Intel's upcoming 10nm and beyond

It doesn't matter what the marketing bs from both sides makes them think they 'want', what most need is a 4C/8T CPU still, even today.

Higher core count CPU's? I'm still waiting for an answer to my pertinent questions from this post here.

We've had higher core count platforms for years now. Nothing has changed.

Spec's, node 'supremacy' and other useless tags and buzzwords that don't bring an equivalent real-world performance increase for the workloads that matter today is the real BS. No matter who is spouting it.

 

If Intel can't deliver desktop / H level laptop 10nm CPU's, 4c/8t or otherwise - resulting in Intel delivering nothing at all, Intel is in deeper doo-doo.

Intel isn't even giving estimates for delivery of 10nm / 7nm desktop or H level laptop CPU's, Intel has all but given up on giving firm dates, and is hoping no one notices.

After 6-7 years of broken 10nm promises, Intel has entered that rarified realm of fantasy that few commercial companies succeed in reaching...

Intel is firmly ensconced, in the Twilight Zone.

Intel's failures with 10nm are so well worn that Intel needs to start spinning new yarns about 7nm, and how 7nm is going to merge seamlessly with Intel's 10nm fantasies, Intel is now zooming right through the Twilight Zone, and is heading to the outer reaches of... The Outer Limits!!

Intel is so far out in fantasy land it's hard to maintain any semblance of interest in their BS.

I'll wait until the 10nm or 7nm desktop / H level laptop CPU's arrive, until then I'm skipping the intervening 14nm reruns.

 

Still can't stay on topic, huh?

Intel doesn't have to deliver anything you want, even if they promised it. They're still supplying what most need.

The only one in fantasy land is you it seems. If you can't wrap your head around the fact that mere spec's/nodes/marketing do not a market make.

 

You are flying under the power of the same rarified fantasy Intel is floating, you aren't even making sense.

" They're still supplying what most need."

Wow, you've given up too. Sorry, it's not easy to take that kind of reality reset, Intel is no longer a trustworthy source for silicon process progress promises.

What a sad day.

 

I asked you and anyone else to show me why a high multicore CPU platform was 'required' today. No answer.

The AMD fairy dust must be powerful if you can keep deflecting a direct question over and over.

For the record, yeah, I've admitted many times that a high core count platform is exactly what a small minority need, but that doesn't change the fact that jack and jill checking their facebook 200 times a day needs it too. The sad part is you believe it to be so.


Edit: AMD is not immune to changing previous rollouts either, here is the latest example.

See:
https://www.tomshardware.com/news/amd-third-gen-threadripper-roadmap,39254.html

 

Agreed. Most consumers today do not need a 6, or 8, or 12 and 16 core CPU. Most would be fine with a dual core, 4 threaded CPU easily and still have power to spare. The server segment is a different beast all together which is why Intel seems to be targeting that first for upgrades. The laptop/ultrabook segment is also another money maker and larger segment than power HEDT users. Again it will see the first roll out next month on 10nm. Again we will see desktop 10nm or 10nm+ when we see it. For now most users with a 14nm+++++ CPU are going to be just fine waiting until then without issue or missing night and day performance jumps. Since I game at 4K 144hz I think I will be waiting for 7nm Intel parts in couple years and focus on GPU upgrades from this point on unless either company pulls a huge performance jump out of their hat.

 

You people (not all of you but a few most vocal) take this brand dislike to another level. I swear, it's like it brings you erotic pleasure, or you get paid to criticize Brand X (replace Brand X by AMD/Intel/Alienware/MS/Nvidia/EA/Epic). Don't you have more interesting things to do? Serious question.

Instead of arguing about rumors or about "nm" values (which will NOT automatically give better performance than other brand's product which uses higher "nm" process) or about hardware or software some of you won't even use because you are too old to enjoy it (meaning you don't enjoy playing games or you don't enjoy maximizing performance of your system in benchmarks) or because it won't fit your usage (in case of 16-core CPUs, which are not required by vast majority but might still be useful to few people) why not do something else like help someone with an issue or perhaps play a video game (if you are not too old to enjoy these), or watch someone else streaming it on Twitch/YouTube? Or perhaps discuss your favorite TV series or Anime series or maybe a sports game. You know, just like rational, mentally stable (without OCD) people do ;-) As of right now, both this forum section and Gaming section reminds me of this comic:

 

Quoting myself (for ~3 years now...) "I asked you and anyone else to show me why a high multicore CPU platform was 'required' today. No answer."

And still no answers to this very simple question, but thanks anyway.

Sunny Cove arch and beyond is very interesting to read about, hope to see it arrive soon at a Lenovo near me. When Intel offers that kind of information, they've usually followed up with better than expected results. What I have also discovered and have repeated here endlessly is that the spec's still are not fully indicative of the platform as a whole either, in every-day use. In other words, buy the latest and most powerful platform, always and period, for the workloads you rely on today, to be as productive as you can.

And, it's not the 10nm node that has me excited about SC, it is the design changes underneath that make it worth waiting for if it will be delivered in the next ~130 days or so. At least for a near-future, mobile platform upgrade.

The reasons stated by Intel for their abysmal 10nm 'delivery', even it that officially happened in 2017 with a single product, actually makes sense and shows how the company has made changes so that something similar doesn't happen again. Much more than I ever heard any other company state when their spin at/near the top went out of control and they were left holding the steering wheel when going downhill full speed and with no brakes either.

AlexusR, it is not a matter of being simply too old to play games or to maximize benchmarks or watch some mindless video, it is a matter of perspective. If that is what you see as 'fun', more power to you. To me, it is worse than work; it is a waste of my time. Just like TV and sports are too, for me. Family and spending my time with them is what is actually important, not what I can do with tech, which I consider simply as a tool, like a hammer or a screwdriver is to a carpenter.

 

IMNSHO, few things are more important than punching through the societal web of lies and deceptions that take advantage of most, holding them back from reality.

