Ryzen vs i7 (Mainstream); Threadripper vs i9 (HEDT); X299 vs X399/TRX40; Xeon vs Epyc

As some have mentioned, last year's coffee lake was really a Kaby refresh, while this year is the full coffee release. So there should be some performance gains, but how much over last year's 8700K, who knows...

 

TechPowerUp Processor Survey Results: The Ryzen Effect is Real
by btarunr Wednesday, July 4th 2018 03:23
https://www.techpowerup.com/245735/techpowerup-processor-survey-results-the-ryzen-effect-is-real

"Late May 2018, TechPowerUp started a front-page poll asking people which processor they use. 37 days and 16,140 responses later, we have a general idea of where the desktop processor market stands among our readers (predominantly PC gamers and enthusiasts). The top-two responses to our survey were 4th generation Core "Haswell," followed by the preceding two generations ("Ivy Bridge" and "Sandy Bridge"). This speaks volumes as to the hole Intel dug itself into, due to lack of competition from AMD. Processors that are 4-7 years old still run today's gaming PCs, and don't bottleneck today's games, as long as graphics cards keep getting faster (where there has been relatively more competition than the CPU market).

Despite being newer, fewer respondents use 6th generation "Skylake" and 7th generation "Kaby Lake" processors than older generations, because those on something like 4th generation "Haswell" or even "Ivy Bridge," don't see the value in upgrading. But then something changed in 2017 - AMD became competitive again, and forced an increase in CPU core counts across the segment. AMD's Ryzen processor family, including both its 1st and 2nd generations, are better received in the market than Intel's competing 8th generation "Coffee Lake" and 7th generation "Kaby Lake." The data stands to validate the "Ryzen effect," the idea that the introduction of Ryzen disrupted Intel's near-monopoly, increased core-counts, and brought innovation back to the segment.

More of our readers use AMD Ryzen processors than Intel Core "Coffee Lake" and "Kaby Lake." So in the period following Intel's launch of 7th generation "Kaby Lake" (slightly before the launch of Ryzen), more AMD processors were installed among our readers. This of course doesn't mean that there are more AMD users, since we're not counting pre-Ryzen Intel generations such as "Skylake" and "Haswell." This seems to suggest that the "Ryzen effect" is not a myth.

In the time since 2nd generation "Sandy Bridge" (circa 2012), very little innovation has been there from Intel for PC gamers. The mainstream-desktop segment has had to content with no more than 4 cores, and there's been very little IPC increments between generations to warrant upgrades. The result is that there are plenty of people with >4 year old processors, which are fast enough for today's gaming. The data also shows that in a shorter span of time, AMD sold more Ryzen chips.

Of course there are limitations to our survey. The data is sourced from a user poll among our readers, in contrast to the Steam Hardware Survey, which gets its data by probing the hardware of a machine. As we mentioned earlier, our readers are composed of PC gamers and enthusiasts, and hence our data isn't in line with the general market (that includes other use-cases).
"
Comments

Wccftech has a few articles tracking AMD vs Intel CPU sales from a German retailer, which show a dip in AMD sales just before Ryzen 2 arrived, with the highest recorded AMD percentage of 53% in November 2017, the lowest after than was 39% in Feb 2018, 40% in March.

Ryzen 2 sales started in April 46%, May 49%, June 45% down likely due to 8086k sales, hopefully AMD share will bounce up again before Intel 9th gen release and again bounce up for TR2 sales, and then Intel has nothing coming - if 10nm is truly dead, so it should be a steady increase for AMD share throughout 2018-2019.

https://wccftech.com/intel-coffee-lake-amd-ryzen-cpu-market-share-june-2018/
https://wccftech.com/jpr-market-share-amd-intel-cryptocurrency-nvidia/
https://wccftech.com/amd-cpu-market-share-jumps-43-in-april-alone-steam-survey-shows/
https://wccftech.com/amd-intel-cpu-market-share-ryzen-2000-up-intel-8th-gen-dominates/

 

The 2990x, just like the 1950x and 7980xe, will count overall for very few sales. Intel is far from gone or being a goner. Now once 7nm gets here the trend may reverse counting on the performance. Intel may no longer have the IPC or Gaming champ and this could truly tell in the end. TR2 may knock Intel from the HEDT performance lead but 12nm just won't get them off the pedestal just yet.

 

As a halo effect product the TR2 release will bring more new owners to AMD Ryzen, even if they don't buy the TR2 level solution right off first time. Any positive news to help grow the AMD line will help potential owners to feel comfortable, realizing it's time to jump off Intel's sinking ship and start on their own AMD transition.

Intel has been screwing up consistently, and widely, and without remourse for years now, and I don't see how they can pull themselves out of their technology tail-spin. I really don't.

Intel will continue to release hardware, some may even compete and sell for a while yet, but I think Intel will continue to fall by the wayside, unless something radical is done inside Intel. Intel really just have no clue what to do, and to stick with it to a successful completion.

