https://www.extremetech.com/computing/296453-welcome-to-the-second-golden-age...

Here is a different viewpoint on AMD's second golden age --- it points out that Intel couldn't completely cope with last year's AMD innovations and Intel has made no real progress in the year since the last match up.   Now AMD has moved the high jump bar up a notch and this implies that AMD's rate of improvement is just going to continue ---- driving Intel into the ground like a tent stake.







Part of AMD’s advantage currently is that it can hit multiple Intel weaknesses simultaneously.  Need lots of PCIe lanes?  AMD is better.  Want PCIe 4.0?  AMD is better.  If your workloads scale optimally with more cores, no one is selling more cores per socket than AMD.

The move to 7nm has given AMD a large advantage in power consumption as well, particularly when you consider server retirements.   STH reports single-threaded power consumption on a Xeon Platinum 8180 at ~430W (wall power), compared to ~340W of wall power for the AMD Epyc 7742 system.

What they also note, however, is that the high core count on AMD’s newest CPUs will allow them to retire between 6-8 sockets worth of 2017 Intel Xeons (60-80 cores) in order to consolidate the workloads into a single AMD Epyc system. The power savings from retiring 3-4 dual-socket servers is much much much larger than the ~90W difference between the two CPUs.   This is enough energy savings payback by itself to justify buying new AMD systems going forward.

So, Intel is forced into Playing AMD’s Game Now

Intel can cut its prices to respond to AMD in the short-term. Long-term, it’s going to have to challenge AMD directly. That’s going to mean delivering more cores at lower prices, with higher amounts of reasonably priced normal memory supported per socket.

Intel's Cooper Lake, which is built on 14nm and includes additional support for new AI-focused AVX-512 instructions, will arrive in the first half of next year. That chip will help Intel focus in on some of the markets it wants to compete in, but it won’t change the core count differential between the two companies. Similarly, Intel may have trouble putting a $3000 – $7000 premium on support for 2TB – 4.5TB of RAM given that AMD is willing to support up to 4TB of memory on every CPU socket.

Intel still wants vendors to invest in building support for its Optane DC Persistent Memory, but it isn’t clear how many are doing so. Nothing that Intel offers now operates at full rated speed unless backed by COPIOUS AMOUNTS of Optane memory in several forms and locations on the Intel motherboard.  

Much of the recent speed increases from Intel have actually been all the large amounts of Optane clustered all around the CPUs.  

The current rock-bottom prices for both NAND and DRAM have made it much harder for Optane/Intel to compete in-market.   AMD's products all use normal memory types and AMD still functions AS FAST OR FASTER than Intel's very very expensive Optane mix while doing so.

Two years ago Intel chose to bet their farm on using lots & lots & lots of Optane Memory instead of doing CPU lithography advances and basic design improvements and this decision has come up a crapper in light of current realities ......

Please note that even Intel's old memory business partner Micron has now given up on Optane as a bad bet as it had never proved out to be the world beating advancement that Intel's brown vapor PR had always touted it to be.






What agonies does next year hold for the long suffering "incompetents" at Intel and their Wintel buddy MS ......




Intel is now throwing a ton of shade at MS for up to 20 historically faulty drivers that they have incorporated into the Win 10 worldview, drivers that once they are mitigated are really really going to hurt Intel's chips performance levels a whole lot more.

Wintel as a pair is going to pay a mighty price for a dozens of years of laziness, back when they first knew about these issues but instead chose to sweep them under the rug rather than fix these underlying issues.    Yep, more class action lawsuits are being posted and people are signing up now to get in line for their damages payouts.

By doing this, Intel hopes that the class action lawsuit bills fall mostly on MS and not on them quite so much ......

So, as of today realize that "Microsoft running on Intel" isn't going to be nearly as fast nor secure nor as controversy free as running Linux on Ryzen will be going on out into the future.

Also realize that this same "avoid the issues, avoid the penalties" mentality still exists at Wintel as Intel in particular wants to continue to put out hardware that completely lacks any hardware fixes these 8 old original predictive mitigations and also has no fixes for these 20+ new driver based mitigations that still need fixing.  

Really, the Wintel pair simply cannot stand becoming any "functionally slower" due to more mitigations or they WILL LOSE YET MORE FACE inside their industry.