Picking at the edges of what is false and deceptive until you can peel off that thin veneer and show the long obvious truth's hidden from view of the in-attentive is my job day to day, and has been for 40 years and more. It's like second nature with me, that's why I can see it from a distance while others can't even see it when it's staring them in the face. That's why I solve problems much faster than most.

It's amusing for some to turn discussions personal - I suppose for you?, I don't find it interesting as I don't get off on it, but apparently you do, right? It's easier to devolve a conversation into personal jabs instead of addressing the issues technical and factual in a real sense, so you can avert your eye's from the uncomfortable truths.

That's why you took it away from being a factual technical discussion based in the reality about how Intel has failed for 6-7 years to deliver products based on 10nm. Intel can't defend it, and you certainly are unable to do so.

And now Intel is so overextended on their promises for 10nm product delivery that Intel's credibility has disappeared completely.

Intel can no longer rely on even the most rabid Intel devotee's to swallow further 10nm BS, so now Intel is doubling down on their 10nm fantasy world by adding promises for 7nm - not because they can do it, but because it's the next expected progression in the process.

" Oh, so Intel is going to follow 10nm with 7nm, that makes sense; I guess everything is going to be ok after all. And, I can continue to buy 14nm CPU's until Intel delivers 10nm or 7nm. Hey!!, now we've got 2 chances to win with Intel!!" - Hopelessly Gullible Intel Devotee's

Even @tilleroftheearth and @Talon can't possibly defend Intel's continuing failures, so now all of a sudden the guys telling everyone that the Intel 8c/16t 9900K overheating when OC'ing is the "Bee's Knee's" while drawing 400w on a 95w TDP, isn't what most people need - they can live with 2c/4t CPU's, that'll do'em!

If you've been blinded by the corporate BS line for so long you can't see or speak the truth, it can be painful to see. I'm one of the few that bothers to try to punch through the BS and wake at least a few people up to look and re-evaluate what's going on in the cold clear light of day.

There are plenty of rah-rah Intel / Nvidia fan's out there. One lone factual truth teller can't bring them down. Don't worry about that. They'll be fine.

The only danger is I might wake you up, and yes - I am talking about you. Whoever you are. Stop wasting your money on their overpriced BS, and start getting value for your money.

Sharp smart, Shop S-Mart

 

It is not required, of course, just useful to some people in a very limited cases ;-) And yes, some exaggerate the importance of it (same goes true for everything else like importance of "7nm" or whatever), but that's just the kind of people they are - they desperately try to find any theoretical reason to win their argument. Because sadly some people have lost any meaning of life other than winning those and even more sad that they cannot comprehend this fact anymore ;-)


But again, this is not important at all. What's important is to spend your time on more useful or pleasurable things.

Well, there you go, this is much more productive or at least more pleasurable than arguing about brands, "nm" numbers, theoretical performance of future architectures and limited usage cases on forums ;-) That was my whole point.

 

It's a question designed to de-rail the harsh light of reality away from Intel's 10nm (/ 7nm) failures into a pseudo intellectual pastime for you, and I am not interested in feeding your fascination with diverting effective discourse into pages and pages of useless drivel.

I have been using high core count CPU's and multi-CPU computers for many decades and the work flow for their use is well known. I've built some of the first of their kind and written the software kernels and drivers to bring them to life.

Please stop asking as I'm not going to waste my time answering and noone else should either. You know you are asking a pointless time wasting irrelevant question that is OT from the main point.

Intel has promised high core count 10nm CPU's for 6-7 years and they have failed to deliver them, and now adding insult to injury Intel is now adding 7nm to the fantasy yarn they spin. Not because people don't need high core count CPU's, because they do, but because Intel failed to do so.

Enough is enough, I'll check in again when Intel's delivered 10nm or 7nm desktop / H level laptop CPU's.

Or, IMNSHO, when Intel finally cries "Uncle", and signs up for outside fab's to successfully spin up their product releases into full production quantities, which is where I see the pressure to deliver taking Intel in the short, and the long term. .

 

I didn't want anyone to miss the link/source for my post#1541 above.

See:
https://www.anandtech.com/show/13699/intel-architecture-day-2018-core-future-hybrid-x86/2

See:
https://www.anandtech.com/show/13699/intel-architecture-day-2018-core-future-hybrid-x86/9

The answers in the second link ring true to me.

This seems like a company that has taken a mega-punch to the gut but still thinks with its head. So far, their actions support their words.

 

And, after reading the same excuses and reasons said in similar and various forms by Intel over 6-7 years I still wish Intel the best in their seemingly never ending quest for 10nm perfection.

I do recommend Intel sign on with a working 7nm process and license it for use internally or contract for fab's to produce their production products until Intel finally nails down their 10nm process.

Hopefully there are rational options for Intel to take rather than continuing to flounder and embarrass themselves further.

At some point, it's going to affect the bottom line, and then who knows what power play will do that could harm Intel far worse than taking action now to produce competitive products.

It's probably already too late, unless Intel has already started down that road of getting external help long ago.

 

You really need to get over what happened 6 or 7 years ago with Intel.

Open your eyes to what is around you today. Companies, like people, change.

 

Ha, I guess that's all you've got. It's not enough.

Intel wouldn't be giving such detailed excuses every time they've failed if they thought that they were done and that was enough.

Yes, Intel's changed, time to move to another platform...

Isn't it nice that AMD has a nice new spin of a successful platform coming up soon? AMD, with semi-regular semi-reliable upgrades in process, design, and hopefully performance at a more reasonable cost to take the place in our lives that Intel used to fulfill?

I was just as excited as the next guy about Intel 10 years ago, still with the hope - and belief - that Intel were going to "make it happen" as regular as rain.

But, like you say, Intel's changed, time to move on.

 

Yeah, Intel's changed, just like I have.

But some things still remain the same after all these years.

