktop (or higher HEDT, workstation or server) silicon that has TPW north of 60W and at any process node that is less then 14nm?
Certainly not for as long as they have one record quarter after the other selling 14nm,there is no company in the world that would be that stupid.
 

Deicidium369

Permanantly banned.
BANNED
Mar 4, 2020
390
59
290
14
So, since I'm the stupidest person in the world, I'll ask, once again, my stupidest question ever. When will Intel release desktop (or higher HEDT, workstation or server) silicon that has TPW north of 60W and at any process node that is less then 14nm? AFAIK, it would appear to be sometime this decade, century or millennia.

Is there some sort of bookie accepting bets for a given year in this decade? I'm thinking February 22, 2022.
I at one time was an Intel HEDT user - Socket 2011v3. When I was building desktop systems for my engineers - i looked into the socket 2066 CPUs and decided to spec dual socket Supermicro Socket 3647 systems. Installed a single 28 core Xeon Platinum and 512GB of DDR4 ECC as well as an RTX6000.

The Xeon Scalable had processors to fit whatever use case, and had the advantage of more memory channels and a BMC. Did try out a Threadripper, but with no ISV optimizations, and it's janky memory controller, it was a joke. At the time my workflow look like this
AutoCAD, Autodesk PowerMill, Dassault Solidworks, Dassault CATIA, Siemens NX, couple of other Dassault programs and for simulation Ansys.

AutoCAD didn't need optimizations, but Solid Works, CATIA and Ansys did and performance on the AMD system suffered significantly.

Since then we have moved over to Creo.

So really, the HEDT market can be addressed by the i9900K and assuming the 8 and 10 core Comet Lake on the lower end and Xeon on the higher end. So to me, there isn't really an HEDT segment anymore. Once again, AMD shooting where the market WAS rather than where it will BE.

And since you are the stupidest person in world - YOUR WORDS

Mid year Tiger Lake will drop - 10nm+ Willow Cove cores (2nd gen) and the Gen12 graphics which are first to be branded Xe. This is Xe LP. They are talking about systems for the holidays - Ice Lake started shipping in volume to OEMs in May 2019. Looks like Tiger Lake will start shipping about the same time to OEMs - I bought my Dell 13 2-in-1 with Ice Lake in October - would expect to see Tiger Lake available in October as well - and hopefully the NUC11 around the same time.

3rd quarter 10nm+ Ice Lake Xeon with upto 38 cores drops - this is for single and dual sockets in the Whitley platform (new chipset, new socket, PCIe4) and the 14nm Cooper Lake (with bfloat16 for AI) will use the same platform, but will only be 4 and 8 socket. 2 very different workloads.

Later this year Rocket Lake-S drops - 14nm but with Willow Cove cores and a cut down Xe compared to Tiger Lake. Vastly improved IPC (1st gen Ice Lake Golden Cove IPC was increased 30%+ and the 2nd gen Willow Cove would likely see a similar increase over IPC on Ice Lake) So new cores, architecture, PCIe4 running on frequency optimized 14nm - so 4Ghz minimum desktop CPU - the last 14nm Desktop flagship - and the first not to be based on Skylake. Z490 boards already shown - socket 1200 for Comet Lake, and Z590 for Rocket Lake/PCIe4.

Sorry Later this year - not 2022 - that would be when the full production Intel 7nm (equiv to TSMC 3-4nm) will drop with the Ponte Vecchio Xe and Saphire Rapids Xeon. PCIe5 + DDR5.

Intel has had ONE single stumble in it's lifetime and all of a sudden they are incompetent - yet TSMC has been the opposite - FINALLY getting a node right, and they can do no wrong. You kiddies have a very limited understanding of the history of Intel and AMD / Global Foundry (before having to sell its manufacturing arm, because they could not afford to plow the money into creating new/next node, and needed to keep the lights on - Core had dropped around this time, obliterating AMD for more than a decade) And AMD has been "destroying" Intel for several decades now - and yet somehow this "destruction" leads to record quarter after record quarter.
 