Tom's Hardware as one of the better test houses has begun listing all their test results as "with mitigations" and immediately listing them again "without mitigations" so as to calmly call out Intel on the sheer size of their self-deceiving processor rating practices.

https://www.digitaltrends.com/computing/amd-zen-3-news-rumors-release-date-ar...

The Upcoming AMD Zen 3: Everything we know so far



AMD announced in early August 2019 that the design phase for Zen 3 was now complete and that its hardware developers were moving on to the design challenges faced by Zen 4.

With Zen 3's designs completed, AMD can now move on to testing and prototyping Zen 3 chips. How successful that process is will determine whether we’ll see Zen 3-based CPUs show up in early summer 2020, or more towards the holiday season.  Either way, we expect Zen 3 to show up in 2020 as part of a full, commercial line up of CPUs.

The full range of processors has a trio of codenames.  Milan will be for high-end servers, a replacement for the Epyc Rome CPUs.  Vermeer will be for high-end desktops and enthusiasts, possibly Threadripper CPUs, while Renoir will be for mainstream desktop CPUs and mobile APUs.

These mainstream AMD chipsets will face off with Intel’s Tiger Lake CPUs on the desktop.

PERFORMANCE
Zen 2 CPUs were a major advancement over the first-generation Zen and enhanced Zen+ designs. It offered double-digit instructions per clock improvements, boosted clock speeds, and additional cores for improved multithreading capabilities. Since Zen 3 will be more of an evolution than a revolution, we don’t expect the same major enhancements, but there should be some notable improvements all the same.

Due to the new manufacturing methods used in the creation of Zen 3, we expect to see improvements in clock speed and energy efficiency.  That could help AMD’s Ryzen processors crest 5GHz for the first time right out of the box.  There’s also the possibility of new instruction sets, which could enhance the entire range’s performance in particular workloads.  Full 7nm EUV+ at TSMC also brings along abilities to triple layer burn features into standard silicone wafers, something AMD has been quick to utilize this past year to add extra cache memory to a few current Ryzen Zen 2 gaming chipsets.

ARCHITECTURE
Zen 3 is based on a 7nm+ process node. While technically the distance between transistors shouldn’t change much, the new manufacturing technique known as extreme ultraviolet lithography (EUV) enables a thinner burn track and a far denser packing of transistors (up to 20 percent) than Zen 2’s fuzzier deep ultraviolet (DUV) methodology. That extra density should improve energy efficiency, which AMD can in turn leverage to enhance clock speed or other facets of the processors’ performance.

In an interview with TechPowerUp in April 2019, AMD’s CTO Mark Papermaster suggested that Zen 3 would prioritize energy efficiency, but that we would still see some modest instructions per clock improvements over Zen 2 as well. That should lead to greater single threaded and multithreaded performance. When combined with additional clock speed, these CPUs should be even better in gaming than the already excellent Ryzen 3900X.

Remember please, that Intel rates everything from stone dead cold (pre-boot cold temps) and that in less than a minute the Intel "rated speeds" simply no longer apply to anything that exists in the real world.    

AMD however rates their stuff fully warmed up and under full load using the AMD supplied stock fan and fin cooler system, so AMD's functional speed ratings actually may well be faster than Intel's never declared fully heated up functional speed ratings ----- but this situation reminds us that Intel ALWAYS intends to mislead their customer base and Intel always wants to "win the PR war" any way they can.

So far, Tom's Hardware is the only one offering fully heated up speed rating comparisons and they are showing Ryzen 2 chipsets have greater fully warmed running speeds than Intel's finest --- even when Intel is running water coolers (in a few cases anyway).

Yes, Intel now recommends and requires WATER COOLING and lots of Optane memory in some of its "stock configurations".          

lower-clock-speeds-on-intels-10th-gen-ice-lake-cpus-arent-a-disaster.html

https://www.pcworld.com/article/3410589/is-ice-lake-better-than-ryzen-9.html




Intel is the overly large on on top, AMD is the squarish one below that.

OK, we are talking new BS brown vapor set up testing by Intel again, specifically the new Ice Lake 10nm as the case in question.  