See:
https://www.cpubenchmark.net/compare/AMD-Ryzen-7-PRO-3700U-vs-Intel-i7-8665U/3433vs3434


The above is what is touted for the new ThinkPad X395, listed as 'coming soon' on the Lenovo website. Hardly revolutionary vs. what Intel has offered for a while now.

See that 30% single core advantage. That stuff is real and really useful in everything you run.

Not just a new 'spin' or a semi-reliable upgrade that is what their competitor is trying today to make stick to the wall.

I don't see any excuses from Intel. I see products that I actually want to buy.

Those Intel hating glasses have you seeing cross-eyed, I think.

 

Bang on, exactly what what I see. I am looking forward to my next 10nm ultrabook (next month?) for the road as it's mostly all I want/need these days. The desktop build will be there for gaming relaxing a bit at home but for road use I want something more portable.

 

Intel has more to worry about 10nm. They got some serious manufacturing problems with yield and quantity.

Microsoft recently blamed Intel for low sales of Windows because.of Intel shortages. That is how bad it is with Intel manufacturing.

I agree. I will believe the 10nm hype when I see it. Also if Intel can meet supply demands.

So far AMD is expected to sell their 12 core 3700x for $349... Can Intel even compete?

 

Intel is as worried about 10nm as you or I am able to build one ourselves.

Which fab do you work at to confirm the problems that you state? Oh, you don't work there, I see.

Microsoft (or any other company) can spin their own version of reality to suit their purposes for their shareholders, and they do. But that same factoid also supports my belief that AMD did not take anything major from Intel's share of the pie either. Sorry, you simply can't have it both ways.

As far as I can tell, Intel is competing very well even with products from 4 years ago now vs. anything AMD can provide today.

The recent steps and about-face that Intel has had to do will once again leave AMD standing at the bus stop wondering when the next one is coming. But it was never a question of can they (Intel) compete.

AMD is the one still trying to catch up. And I applaud them for their efforts and raise my glass to their continued success.

Anyone that actually needs a high core count platform better not-have waited for AMD to supply it to them (I'm talking for paid-for work here). Those have always been available from Intel and the used market.

Anyone that buys a 12 core though just because they can? Well, more power to them.

See:
https://www.cpubenchmark.net/singleThread.html

Note that on a 4K monitor at the native resolution I can't see where the first AMD system shows up when I make the first CPU the Intel Core i9-9900KF @ 3.60GHz (top CPU, yeah; you know the one that everyone laughed at not so long ago).

This is where the real $$$$$$$ are (for the masses) and where Intel hasn't been touched even with a 3 year AMD invasion so far...

The facts, today and still, speak for themselves.


See:
https://www.cpubenchmark.net/compar...2970WX-vs-Intel-Xeon-W-3175X/3316vs3345vs3420

The Intel server variant in the comparison above shows what Intel can do with 5W more TDP.

I can already hear the shouts, but the cost! Yeah, if I could double my output today for twice the one time cost? Done. Can you deliver yesterday?

If actual performance is required, Intel is far from being written off anytime soon.


Back to the non-issue of the nodes, 10nm is effectively done and merely a stepping stone for Intel today. 7nm and beyond is where we may see it hitting on all cylinders again. The node is not the story anymore (if it ever was, Intel's 14nm++ process is still proving that today).

See:
https://www.anandtech.com/show/1431...oadmap-refined-nodes-specialized-technologies

See:
https://www.anandtech.com/show/1427...tion-of-10nm-ice-lake-cpus-raises-10nm-volume

Even if nothing quoted above comes to fruition, Intel is hardly worrying about nodes and other such nonsense today.

AMD and their supporters better take heed and forget about chasing mere 'scores' with more cores, while leaving the actual performance a user can use and feel on the table for Intel and others to run with.

Of course, this little snapshot of today isn't where the future rests, for either company.

I know that Intel offers real and tangible improvements for its newest platforms as a whole, and can pinpoint their customers with precise accuracy, even if those platforms aren't applicable to me, every single time.

AMD, on the other hand, is still in a kind of self-promoting mode for 'more cores for the people' and I think is ignoring what they should be focusing on, instead (actual performance increases over existing platforms, theirs and their competitors).

The next few months will be very interesting. With all the IP Intel is putting into its (near) future platforms, the future does look good indeed. Node warriors need not apply.

 

Still with this nonsense I see.

OK, first, here is what I posted in another thread a little bit ago:

"You are correct on Intel. 10nm+ (Ice Lake with Sunny Cove) should have LOWER transistor performance than 14nm++. That is from Intel itself. But, 10nm+ will be CLOSE to 14nm++. Then there is the IPC increase from Sunny Cove. That should give 11% IPC, which is the average Intel generational IPC jump from new core arch. To put that in perspective, from all rumors out there, that is a 4-5% estimated IPC over Zen 2. Not that impressive, but still a win. So the question is how much of a frequency regression Ice Lake will have.

Further, Intel plans for "S" skus is 14nm Comet Lake. Rocket Lake is rumored to be 14nm++ as well, meaning if they don't move that sku to 10nm, then Intel is trying to jump straight to 7nm in 2021-22. [Tiger Lake-U is the current plan for 10nm next year, not Rocket Lake]

Now, we also have the rumors surrounding AMD, such as the Zen 2 12-core running at 5GHz. We do not know if that is single core boost or all core boost.

If we look at the AMD page ( here), we see they only list the single core boost. Even assuming the same, that means reduce 5GHz by 200MHz and you have a chip estimated to run 4.8GHz with 50% more cores.

Now, we have already seen the 65W 8-core match a stock Intel 9900K. The only question is what frequency was used. If the 13% IPC number is to be believed, then we are talking about around 6% IPC over the current Intel chips. At 4.8GHz all core, you should see the equivalent performance of an Intel 9900K/F at 5.088GHz, or 5.1GHz, estimated. If the 12-core CPU does hit 5GHz, and there isn't a regression from having two dies on the chip, then that would be equivalent to the 9900K running at 5.3GHz, except that it also has 4 more cores, meaning it would be matching heavily overclocked Intel CPUs in single thread, potentially, while also having more multi-thread power.