Deicidium369

Permanantly banned.
BANNED
Mar 4, 2020
390
59
290
14
Certainly not for as long as they have one record quarter after the other selling 14nm,there is no company in the world that would be that stupid.
They do not understand business. IF that 14nm process was at TSMC the first + would have made it 13nm and the 2nd + 11nm and would be labeled as 10nm in it's current form. Intel doesn't care - doesn't need to impress Apple or AMD or Nvidia... so the heavy marketing about "7nm" (really 10nm class) isn't necessary.

Intel knows what the market wants - they have deep ties to the OEMs, and seems as though they are fine with Intel's 14nm and will happily build systems based on the 10nm stuff as well. When Ice Lake Xeon is released, PCIe4 will finally go mainstream and in high volume.

I love it when they start to talk stock prices - and you have to explain P/E ratios and how AMD has one of the highest at 160x (higher is NOT better or stronger) and the market average is 16x and Intel is 12x. IF AMD stock price was calculated at 16x *industry average) then it's stock price would be $5 not $50.
 
First generation Ryzen and threadripper were quirky, that's for sure. Had Intel answered then, no one would have cared for AMD outside of console space.
Intel tacked a couple cores to their existing lineup, and got hit with Meltdown and Spectre.
Then AMD solved most of the problems of first gen Ryzen with BIOS updates (including Spectre) and solved the rest with second gen (Zen+) and 12nm (which is very much like 14nm+++)
Intel answered by taking a couple more cores to they existing lineup, people started saying that Intel wasn't doing much.
Now AMD put third gen Ryzen out - and this time they double the amount of cores overnight, go to 7nm (which is very much like Intel's unusable 10nm), get the IPC crown, the performance per watt crown, the price per core crown, on their whole lineup including ones Intel can't hit (64 cores per package) so much so that industry leaders have to revise their licensing models...
Intel's answer is to activate smt on their lower tier stack.
For the first time since x86-64 came out, Intel is again a follower instead of a leader. That always hurts in high tech.
You can say what you want, nowadays the best performing HEDT machine has a 64-core AMD CPU in it and uses PCI-E 4 NVMe. You can't have that with Intel.
 

everettfsargent

Reputable
Oct 13, 2017
111
22
4,585
0
Since we all seem to be discussing vaporware CPU's, what if AMD releases a 5nm chiplet this year? Say like a 4950X with 32 cores (64 threads) running at 4.5 GHz all core turbo for an MSRP of ~$750US.

Oh and this is Intel's 2nd stumble, they are repeating their Pentium GHz at all costs back when AMD released the Opteron or some such. Again, I know that I am the most stupidest person ever on this planet, but for all intents and purposes, heat dissipation below the 14nm node is Intel's main problem, no chiplets to spread that heat and so 5+ years to a so-called real 10mn desktop node that will forever run slower then there 14++-++-++ node.

2nd stupidest question ever asked by the most stupidest person ever on this planet. Will Intel ever release a CPU smaller then the 14mn node that has a top all core turbo speed over 5.3GHz with 12+ cores? In other words, can Intel defy/defeat the conservation laws of thermodynamics? Meaning something that does not requite LNG as in a 247 DIY desktop or some such.

Right now, chiplet CPU's appear to be such a no brainer.
 
Last edited:

bit_user

Splendid
Ambassador
decided to spec dual socket Supermicro Socket 3647 systems. Installed a single 28 core Xeon Platinum
...
Did try out a Threadripper, but with no ISV optimizations, and it's janky memory controller, it was a joke.
If you ever populate that second socket, then you're in the same territory as ThreadRipper's "janky memory controller".

And its memory controller is not janky - it's just effectively a multi-CPU system inside a single package. So, if/when you go multi-CPU, you'll be right in the same place (unless you were talking about the 32-core TR, which I'll grant was something of a weird bird).

So really, the HEDT market can be addressed by the i9900K and assuming the 8 and 10 core Comet Lake on the lower end and Xeon on the higher end. So to me, there isn't really an HEDT segment anymore. Once again, AMD shooting where the market WAS rather than where it will BE.
Strange you should say that, after crediting Intel for their 8-core desktop part. Now who was first to make 8-core desktop parts? Hmmm...