You were just faithfully promised a 18% speed/performance and battery life upper by INTEL.    The actual first test samples (as cherry picked and installed and tweeked inside INTEL owned laptops did not give these +18% results, but instead tested same - same to the old 14+++nm Intel laptop chipsets that are commonly out there right now with disappointing (no improvements seen, actually really not as good battery life results were seen).

Intel just got slammed hard for the crappy battery life results and flat out questioned as to why anybody would build or buy a 10nm Intel Ice Lake product because of these somewhat poor test results.

Indeed, there suddenly are no plans by the big builders to carry forward with the entire Ice Lake line of desktop products, as they are SLOWER than the old INTEL 14+++nm desktop chipsets they would compete against and Ice Lake poses no competition to the latest 7nm all-in-one Ryzen processors from AMD much less the Ryzen lineup we have today.

Because of these vocal and misleading INTEL challenges, AMD has brought forward their next APU series of processors early and PLEASE watch out what the Intel to AMD "comparisons" you are being shown are really comparing what to what, PLEASE --- there are too many INTEL benchmarking games being played here to even list all of them.    

The easiest way for you to catch the BS is to use this chart and spot how far back in time (how far out of sync) the INTEL match ups you are being shown really are.   INTEL does not want to compare themselves to any current modern state of the art AMD processor of any current generation it seems like ..... nor do they want to compare 10nm ICE LAKE against the most current crop of modern 14+++ nm INTEL processors either.    (except for the new INTEL graphics included that is, INTEL can claim an improvement to their old stuff there)

TAKE NOTHING AT FACE VALUE, there is lots & lots of lying vaporware and BS are being applied lately by INTEL because, quite frankly they've really got bupkiss to show you right now, but are instead trying to sell some raw freshly produced INTEL PR on some some vague future brown vapor bupkiss, a set of misleading intentional lies  that is going to cost them a bundle in class action court rooms soon enough ......



Spot where you really are on this graph, KNOW ABOUT the two items you are comparing to each other.

Lastly, always remember that INTEL rates processor speed from a stone cold starting point (rating off the one very best core in the set) and that INTEL does not apply security mitigations for all their various illnesses to test sets (because it could cost them 15-25% of their already marginal performance).


PART OF THIS MESS IS THE INTENTIONAL CONFUSION BETWEEN THE LOW END ICE LAKE UNIT AND THE "i7" UPPER END ICE LAKE UNIT.   I have to confess to still  being confused by the two of them, and I suspect this was intentional on Intel's part.   Intel's lower end Ice Lake unit appears to be a stone lose-lose situation no matter what you compare it to, while the upper end Ice Lake unit should be able so show break even to a bit better than break even due to the improvements in the new generation of Intel graphics.

https://www.tomshardware.com/news/amd-cpu-sales-intel-ryzen-german-market-ret...

AMD Comes Roaring Back: Analyzing Five Years of CPU Sales at Mindfactory



The graph above is with AMD consistently sandbagging unmercifully on all their newest stuff as Intel isn't improving any at all (save for a dubious very few 10nm laptops).  

AMD is being very careful to keep the range of actual performance improvement and pricing differential at about the same distance apart compared to Intel --- this fuels their removal of Intel market share at a rate that AMD can support.  

AMD is also being careful to post rapid fixes for anything Intel starts to use as a FUD talking point.

In the past month Intel and its paid talking heads have pushed a lot of FUD nonsense that said:

AMD processors runs hot, hotter than Intel ----- this was debunked as simply not true unless you were running ALL of the AMD cores at full tilt during a benchmark test (a thing that really does not happen in the real world).   Intel's boys carelessly lost track of the fact there are TWICE AS MANY OR MORE AMD CORES running at full tilt during these tests and that the fact the entire heat load of all of the extra AMD cores did not exceed Intel's normal heat rise by very much at all.   Each AMD core runs a lot cooler than an Intel core can run.   But yes, there are a lot more AMD cores to make heat during a full on benchmark test.

Intel made a lot of noise several weeks ago about a few low performing AMD cores that were seen in testing, and a lot more noise about AMD having "throttled their cores 25% more due to suspected (predicted) low service life".   AMD responded that their service life is fine, and yes out of the 2x more than Intel cores in an AMD processor some cores do perform slower, but all are within the now understood published acceptance ranges.   AMD also points out that Intel does the exact same thing, but Intel allows some cores to vary much more than AMD allows.   AMD's response was prompt and factual, and these facts were verified by Tom's Hardware.