Meanwhile, AdoredTV released a vid discussing the upcoming Zen 3 Milan server product. There is not an assurance it will come to consumers. With threads, although it speeds up workloads, it is not linear. For example, SMT can add up to 33%, approximately, over a non-SMT CPU (talking both AMD and Intel). By adding two additional threads per core, what you are doing is adding a more efficient queue system to lower down time of the cores, thereby speeding up processing, roughly. This also needs a more efficient and better scheduler to cause that not to cause slow downs. So you may get a case where it takes SMT from 33% up to 50%, but it will not be doubling the performance. This also can increase core heat, which can require lower frequencies (not always the case, but it IS a possibility that one would need to be aware of).

As to the stacked mem I/O, that would really be a boon. People have argued with me saying that the latency is too high on HBM for this use, but they have ignored the lower latency optimized HBM2, as well as the higher speed HBM2 chips that have been shown, which faster speed with roughly same latency means lower real world latency. Then, you just need to keep the HBM fed from DDR. Imagine 16GB-32GB of HBM2 or HBM3 on the die, which would have a latency between 30ns-60ns (note, AMD's memory call latency is already around 60-80ns), which would be fed by a larger ram pool off chip. This is coupled with the bandwidth of HBM2, which would be 512GBps to 1TBps bandwidth, which would be fed from an 8-channel DDR4 or DDR5 system, which would be 160-320GBps, approx. Also, those speeds are peak, not continual. Overall, that means, even with the latency, the bandwidth should make up for it in the context of a datacenter CPU.

So, either way, we have about 2 weeks to Lisa Su's keynote at CES. Then, it will be about 5 weeks from that that the first Zen 2 CPUs drop. That means in about 7 weeks, all the reviews will be out and we can put the silliness on saying AMD isn't competing on the high end to bed.

Also, when I mentioned 10nm, I have to re-emphasize Intel has ONLY said they will have low core count "u" and "y" type variants and entry level Xeons. That is NOT desktop parts. That is not high core count server parts. That is very limited."

With that said, we can discuss Intel lying on power consumption that throws off TCO. Although not perfect, AMD's TDP seems more accurate than Intel's, especially when estimating power consumption. With Intel's 14nm chips going to stay power hungry, while AMD's chips are already shown to be doing well with power consumption, it really is easy to see which way the wind blows.

Then there is the rumor of AMD shooting for a Q1 2020 Zen 3 chip for servers. If true, so long as this is a true full re-arch, we could see another 6-11%+ IPC jump. Now, there are some rumors that TSMC 7nm+ has some frequency regression, even though it will have an area reduction and power savings. So it is in a similar situation of going from 14nm++ to 10nm+ for Intel. That leaves the question on whether the IPC gains overcome that of the frequency regression, which, as I mentioned on the four-thread SMT, that could bring up to 16% performance for workloads able to use it. This means that Intel's Ice Lake-SP, which may be limited due to it being alongside the Cooper Lake-SP, would be facing another jump from AMD, which is bad news bears.

And if you cannot tell Intel is scared from their own words, you certainly are not paying attention.

Already, AMD is estimated by analysts to reach double digits in server either end of this year to the H1 2020. That is a HUGE windfall of income for AMD, which they have been investing heavily into R&D.

What is more interesting is that due to the Intel shortage, AMD is filling the gap. But AMD cannot fulfill ALL of the needed chips for the shortfall. There is still limits on TSMC's free fab time for ramping 7nm. With both CPU and GPU releases, this could be why TR was pushed from 2019. Another possibility, with the news of Zen 3 Milan server chips in Q1 2020 is that TR will go directly to Zen 3, bypassing Zen 2 altogether, which would also be an interesting turn of events.

But, Intel's shortage is confirmed to exist through Q3 2019. This is why Intel has only produced high margin parts. They have cut production of high-volume, but low performance and margin parts, which AMD has taken that segment of the market pretty much. If AMD does deliver on an 8-core 105W chip that performs as well as an overclocked 9900K for nearly half the price, then Intel is in SERIOUS trouble. Why? Because Intel, while having low production, would then be stuck in a price war which will harm their margins, which low volume mixed with lower margins means lower revenue and profit. They just had to revise their 2019 estimates down considerably. With new AMD CPUs putting pressure on Intel's pricing, AMD will likely have trouble keeping stock due to demand, and they know it, which is Intel's ONLY saving grace.

 

Too bad Intels TDP doesn't even make reasonable sense in any known universe.

That particular chip pulls 380W at max load (God knows you can probably hit 400W somehow). Whilst the 2950x pulls 180W.

Also please read the article before you reference it.

 

Please don't give me a wall of text that I have already read in another thread. A link would have sufficed.

If you can't respond directly to the points I've made, please resist attempting a simple and half-hearted reply that has no bearing on what I say above.

Go read the links I've posted in the posts above, your post simply repeats many of the same things but with an AMD twist them. Yawn.

Where are the counter-arguments to the core of the real issues here? AMD still playing catchup three years after Intel's monumental screwup(s)? And yeah, I agree their screwups were plural. But in all that time, Intel is still ahead of AMD where the meat of the workloads lie, even today.

I am showing you what Intel is, and you and others keep telling me what AMD will be. So far? I haven't seen any of that future to have panned out. Great, we have cheap (high) multicore platforms for people to play with. So? Those have always existed in retail and the used markets. I'm not taking away anything AMD is providing to the people that actually need those platforms. But those needs are few and far between in the entire market that Intel and AMD call 'home'.