Also, AMD was first with PCIe 4.0 - and yet they're not shooting for where the market will be?

Intel has had ONE single stumble in it's lifetime and all of a sudden they are incompetent - yet TSMC has been the opposite - FINALLY getting a node right, and they can do no wrong. You kiddies have a very limited understanding of the history of Intel and AMD
Wow, someone here sure needs a history lesson!

Just a few of Intel's other notable stumbles:
  • Their i740 GPU
  • Itanium (and putting all its 64-bit eggs in that particular basket).
  • Betting on RAMBUS, which delayed integrating its memory controller, giving Opteron another andvantage.
  • Netburst (i.e. the infamous Pentium 4 uArch)
  • Hyperthreading - releasing it prematurely, then having to walk it back, in the Core/Core 2 era.
  • 14 nm delays, forcing the cancellation of the Broadwell desktop parts and release of the Haswell Refresh chips
  • Xeon Phi
  • Trying to compete with ARM-based cellphone SoC
  • Purchasing Nervana, only to kill it off about 2-3 years later.
There's a small history lesson, for you.
 
Last edited:
Reactions: TCA_ChinChin

bit_user

Splendid
Ambassador
Certainly not for as long as they have one record quarter after the other selling 14nm,there is no company in the world that would be that stupid.
The problem is that Intel can't afford to get left behind. And that's exactly what'll happen, if they stay on 14 nm, too long.

For one thing, 14 nm has obvious deficiencies in perf/W, which leaves open a market window for not only AMD, but also ARM-based server chips to start gaining traction. And once those guys get a foot in the door, it's going to be difficult and expensive (i.e. via reduced margins) for Intel to claw back that marketshare.

Right now, the datacenter market is growing so fast that everyone is having trouble keeping up with demand. Intel will continue to own most of this market, if for no other reason than that AMD doesn't have enough allocation of TSMC's production capacity to address much more of it than they already are.

It's not hard to succeed in a growing market. Just wait 'till demand falters, and then we'll start to see how much of that market Intel truly owns.
 

bit_user

Splendid
Ambassador
I love it when they start to talk stock prices - and you have to explain P/E ratios and how AMD has one of the highest at 160x (higher is NOT better or stronger) and the market average is 16x and Intel is 12x. IF AMD stock price was calculated at 16x *industry average) then it's stock price would be $5 not $50.
Did you ever wonder why Intel's P/E is so low and AMD's is so high?

It's a signal that the market is betting big on AMD, and is wary of Intel. And this is in spite of Intel doing stock buybacks and paying out dividends, while AMD has done neither. So, that really casts a harsh light on big investors views on Intel's prospects.

I'll agree with you on one thing: when a company's P/E is out-of-whack, you have to really ask yourself why, and question whether you're truly smarter than the market. But, that goes as much for low P/E companies as high ones.
 
The problem is that Intel can't afford to get left behind. And that's exactly what'll happen, if they stay on 14 nm, too long.
They aren't left behind,they have 10nm products on the market for years now,there is just no reason for them to release them for desktop because very few people care about perf/watt on desktop...let alone on enthusiast.
For one thing, 14 nm has obvious deficiencies in perf/W, which leaves open a market window for not only AMD, but also ARM-based server chips to start gaining traction.
Intel has a ton of products for the server marketspace that are much more efficient than their 14nm CPUs,they released nervana NNP in m.2 and pci card format that are hardware accelerated AI chips and you can update your existing systems with them, stick as many of them into the same system as you have power to feed,and here's the kicker intel provides software for it and doesn't rely on the users to make their own.

As someone else already mentioned intel looks forward while AMD caters to a market that used to exist.
https://www.techarp.com/business/intel-nervana-nnp-i1000-details/

It's a signal that the market is betting big on AMD
Yup intel is a boring and save stock,that's why people are betting on AMD,it's not save,people hope that it will continue to go up but the way they are forced to burn through nm they are going to slam into a brick wall pretty soon.
 

bit_user

Splendid
Ambassador
They aren't left behind,they have 10nm products on the market for years now,there is just no reason for them to release them for desktop because very few people care about perf/watt on desktop...let alone on enthusiast.
First, that story line is at odds with all the news and the slipped schedules and everything else that one would know about the situation, from following the news over the past few years.