Also, within 1 week AMD's vendors were pushing out BIOS updates for their motherboard products that caused these slow cores to speed up some and this same BIOS update also changed the scheduling portion such that the slower AMD cores got less heavy traffic put against them by the scheduler.   Intel's current talking point FUD issues basically disappeared at that point in time.

So what is really happening?

Intel simply is losing a lot of ground to AMD and Intel hates that fact a lot.  

Fear, Uncertainty and Dismay has always been an Intel marketing tool and that aspect has not changed at all.   This week's Intel FUD is saying that AMD is going to run out of manufacturing capacity and the "current shortages" of AMD chipsets are proof of this.

Look at the graph up at the top.   AMD is selling so many more chipsets (and Intel is selling so few) that yes, AMD is being careful about the timing of introducing new advanced products to be sure that they have enough 7nm chiplets flowing to build yet another hit product.  

In reality, Apple kicking into their full new 7nm product production cycle has stressed TSMC's ability to make everybody happy all at the same time at 7nm++ right now.   Apart from AMD and Apple, TSMC's lead times have just tripled for all the little guys who are just now placing  their very first 7nm orders.

TSMC's solution is to rapidly ramp up 5 more brand new ASML 5nm lines.  TSMC wisely spends their money on next year's best and brightest, not on old stuff with less than a year to go on it's primo lifespan.  

TSMC has released all of the build tools for 5nm to their customers now and is getting 5nm 14 layer deep burn orders already placed on them from AMD, Apple and others.   Full 5nm Production at TSMC begins soon, like March of next year type soon (or sooner if your name is Apple).

So, TSMC isn't buying any additional 7nm lines, but is spending all their money staying ahead of the coming 5nm wave.   When Apple and AMD shift over to 5nm, then some of the current pressure will come off of the existing 7nm production lines.   Remember please, that the new 5nm lines can do a 7nm product in a pinch, just that all of the rest of the 14 layer stuff won't be used when doing 7nm but if the demand is there and the new 5nm line is idle ....... well,  heck, go for it.

Bean pickers are funny, new production lines need to be booked solid so as to generate a revenue flow to pay for the new line out of its own cash flow (while you are paying for it actively, anyway).   TSMC likes to be booked solid with a waiting list so they KNOW they get their full utilization of the money they just spend on new production equipment.


===================================================


Good news, Intel has finally placed their first real orders for some real 7nm++ EUV production equipment --- this puts them 1.5 years minimum away from making a real 7nm full production line of chipsets ...... and Intel is 1.5 years behind all the hockey stick boys who had ordered their stuff late this spring and are most of the way through their wait period.

Intel is calling their old home grown 7nm lines "10nm lines" now, so expect Intel to make a whole lot of 10nm stuff before they make any real quantities of "new" Intel 7nm++ fresh off of their really new new really really 7nm EUV lines, the ones that Intel just placed orders on ......

            .....  i.e. by the time everybody else will be moving from 5nm down to 3nm

https://wccftech.com/intel-guts-cpu-pricing-response-ryzen/

Intel cuts some prices by 44%



HEDT SKU      Cores / Threads      New Price      Old Price*
Intel Core i9-10980XE   18/36      $979      $1979
Intel Core i9-10940X      14/28      $784      $1387
Intel Core i9-10920X      12/24      $689      $1189
Intel Core i9-10900X      10/20      $590      $989

OK, Intel is sorta kinda getting real now ...... they have half the cores of AMD, a quarter to half as many computing threads, and whole warehouses sitting full of older model chipsets that are NOT MOVING ANY AT ALL ...... what is a boy to do  ????

Mind you this newly announced 44% price cut is a partial move on Intel's part, it does not make them price or performance competitive to AMD, but it does help Intel to turn warehouse stocks into ready cash which Intel needs very badly.    

And as a first move at least it is in the right direction.   Intel post price cut pricing is within 10% of AMD's real selling prices.   But you can't buy from Intel at this price, only Intel's historical builder buddies can buy at this 44% off price right now.