And don't get me wrong, I want AMD to take the lead so Intel needn't hold back any more than it has to. Right now? Yeah, it has to. If Intel repeated the Core performance from so many years ago today, we would be seeing Intel getting pieced apart, all in the name of 'fairness' to an underperforming AMD. I don't want that, Intel doesn't want that and neither do their shareholders. I think they know what they're doing to get maximum return on their investment.

Before replying again with what you wrote to someone else in another thread, make an effort and try actually replying to the points made here instead.

 

Eh... Seems to me for majority of workload AMD is ahead, at least for anything I do. For 1440p gaming Intel performance lead is negligible or worth it to me. Ryzen wins for me at least. I do simoke compiling and compression stuff, Ryzen wins easily for what I want. To each their own.

I don't see evidence AMD is playing catch up. I see the exact opposite, Intel is playing catch up.

I am excited for Zen2. I have zero enthusiasm for Intel's plans.

 

Why does it have to make sense to you? The only issue(s) for me is to have a stable system and let TDP fall where it may. As I always have. The design choices made by the manufacturer of the platform I am considering are secondary to the performance/productivity it can provide.

I'm talking about pure, raw performance here, discussing power numbers is silly. Just like anything in tech, if all else is equal, more power in = more output (productivity). This isn't rocket science.

If I or anyone else needed something like the W-3175x, do you think that actual power drawn would deter them or me? Especially if I was twice as productive (not a given) as my competitor? For a few $$$ dollars difference a year?



Look at the ridiculous assumptions I've made in that screenshot above. Yawn.

A lower power option makes sense in mobile (mostly) and when required to have a few dozen to a few hundred or more systems in the same area/building. A CPU like this, if I could effectively use it, would be bought by the single units (max 9 and only for redundancy/backup purposes). At these low numbers, for non-server workloads, what any particular chip 'pulls' vs. another, much weaker, chip, is a non-issue.

 

Eh... For me TDP and power consumption is VERY important. I think for many people innovation isn't just about pure power, it is also efficiency. No one cares for example if an engine has 2000 hp if it is too heavy and consumes too much fuel.

250 watt or 300 watt CPU is a meme to me. Hell my cooler is a 200 TDP... And well I like that the 2700x runs under 70c with superpi.

Glad to see 3700x will be 105 TDP that will sink an i9 for 2/3 price.

 

I am genuinely happy for you.

You're still wrong about the majority of current workloads having AMD being ahead, but I understand the effect those glasses have.

Note that you do concede Intel is still in the lead, at least for some of your workloads. The 'worth it part' I leave up to you.

A 'win' for me would be to be above where I am now, AMD isn't in the running (look at the PassMark links above to see why).

What was your previous platform before your AMD desktop you're enjoying today?

 

For my workload Intel is sorely behind....

I haven't used AMD since AMD64. Dunno how that effects that 2700x is King for what I wanted.

Seems you are accusing me of being a blind fanboy. Just saying you seem to be coming off as an AMD basher.

Last AMD GPU i used was HD5770 if that matters.

 

Please expand on why a single platform with lower TDP and power consumption is so important to you? Curious about the reasons and to the numbers behind your decision to make it so.

If we're chasing performance here (and that is what I'm talking about), your engine example assumes the worst, but in Intel's case, it's very far from being 'too heavy' (over double the 'score', remember) or that it consumes 'too much fuel' (see the power table above).

 

Better question is why would I want a 300 TDP CPU? I don't need another heater for my housing

 

Let's cool it with the fanboy comments and such. I've already deleted enough posts today. I can always give some vacation time out as well...

 

That is the exact opposite of what you said and I quoted you saying above.

I don't know 'AMD64', but that still doesn't answer my question about your previous desktop platform?

So, even after acknowledging your choice and trying to have a conversation with you makes me an AMD basher, huh?.

The facts I've presented don't make me anything except someone who values core truths over marketing bs, from both sides.

No, a better reply would have been to just answer the question.

I don't like lasagna, but that doesn't make it a non-choice for others.

 

You are being absurd, not addressing the IPC changes and estimated clock speeds, and calling it catch up when Intel has nothing. You don't respond to legit critiques.

Then, you try saying people will want to buy the $18K Cascade-SP over buying the 64-core Rome Zen 2 for likely 1/3 the cost (potentially less).

Are you being intentionally obtuse?

 

Edit: Sorry, I just saw in the cleanup my post was moved to page 7. My apologies!!! Sorry about that.

 

No, not intentionally, but I could ask the same about you.

Show me something that is available today, right now that I can buy. That is my point. Period.

Playing crystal ball future gazing is not my forte, and neither is gambling with my tech budget either.

I hope that anyone buying this kind of performance isn't just relying on the specs on the box?

A 'legit critiques' is one which I have made a point about, above. Not a random post dropped from another thread.

 

No one buys a CPU unless forced when a release is incoming in less than two months!

In fact, almost every tech reviewer has a wait recommendation for at least until Computex to see if AMD delivers. It also is when Intel is dropping Cascade-SP, possibly Cascade-X, and Ice Lake-U on 10nm with limited production. As such, it really isn't a waiting game, it is 2 weeks from the conference reveal of performance and tech reviewers seeing performance on sample machines, which then the release is a month and a week from that date (enough time to get the money to buy one).

Also, that post was my response to Ole, which wanted to get my analysis on the Zen 3 rumor with 4 thread SMT, which also required discussing where things are with the mobile parts for laptops, which I did not include all of that information in reposting.

Now, on GPUs, that is a different story. Intel will have 10nm and 7nm graphics parts starting next year and 2021, respectively, if rumors are correct. I see them focusing on datacenter rather than consumer. Also, the amount of performance increase on iGP is nice, but it depends on which AMD CPUs are being compared. When you look at the "U" series AMD parts, it is a winner. When you look at "G" series AMD parts, Intel only finally reached the performance of existing "G" series APUs (referring to the 2000 series APUs, not 3000, and definitely not the next round with Navi). So it is curious. I am hopeful that they can really get that off the ground and it seems they have plans to use an interposer with multiple compute units split up into chiplets. That, plus their AI chiplets, and other custom tailored chiplets, for datacenter workloads really could give Nvidia a run in that segment. As for consumers, I really think they are just going to hit the midrange next year. 2021 might be a shot for the high-end.