Second, it doesn't explain why Ice Lake server CPUs have slipped, and slipped, and slipped. You can argue that the consumer desktop doesn't care about power-efficiency, but you can't argue that about data centers.


Intel has a ton of products for the server marketspace that are much more efficient than their 14nm CPUs,they released nervana NNP in m.2 and pci card format that are hardware accelerated AI chips and you can update your existing systems with them, stick as many of them into the same system as you have power to feed,and here's the kicker intel provides software for it and doesn't rely on the users to make their own.
Huh? What does that have to do with anything?

As someone else already mentioned intel looks forward while AMD caters to a market that used to exist.
https://www.techarp.com/business/intel-nervana-nnp-i1000-details/
LOL. They killed their Nervana product line.


So much for that point.

Yup intel is a boring and save stock,
I think you mean safe, and it doesn't explain why Intel is trading below average. The only explanation for that is that investors fear Intel will underperform in the foreseeable future.

that's why people are betting on AMD,
I know Wall St. gets criticized for being a bit of a casino, but most stock is held by institutional investors, and they're not usually irrational enough to let something get so over-valued unless they think there's a realistic shot of it rising to justify that valuation.

the way they are forced to burn through nm they are going to slam into a brick wall pretty soon.
When AMD designed Zen, it was in bad financial shape. They are now in massively better condition, and that means having a lot more resources they can pour into designing future iterations of Zen. As they do, they'll be less dependent on TSMC's manufacturing advantage, and the architecture can better compete on its own merits.

I'll agree with one thing: I'm not buying AMD stock (nor do I own any). That P/E is outside my comfort zone, for sure.

Anyway, as with all predictions and prognostications, only time will tell for sure.
 
First, that story line is at odds with all the news and the slipped schedules and everything else that one would know about the situation, from following the news over the past few years.
OMG you mean to say that click bait articles make everything seem worse than they are?
Second, it doesn't explain why Ice Lake server CPUs have slipped, and slipped, and slipped. You can argue that the consumer desktop doesn't care about power-efficiency, but you can't argue that about data centers.
Because server CPUs in general sliped? Because they shifted to GPUs and AI? Because whatever else?
Huh? What does that have to do with anything?
What do inovations that are important for the server space have to do with the server space?
LOL. They killed their Nervana product line.


So much for that point.
Yes they made money with nervana,found an even better alternative and are going to make as much and probably even more money with the new thing,so much on how to make money.
I think you mean safe, and it doesn't explain why Intel is trading below average. The only explanation for that is that investors fear Intel will underperform in the foreseeable future.
Intel themselves came out and said that the amount of money they make right now can't keep on forever, they said that 10nm will be less profitable so of course not many people are going to continue to buy up stock.
The last two years intel's profit was double than normal,it wouldn't be natural for this to keep going.
When AMD designed Zen, it was in bad financial shape. They are now in massively better condition, and that means having a lot more resources they can pour into designing future iterations of Zen. As they do, they'll be less dependent on TSMC's manufacturing advantage, and the architecture can better compete on its own merits.
Irrelevant,intel also improves 14nm again and again and that is considered a bad thing(and they are not stuck on 14nm it just doesn't make sense to go lower yet),AMD doing that on whatever they end up on will be a good thing?
 

bit_user

Splendid
Ambassador
Yes they made money with nervana,found an even better alternative and are going to make as much and probably even more money with the new thing,so much on how to make money.
No... they didn't even recoup their investment on Nervana. They killed it before they shipped any new product, and only one product with a few customer commitments was kept alive in order to fulfill those commitments.

they are not stuck on 14nm
Wow, it's like you're in a parallel dimension.

How do you explain Comet Lake? Why on earth would they release another 14 nm chip that only gains performance by virtue of running yet hotter, if they could just switch to 10 nm, whenever they wanted?
 

InvalidError

Titan
Moderator
Irrelevant,intel also improves 14nm again and again and that is considered a bad thing
Improving 14nm can only go so far. The power and die size are becoming unmanageable.