Intel has replied to AMD's last announcement, this time in a response that actually could meant something.

https://liliputing.com/2019/10/the-new-fire-hd-10-is-amazons-first-tablet-wit...



Many times I have said, if you want to know the real "state of the art" look to the lowest common denominator, the low end products, to judge if progress is really happening or not.

Right now, the low end stuff is best represented by the AMAZON FIRE line of tablets and the WALMART ONN line of tablets.

In the past six months, these lines of products has gotten 30% stronger processors, better pipelines and throughputs and Amazon has picked up USB C and the ability to use USB C "on the go"  (Walmart is behind on this, but not for long I suspect).

Now saying this, I remember doing a reality check just last year and finding out many of these new Amazon Fire advancements DID NOT WORK AT ALL because Amazon was feuding with Google again and Amazon Fire products didn't load the current Google drivers, etc.

This Amazon/Google feud is still going, Amazon does NOT COME SOFTWARE COMPLETE OR CURRENT, it remains 3-4 years back in time compared to the rest of the Android world.   And this means the Amazon hardware although it may be current doesn't work as expected.

This situation has prompted Walmart to come out with their own line of tablets that do come with some improved hardware and DO SWING A MODERN 9.0 ANDROID COMPLETE WITH ALL DRIVERS.

Walmart's unit is running the current Android 9.0, but all the internals are the older mini USB standard and the processor and the memory are not as good as the current Fire products that are swinging USB C.   But Amazon with the more modern hardware is still being self-crippled by not using the current Android 9.0 software and drivers.

Walmart will be quick to fix this situation, Walmart will probably have new units out in time for Christmas with modern guts swinging Android 10 as Walmart is attempting to outdo Amazon in all things lately, but this one Walmart effort seems to be working out somewhat better than the rest of  their Amazon beaters ......

So, buy what your are gonna buy anyway, just be aware that better stuff is only a month or so away.  Justin will buy his wife a new Amazon Fire 8" when the time comes  (amazing ain't it -- just how regular this "spontaneous Fire death" occurs, huh?)

My wife is still using her Fire 8" tablet regularly, but her battery acting sorta weak so I am watching this newest Fire 10" as a potential upgrade Christmas present for her for next year.  

Her Fire 8" is jailbroken and it gets Android 9 updates directly from Google, so I doubt she will have this "spontaneous Fire Death" stuff happening to her tablet -- a simple replacement by me due to "poor battery life" is a much more likely occurrence.

https://www.notebookcheck.net/The-Ultimate-Core-i7-1065G7-and-Ryzen-7-3700U-2...

There is a BIG whopping graph on this one, I can't put it in the list text box at all and there is no net reference to give you as it is a text based graph being generated off a database.   You just got to click on the red address link to go there and take a look at it.

https://www.notebookcheck.net/The-Ultimate-Core-i7-1065G7-and-Ryzen-7-3700U-2...


The first Ice Lake laptops are finally on store shelves just as Intel had promised back in CES 2019. Dell, HP, and even Razer have all begun shipping their respective premier models each equipped with the 15 W or 25 W 10th gen Core i7-1065G7 designed to replace the last generation Whiskey Lake-U Core i7-8565U and compete directly with the mobile AMD Ryzen U series.

I have said repeatedly that Intel lies a lot, claiming loudly to have made a lot of progress when they actually have NOTHING REAL that is good to show you.

Notebook Check has left the ranks of paid Intel puppets now, they have posted an honest review of the two new "10nm" Intel all in one Laptop Processors, the 15 watt and the 25 watt versions.

What do you need to look carefully at the big graph to see ......   you look to see how Intel only ranks their processors from a cold dead start and the Intel BIOS built into the chipset runs them full out until thermal limits are hit then the BIOS THROTTLES ITSELF very strongly, down to a MUCH SLOWER rate that the chipset can actually support for more than a minute.

The AMD horizontal blue circles (second line from the top) are what you should really focus upon.   Also ignore the fact that Intel won't compare itself to current just released AMD processors, but insists on comparing back at least one generation ---- even using these dirty trick AMD's old stuff still outperforms Intel once the magic minute has passed and everyone is running at steady state.

So, Intel has built a gimmick into their heat controlling BIOS to allow them to show a false performance advantage over their competitors that is pretty much meaningless in the real world that you actually live in once that magic minute is past.