Intel is Building the 5G Future Now
Intel is making 5G real now—bringing together wireless, computing, and the cloud for a seamlessly connected, powerfully smart future. Across the globe, we’re bringing 5G out of the lab and into the field, collaborating with partners across industries to develop and test 5G technologies and prototypes, while breaking through technological barriers to achieve the record-setting “firsts” that will accelerate commercial deployment. At the same time, we’re innovating rapidly to build the 5G silicon solutions today that will enable the next generation of phones, PC’s, cars, and more.
https://www.intel.com/content/www/us/en/wireless-network/5g-technology-overview.html

Intel Kills Off 5G Modem Project After Losing 2020 iPhone Contract
2020 iPhone could run on MediaTek modems instead
Jul 5, 2018 09:57 GMT
https://news.softpedia.com/news/int...fter-losing-2020-iphone-contract-521845.shtml

Anyone want to compile a list of the projects Intel started and then canceled, and a list of the failures to innovate over the last 10 years?

Does anyone out there really believe Intel is going to pull out of this series of disasters successfully, any time soon?

Apple Inc. Dodged a Serious Bullet
Avoiding this unreliable chip manufacturer was a good bet on Apple's part.
Ashraf Eassa, ( TMFChipFool), Sep 6, 2017 at 12:45PM
https://www.fool.com/investing/2017/09/06/apple-inc-dodged-a-serious-bullet.aspx

"...
Betting on Intel 10nm would've meant disaster
Had Apple bet on Intel's 10nm technology for the upcoming A11 Fusion chip -- which Intel continues to publicly claim is about a generation ahead of competing 10nm technologies -- it would be in a world of hurt right now. TSMC has been cranking out chips for Apple -- as well as others, like MediaTek and Huawei -- using its 10nm technology for months now. TSMC will need to deliver tens of millions of units to help Apple meet what could be insatiable demand for the latter's upcoming new iPhone models.

Intel, on the other hand, has seen its schedule for 10nm production continue to slip. In late 2013, Intel told investors that its 10nm technology would be production ready by the end of 2015. After many delays, Intel is now hoping to get this technology into volume production sometime in the second half of 2017, and doesn't expect volume shipments of its lead product -- known as Cannon Lake -- to occur until the first half of 2018...".

The one thing Intel is supposed to be good at, Intel's Core Competency - which spent $100B's over the years to maintain successful continuity, and what a surprise, Intel still hasn't delivered Cannon Lake 10nm in the "first half of 2018".

And, Intel has failed at everything else it has tried outside of it's Core Competency - and if Intel don't have that moving forward, what does Intel have of value?

 

News Corner | AMD Destroys Nvidia in BFV? Intel Details 9000-Series CPUs
Hardware Unboxed
Published on Jul 6, 2018
02:55 - AMD Destroys Nvidia in Battlefield V Closed Alpha?
06:33 - Asrock Introduces RX Vega Graphics Cards
08:10 - First Ryzen 3 2300X Benchmarks?
09:22 - AMD Inching Closer to H-Series Ryzen CPU Launch?

Spoiler: Show index

Support us on Patreon https://www.patreon.com/hardwareunboxed

News Topics:
00:24 - Intel Reveals 9000-Series CPU Names and Specs
02:55 - AMD Destroys Nvidia in Battlefield V Closed Alpha?
05:02 - Acer BM270 Monitor Packs 4K, HDR and FALD
05:56 - Philips Momentum 43-inch 4K HDR Display Now Available
06:33 - Asrock Introduces RX Vega Graphics Cards
06:54 - Cooler Master Launches MasterCase H500
07:28 - Sony Uploads Entire Movie to YouTube Instead of Trailer
08:10 - First Ryzen 3 2300X Benchmarks?
08:56 - Nvidia GTX 1180 Appears at Vietnamese Retailer?
09:22 - AMD Inching Closer to H-Series Ryzen CPU Launch?

Sources:
https://videocardz.com/76720/intel-re...
https://www.pcgamesn.com/amd-nvidia-b...
https://forums.battlefield.com/en-us/...
https://www.sweclockers.com/test/2594...
https://techreport.com/news/33871/ace...
https://www.anandtech.com/show/13033/...
https://www.techpowerup.com/245700/as...
https://www.techpowerup.com/245701/co...
https://torrentfreak.com/sony-blunder...
https://videocardz.com/76698/first-be...
https://www.techpowerup.com/245657/nv...
https://www.techpowerup.com/245713/am...

 

So, Der8auer explored overclocking an EPYC 7601 to get an idea of performance on the 2990X. Here is his video:

 

Here is the problem. Epyc does not revert too pushing, with quad channel memory, to two CCX's using dual channel and then the other two CCX's going through the IF. In that case it takes each CCX and divides the memory to single channel to each CCX. This is obvious by at 3.4GHz getting the same score as about a 1950x.

 

It's fun to watch, but in the long run we need the actual part to see what's what.

This simulation only had one useful outcome that I noticed - the motherboard couldn't deliver enough power to simulate the potential of the new 32c / 24c TR2 CPU's.