As for "power efficiency not mattering for dekstop", pretty sure that given the choice between two systems of practically identical performance where the largest difference is power draw, most people would pick the more power-efficient option. The only people who really do not care about power efficiency are enthusiasts since efficiency gets worse at the high-end and flies out the window with overclocking. Most people get increased efficiency from entry-level to mainstream parts gaining efficiency much faster than high-end.
 
No... they didn't even recoup their investment on Nervana.
Well, if you say so,I'm sure you looked into intel's wallet.
They brought out two products one on m.2 and one on pci and unless you have specific numbers they made a profit.
Why would they buy havana if nervana was a failure?
Wow, it's like you're in a parallel dimension.

How do you explain Comet Lake? Why on earth would they release another 14 nm chip that only gains performance by virtue of running yet hotter, if they could just switch to 10 nm, whenever they wanted?
For intel to reach ryzen's performance in cinebench they need to match core/thread numbers,if they went for 10nm...they would still have to match core/thread numbers...
It just doesn't make sense,they are going to increase cores until they reach the same amount of cores,or the maximum they think makes sense and only then are they going to go to 10nm.
Improving 14nm can only go so far. The power and die size are becoming unmanageable.
The 9900k has a smaller die size than the I/O plus cores of the 3700x while including an iGPU that is something like 40% of the die area.
Power usage is great unless you heavily overclock to run AVX.
As for "power efficiency not mattering for dekstop", pretty sure that given the choice between two systems of practically identical performance where the largest difference is power draw, most people would pick the more power-efficient option.
At idle and looking at the web intel's 14nm is drawing the same if not less power than ryzen.
Even gaming at 5Ghz is way below TDP.
https://www.tomshardware.com/reviews/intel-core-i9-9900k-9th-gen-cpu,5847-11.html

 

InvalidError

Titan
Moderator
The 9900k has a smaller die size than the I/O plus cores of the 3700x while including an iGPU that is something like 40% of the die area.
About 2/3 of the Zen 2 CCX die area is L3 cache. As for the IO die, can't do miracles when the die size is effectively dictated by how long the edges have to be to cram all of the IOs including the two CCD interfaces that end up consuming ~15% of the total IOD area, leaving the middle of the IOD with a fair amount of empty or very low density space. Area-efficiency on the IOD clearly wasn't a design priority. A multi-chip design will always have a significant handicap on total die size because of the chip-to-chip overheads.
 
About 2/3 of the Zen 2 CCX die area is L3 cache. As for the IO die, can't do miracles when the die size is effectively dictated by how long the edges have to be to cram all of the IOs including the two CCD interfaces that end up consuming ~15% of the total IOD area, leaving the middle of the IOD with a fair amount of empty or very low density space. Area-efficiency on the IOD clearly wasn't a design priority. A multi-chip design will always have a significant handicap on total die size because of the chip-to-chip overheads.
My point was that the die size isn't unmanageable not to bring down AMD.
Also the L3 cache has an immediate influence on a few benchmarks while nobody takes the iGPU into account when talking about multitasking or multithreaded workloads.
 

bit_user

Splendid
Ambassador
Well, if you say so,I'm sure you looked into intel's wallet.
They brought out two products one on m.2 and one on pci and unless you have specific numbers they made a profit.
They announced and then cancelled them.

Why would they buy havana if nervana was a failure?
The two companies are competitors! Intel bought Habana because they realized that Nervana couldn't compete. So, they ate their mistake and forked over $2B, which isn't exactly chump change, even for Intel.
 
They announced and then cancelled them.


The two companies are competitors! Intel bought Habana because they realized that Nervana couldn't compete. So, they ate their mistake and forked over $2B, which isn't exactly chump change, even for Intel.
Ohhh so they canceled nervana without bringing out any product and then they bought out havana so that havana couldn't compete with the products intel didn't make...makes so much sense.
Then maybe you should be complaining about how much Intel increased the price of their 8000 and 9000-series CPUs.
Why should I be complaining?!?
For existing 7th gen SKUs Intel added cores or at least threads while keeping prices the same,you can get a real 8 core now with higher boost clocks, for the price of the 7700k.
 

ASK THE COMMUNITY