What is now true (and is making Intel quail in their boots at the moment) is that the new Gen 3 AMD stuff coming out now (that was just announced yesterday) picks up enough throughput advantage to make up even a bigger steady state lead, and on top of that it picks up a solid 20% speed upper that makes the Intel cheat "magic minute" hump at the very first pretty much flush with AMD even in the first few seconds of heat build up.

I find it refreshing to have computing magazines are showing the public all of this stuff openly instead of just hiding it like they used to.

This goes a long way towards explaining how the Intel press says Intel is "so much better than AMD" yet the impartial folks like Tom's Hardware and ArsTechnica consistently say something quite different.


===================================================


In line with this Intel 10nm "false performance advantage" folding up under impartial scrutiny, there are now rumors flying around that Intel's main builder/customers don't want anything to do with this very very expensive yet half arsed all over again Intel 10nm.   14nm gives the same performance and costs a LOT less .......

Now Intel has supposedly had to drop all their rapid expansion plans to go across the board with their version of 10nm because it simply doesn't offer any real advantage over the current Intel 14nm and Intel 10nm certainly is no competition to anything built by AMD and accordingly at this stage of things none of the Intel builders want any part of it.   This article says it will be year 2022 before you see anything that is functionally better than the Intel 14nm stuff out of Intel.

https://www.tweaktown.com/news/68127/intel-cancels-10nm-desktop-14nm-hold-unt...

Loud screams coming from Intel today tend to say there is a little fire burning over there underneath that dense Intel smoke screen ..... and Intel's hand is being held in that fire by their own builders.

https://www.tomshardware.com/news/intel-yes-there-will-be-10nm-desktop-cpus

...... and yes, Intel has admitted to Tom's that they have both low yield and low performance issues yet to overcome with their 10nm products.  Tom's Hardware is quoting inside sources that say that widespread Intel 10nm will likely be delayed until sometimes in 2021 which smells like 10nm will never arrive as Intel will have real 7nm production by that time off their new 7nm equipment that they just ordered and that will be delivered and hooked up and up and running by 2021.

Me, I say Intel lies a lot, and spending your time paying attention to what Intel says right now is really kinda futile ......   just watch what they really do.

AMD is rolling into the ......

F
U
T
U
R
E

...... while Intel stays stuck 2-3 generations in the past.    

In the last week Intel has come up a crapper with their brown vapor 10nm across the board plans again, and Intel has run out their 7nm plans into 2022 for their 7nm implementations.   This puts Intel firmly 2-3 multiple generations stuck in the past again while phones and Android march quickly into the future, with AMD going along with them instead of with Intel.

Ask Global Foundry what happens when you follow Intel's lead instead of Android's -- oh, that's right, Global Foundry only took a big 2 years of stagnation to become totally passe' and to drop away from all relevance.   Now you can buy parts of Global if you want them, the Arab owners have it all for sale now.

AMD's next Gen 3 design for 2020 is in the can now at a 20% throughput and 10% speed improvement and production has already begun on it.   Intel cannot even start to compete with this even using their largest mainframe chipsets run up in over-voltage speed using all the overclocking tricks the gamer - hobby boys have built up over the years.  

Watch the 16 core Ryzen announcements for the first fruits of AMD Gen 3 --- the new test results are coming in as we speak.

AMD Gen 4 is 80% design complete at first early production level of 5nm (15 burn levels) using 5 new production lines that TSMC has already contracted with ASML to purchase.   First lines have been contracted and are being built now.   This rolls into place in 2021.   Apple is footing the bill for the first new TSMC 5nm lithography lines as they are wont to do as they WANT and NEED that critical 1-2 years worth of tech edge over everybody else.

AMD Gen 5 is also under early design using the very newest ASML scanners that are being designed for Apple and TSMC which will be 5nm with 24 burn levels instead of just 15 burn levels.  This is planned for 2022-2023 and AMD has their two leapfrogging design groups functionally keeping the AMD improvement pace up there at an Intel crushing level.

There are prototype 3nm ASML high energy scanners being designed and built, Apple wants them ready for 3-4 years from now and we all know that Apple generally gets what they want since they are paying for it.  AMD will have a design team working on that as soon as the specs become known.