You also keep pointing to single core performance. When the win in that will shrink in 2 months once released to be fairly negligible, but the cost for the competitor CPU, which also will have equal or more cores, will cost 50-70% of what that 8-core costs, people will buy the lower cost part.

This already has been proven with AMD market share for the DIY community, where you see last fall as an inflection during the black friday sales, which has continued all the way until now. Once the new chips drop, Intel may hold the crown on single core performance, but very few people pay 60-100% more for a product with only single digit single core performance wins over another product, especially if the other product can scale when multi-core is needed.

That is based on mindfactory's numbers, not me making it up. It also is seen in GN's data from people using affiliate links.

Edit: I forgot to mention that Intel entering in mid-range actually would be great on strategy. People love seeing the top in high performance, but they forget that the largest volume of cards comes from mid-range sales. As such, Intel would do well doing that with a 10nm limited production capacity, while then going high-end on 7nm, which is equivalent to 5nm/3nm TSMC or Samsung (wikichip has it closer to 3nm, if using Intel's SRAM data estimates for transistor density, which may not actually have that density on HP designs).

 

I can agree on many of your points below. Though many are still based on future events, like I keep saying makes your points moot today.

I too hope people actually know what their primary workloads are and purchase accordingly.

For the Jack & Jill masses, why buy a new platform when their old one is just as fast?

Conversely, anyone that needs true high core count platforms doesn't need me or you to tell them this. But for the masses, blindly saying you can save $$$$ by buying a slower system today seems like a disservice to AMD, in the long run.

That is why I keep saying test against your current workloads with your current platform vs. any newer platform you're considering. If there is no improvement, then merely saving 60% to 100% or more over a competitors option where you do get some increase in performance makes no sense because that is still $$$$ down the drain for no material gains.

If someone needs something today, they simply buy today. If they can wait, they will. No argument there.

Saying that what is available today will be negated in some future tomorrow can be played by both sides. It's only a game if you play along.

I refuse to play those games. See the above on why doing so may get you in a not-so-good position, tech-wise.

 

A lot of AMD talk in here for an Intel thread.

https://fuse.wikichip.org/news/2293...and-packaging-plans-10nm-in-june-7nm-in-2021/ -- Some pretty interesting stuff and interviews with Intel guys. Apparently 10nm went into high volume production at the beginning of 2019.

"Intel also gave an update on their 10-nanometer node. 10 nm has gone into high-volume production at the beginning of the year. “We are delivering, without compromise, on the original performance and scaling targets,” said Renduchintala. Their first-generation 10-nanometer node will follow by two additional inter-node optimizations – 10nm+ and 10nm++ planned for this year through 2021. It’s worth noting that even after 10 ramps, Intel will continue to leverage their 14-nanometer node for mix applications such as cost-sensitive products and to speed various developments wherever it makes sense."

 

There are times to not wait and times to wait when worrying about future tech and upgrades. Many times, if you have to buy today, like a system dying on you, you may not have saved for the optimal setup you want and just get something to get by. I've done that with a cheap laptop after my MB died. It worked sufficiently for small tasks, but wasn't up to the full cadre of my workload.

Aside from those limitations, there are other things to consider, such as release cadences and whether the hardware has an upgrade path. For example, you could pick up an X470 board and have support from the MB manufacturers to support the upcoming Zen 2 CPUs. On Intel's side with the Z390 boards, we do not know. We know Intel has planned two different chipsets for the upcoming Comet Lake CPUs. Now, some great BIOS engineers (like Prema) have allowed for the newer CPUs to work on older systems. But that isn't a guarantee.

Moreover, if the difference is a matter of, just as an example, 6% in single threaded performance, but Intel is nearly 40% behind in multi-threaded workloads, then it is a disservice to oneself to pay the Intel premium, not a disservice to AMD. It is not blindly saying to buy AMD, it is literally calculating the price to performance ratio, and if the premium for the faster chip is above a certain preset premium to make it worth it (which very well may be around 10-20% more than the other chip), then, even with adding in the premium, it is not worth it. When you do that on TCO, where the power consumption numbers actually come in, as well as computing density, then you also have to examine how Intel's 14nm performs on power consumption versus the 7nm TSMC process, which considering a stock 9900K was matched with an ES 8-core 65W chip, and that the 105W 8-core at stock may be able to compare to Intel's 9900K overclocked to 5.1GHz while consuming around 170-190W, then we are starting to see a turn. Or when comparing AMD's 250W CPUs to Intel's over 350W power draw, once again, when talking datacenters, that TCO equation starts shifting. Intel's saving grace, however, at least for now, is that they have a higher single core performance which is better for per core licensing. On socket licensing, AMD has the win. Now, AMD tested the waters with a limited run on Epyc 7371 chips, which were speed optimized chips running at 1950X stock speeds with Zen 1 chips. If AMD can put out a speed optimized Epyc chip in volume (another theory why AMD dropped TR from the 2019 roadmap), then even with the couple hundred MHz boost expected on Cascade-SP chips, AMD may be able to compete for single core licensing. That is theory and not today, however.

And, let's discuss expected release times for both sides, shall we?

Intel:
Cascade-SP released around computex 2019
Ice Lake-U released around computex 2019
Comet Lake-S released Q4 2019-Q1 2020 for consumers/ Q2 2020 for commercial systems
Ice Lake-SP "before late 2020" (sometime before Q4, possibly H1 2020)
Cooper Lake-SP possibly H1
Rocket Lake S - possibly Q3-4 2020, with possible slip to 2021
Tiger Lake-U 10nm++ - around mid-2020

Now for AMD:
Zen 2 Epyc and Mainstream - Computex 2020, available for sale July
Zen 3 Milan EPYC CPUs - Q1 2020
Threadripper - no show, possibly Q1 2020
Zen 2 APUs - between Q3-2019 and Q1-2020 (depends on the rumor)
Zen 2 commercial desktop offerings - if following cadence, between Q4 2019 and Q2 2020

Note - AMD's dies are scaled across all offerings, so the lack of names in comparison should not be taken as a lacking on their part of products for the different segments. If I had used like Sunny Cove uArch for Intel, it would have simplified their list, same if using process for 10nm or 14nm, but as Intel has more of a mix, the names need spelled out.

I also said we must wait on reviews and reveals. And there is a chance that the 10-core comet lake on 14nm will beat the 12-core AMD offering, which is why AMD has considered holding back the 16-core (along with it destroying part of their threadripper lineup).

This is why the process discussion needs to take place with a view to uArch development. Intel has 3 uArchs already developed. AMD has Zen 2 (released soon), and Zen 3 in the can, so to speak, and are working on Zen 4 (which Intel is tweaking their third architecture still as well, planning on 7nm, which is still in development).

That is why the race is closer than ever and just buying the fastest today, especially when for 8-core offerings you are paying $500 approx., when the Zen 2 will be between $220-$330, means seeing what may come, which is revealed in 15 days if memory is served (keynote is ahead of the conference on either May 27 or 28), is what any consumer that can wait should wait. Why? Because they can theoretically get the X470 or X570 boards and drop a throw away chip in that board while waiting 1 month.

Now, that isn't possible if needing a production rig. That only speaks to end consumers, not professionals. But even buying a 2700 for a couple hundred, using that for a month and a half to two months, then dropping in the new chip would make sense for many people, while reselling the 2700 to recoup maybe 50% of cost (so the producer loses $100 on that deal). For some, that will make more sense than buying a $500 CPU today, while the competitor drops their CPU for $350 with 50% more cores and single digit losses on single thread performance a couple months later.

This will vary by use. And with changes to offerings, the equation will shift. As mentioned, Intel's Comet lake may use the Sunny Cove uArch, which would increase their IPC by approx 11%. That will increase single core performance, but is at minimum around 6 months out. If you always play the wait and see game, you will forever be waiting. But, you also do not want to be caught in a situation like me, where I bought a 6700K 1-month before the 7700K dropped. I knew the risks, but I also needed the build. Hence where your example of buying for now comes in. If you can wait, you should wait, at least when you are within a couple months of the new drop (since Intel doesn't cut pricing on old chips).

 

Don't Intel's own latest released timetables say Ice Lake-SP is sampling now, with production shipment 1H 2020..

https://fuse.wikichip.org/news/2318...u-and-i-os-to-follow-by-tiger-lake-next-year/ -- YES they do.

 

Can you cite what volume capacity Intel has on 10nm?

Rumors are that Intel scaled back 10nm at three of their four fabs that had 10nm ready, while the fourth one scaled back the number of lines at the fab in Oregon.

Intel has been a paper tiger on releases as of late, which is uncharacteristic of Intel, but has become a pattern. So, if you can show me why I should believe volume of Ice Lake-SP in 2021, I would love to see it (other than Intel's own PR spin).

Also, if we are going with that rumor, then Ice Lake -SP with limited production will be facing Zen 3 Millan chips in Q1 2020, with whatever IPC changes and whatever 7nm+ will bring to the table (which I previously noted may include a frequency regression from 7nm, which is also what Intel's 10nm+ is compared to 14nm++).

Edit: the reason I say this is 10nm allegedly is in mass production, but there are limited supplies of Ice Lake-U until around Q4 2019. Considering the numerous delays and lack of volume, there is reason to doubt Intel will hit their Q1 2020 target. They very well might, but it could also slip a couple quarters for Ice Lake-SP. Those will likely be lower powered server offerings, considering it will be sandwiched with Cascade-SP still out, followed by Cooper-SP a couple months after the stated release of Ice Lake-SP. That suggests Ice Lake-SP is a lower volume product as far as output. Now, they will need to get it out in the first couple quarters as they need the 10nm capacity to release the Intel GPUs around mid-year, likely Q3, maybe Q2. Then you have Tiger Lake-U mid-year for laptops and netbooks.

This plays into why Apple is trying to get an ARM chip for their laptops by 2020 to 2021. Already Intel has shuttered their 5G modem and is in talks to sell the IP to Apple, which produces on TSMC's fab.

As such, they said they will use 14nm to fill out their line gap, meaning that due to expanding 14nm capacity, they need to recoup costs, and plan to do so on the above stated products.

Also, there is a question of yields, which rumors state 10nm yields may be as low as half that of 14nm at the moment. If true, that is less than 50% yields when TSMC and AMD are rumored to have already reached over or around 70% yields per wafer. This likely is why Intel's GPU is using multiple chiplets, which means smaller chips, similar to the small nature of the Ice Lake-U chips, thereby increasing effective yields due to defect density.

So capacity and yields is important to remember, and the rumor mill, yes it is still rumors (like Intel would say such negative information publicly), but rumors can sometimes bear fruit.

 

Rumors.. Rumors are worth absolutely nothing to me. Additionally this latest meeting was an investor meeting, stating they are accelerating 10nm production, have been stock piling chips since early beginning of the year for their 10nm launch next month in ULV laptops, and server segment 1H 2020. I highly doubt Intel wants to open themselves up to lawsuits by lying to investors about what they're actually doing.

None of the major tech outlets after the Investor meeting reported on rumors of Intel scaling back 10nm. All I've seen are articles reporting on their latest official timetables and articles about ramping up production and accelerating product launch intervals.

Again this an Intel thread discussing their 10nm and 7nm future products. Linking news articles and the most recent timetables released from Intel officially (not a leaked, possible fake or rumored timetable) is more productive. Rumors are cool and all but official timetables that came out a week ago are the most accurate information we have. Intel will have competition, no doubt, but being an Intel thread I don't think we are here to discuss what AMD will be bringing to the table. Again there is a dedicated Ryzen and EPYC thread if you would love to discuss AMD as posting it here detracts from the point of the thread.

 

I've already stated that keeping chipsets M/B's for longer than necessary stagnates actual progress. I've seen nothing to dispel this myth that M/B longevity is a positive aspect of a platform when the end goal is increased productivity.

When I too was once naive about purchasing a new platform vs. upgrading one, I too made those mistakes at a high cost in the end. Tearing down a working system for something marketed as 'better' is a fool's errand when actual work is waiting to be done.

Re-read my posts, they are not ambiguous. I've covered all the possible cases for which my points are valid. Walls of text and repeatedly going back to if/when 'x' happens for company 'y' and other currently unsubstantiated rumors and musings is great if you want to keep convincing yourself of something that isn't currently true, but may be, one day.

Worse, you keep repeating what I've already said, except for trying to put an AMD spin on it. I've already said it!

 

How are you ignoring that what they said DOES NOT contradict what I said?

Only doing low core count low power variants this year, with allegedly nearly 6-month inventory piling? 3 months is pretty normal, 6 months isn't, which speaks to yield issues on the node.

Then, you have forward looking statement protection on their estimate to market regarding the server segment. Cannot be sued for that unless it is misleading because of omissions relating to that statement, which you then have to get to the embedded inferences. I will not bore you with the details and nuances of securities law (you likely do not care).

And if we trusted Intel on release dates, we would have had cannonlake back in Q4 2016, I think it was (may have been 2017, but I think it was originally Q4 2016, then pushed to 2017, then 2018, which they had a token 10nm chip in laptops by Lenovo which never saw volume production).

How are you not looking at those aspects for their statements?

 

Yeah, that horse has been beaten to death, yet Intel is still here in the now.

Time to move on from 2016, methinks.

 

That is crap. You do NOT need artificially gimped progress each year. Such as artificially limiting Optane use.

Now, with that said, we are on the cusp of switching multiple tech generations that says sticking with the old board is not as good. For example, consumers get PCIe 4.0 from AMD this year (Intel is planned for next year). You have DDR5 coming around 2021. You have PCIe 5.0 rumored for AMD servers next year (no word on Intel that I have seen yet), but that might not come to consumers until 2021 or 2022, if at all with the cost of implementation and consumer uses not using all the bandwidth in PCIe 3 at the moment, which it will take awhile to saturate PCIe 4.

But, there are many times where the "features" on the new boards DO NOT benefit the user in any significant way.

 

An upgrade to a new part within a platform is nice to have if I can have it along with other benefits (more capacity, in RAM or storage...).

A platform upgrade offers more.

See:
https://www.anandtech.com/show/1404...el-core-i7-2600k-testing-sandy-bridge-in-2019

The article above tries to show that but fails in my eyes. Why? Because what was actually available in 2011 isn't what they're testing with today.

Try running a 2011 platform today and see if your performance and productivity doesn't tank.

You may not be cognizant of the advances and time savings a new platform makes, over and above any CPU improvements. Your loss.

Am I stating that you need a new platform each time? Of course not. Stop stating the obvious fail cases. What I am saying is that when it is a viable option for all the reasons we both know it can be one, it is usually worth more than just the sum of the parts that you make it out to be.

To put this in car terms so that others reading along can follow easier; yeah, put a Porche engine into a GM 'sports' car, but the overall experience will still be junk. Just that the junk will go faster.

 

And was Intel forced to push it out due AMD's hard competition? Nope. AMD haven't other choices. If not they would been pushed backwards.

 

The article doesn't try to show that. It tries to compare CPU performance increases over the past 8 years. Imagine that, going to a faster memory standard, which was not isolated at any point in that article, can be shown to be a worthy upgrade. Or changes in connectivity, which also was NOT shown off in that article, meaning the PCIe lanes to connect faster storage, etc.

That article's sole purpose is to show that the CPUs have improved to a point that a person should consider upgrading their rig, with caveats of not being able to afford an upgrade or if playing on 4K for gaming. That is quite clear at the end of that article.

So that article DOES NOT stand for the proposition you think. Also, it does NOT suggest Intel should be the only considered chip for an upgrade. That is thinking from nearly a decade ago.

 

Unlike yourself, I don't take the article at face value. I can and do use the data presented to draw further conclusions when my experience is included too.

I don't know what you think that article stands for, to me. But you are obviously misreading it all.

 

When you do not share your experiential data to transform the information contained in the article, then all we have is the article, and as such the article, by itself, does not stand for your proposition.

See, there is no analysis of the change from DDR3 to DDR4 there. There is no analysis on how additional PCIe lanes direct to the CPU and NVMe compares to being bottlenecked by the PCH controlling the storage and sharing a x4 connection to the CPU, etc. I've read articles on all of that. THOSE articles stand for the proposition that platform changes can bring benefits.

I've also read how the Z170 was artificially not supporting the Optane cache drives for OS use, trying to get consumers to purchase Z270 boards instead with the release of the 7700K. As it turns out, Optane didn't give much value for my uses, or most people's uses, anyways.

So, please, put it into context for me. Explain it so that I can see how it stands for that proposition. Take snips from the article, then write how it supports your thesis.

Oh, right, that is what I do in those "walls of text" you keep complaining about. Forgive me, I was trained in using the "make a statement, explain the support for the statement, draw conclusion", and "Issue, rule, analysis, conclusion," and "conclusion, rule, explanation of rule, analysis, conclusion" forms of writing. I try to make it so that anyone that reads what I wrote can understand and pick apart my underlying assumptions and analyses. You don't even address the points I make half the time or more, rather you are just dismissive of them, or try to downplay them.

So please, take a page out of my book and explain to me how it stands for the proposition.