Even though this is an Epyc motherboard, this power / VRM thermal issue says to me we should wait for new motherboards designed with beefier power delivery and VRM cooling - designed to OC with the new TR2 ThreadRipper 32c / 24c CPU's.

There may be better power efficiency with the new TR2, but again, we won't know until a real TR2 is plugged into a new beefier motherboard designed to get the most from TR2. And, for comparison TR2's plugged in to previous generation motherboard's to see what if any power limitations are found.

 

Well, I did notice that, and seeing as the numbers of being in 8 channel more closely meet what has been seen with the leaks, it is something to watch for and get more info on.

This was a six phase. I do not know if it was a 50 or 60A mosfet or not. Either way, it suggests that capping out would happen not too far above what he got considering you would only have two extra phases. But, this is why I posted it, so we could talk about the findings and known issues that will come up.

Also, notice how his extrapolation roughly gave numbers in line with the TR2 leak.

 

Like I said it was fun to watch, but without the actual part on dual channel memory on a x399 motherboard - either current TR1 power design or TR2 power design, we won't really know if it's close or not.

Those scores you posted from the rumor are even higher than he guestimated, 6100 vs 6250 @ 4.0ghz... and on up to 6400 @ 4.1ghz.

I was perfectly happy to wait until release and owner testing arrived, but der8auer needs to get traffic, so I'm fine with his entertaining video.

 

Well, I'd say both are within reason. You have the memory channel issue, depending on how wired, IF speed, and the speed of the rdimms used.

You can see an uplift of around 15-20% with fast enough ram and tight enough timings on TR1. So, you estimate it gets knocked down to 5200 points if you use 2133, not out of the realm of possibility with his numbers. The higher score comes from 3200MHz+ ram and tight timings, comparatively. That also speeds up the IF. So, 5200-6400 I would think is within the realm of possibility depending on hardware and knowledge of the OCer.

 

FYI, it probably is 6 phase 60 amp but even at 50 amp it can handle 300 amps and at 1.25v that is 375w at least or 450w package. Enough for 3.4 GHz on a short run, apparently. Nice to know Epyc overclocks so well anyway.

 

Salt shakers please;
https://wccftech.com/amd-threadripper-2990x-32-core-benchmarks/
https://segmentnext.com/2018/07/16/ryzen-threadripper-2990x-performance/

 

Interesting poll here to measure interest for 3200MHz ECC memory

 

BTW, if anyone wants to try to beat my scores on sisoft sandra, here you are:
http://ranker.sisoftware.net/show_u...e9d4e5c3ab96a385fdc0f1d7b2d7eadafc8fb282&l=en

Got some pretty good scores on there if anyone wants to take a gander.

 

Some more wood to the fire;
https://www.tweaktown.com/news/6252...990x-simulated-tests-32c-64t-beast/index.html

And then;
https://www.hardocp.com/news/2018/0...ming_threadripper_air_cooler_silver_arrow_tr4

 

Actually all these articles are repeats of the same info derivative of der8auer's video, and it's another reason I felt that it was a waste of time.

All he did was OC an Epyc 32c CPU - he even state's that as the subject of this video - straightforward OC'ing an EPYC CPU, nothing to do with TR2 until he tried to up the view count by trying to tie in with TR2.

So it's still not TR2, but now we need to waste time looking at all the articles and news echoes back from his video as if it's TR2 news, yet there is nothing useful that I can see that came from it all.

 

Agreed but we can only forward the links. Amazing how many will just jump on the bandwagon without further research.

 

Xeon in socket 1151 news from Eurocom:

EUROCOM Tornado F7W
17.3-inch; FHD 120Hz (1920x1080)or UHD 60Hz (3840x2160)
http://www.eurocom.com/ec/specs(452)TornadoF7W

Launch Date: October 9, 2018

Launching Eurocom Tornado F7, heavy-duty GPU and CPU upgradeable,Intel Xeon laptop with an all-in-one design with integrated display, keyboard and built-in UPS/battery pack, ensuring you have dedicated on-the-go capability anytime, anywhere. EUROCOM Tornado F7 supports up to 22TB of storage with four physical drives in RAID 0/1/5 and up to 64GB of ECC memory.

Weight/Dimensions: 4.14kg (9.1lbs); 428x314x24-51mm (17.12x12.56x0.96-2.04-inch)
Operating Systems: Microsoft: Server 2016R2, Windows 10; Linux/Ubuntu
Core Logic: Intel C246 Chipset
Processor: 6-core/12-treads Intel XEON E series of i7 8th generation; LGA1151
GPU Technology: GPU-upgradeable; MXM 3.1; Quadro: P5200, P4200 and P3200
Memory: up to 64GB; DDR4-2400 ECC (with Xeon) or non-ECC (with i7); 4x SODIMM sockets
Storage: up to 22TB of storage with four physical drives (2x M.2 NVMe + 2x 2.5")
Redundancy and Performance: RAID 0/1/5
Network:1GbE; RJ-45 LAN on-board
I/O Ports: 1x Thunderbolt 3.0; 5x USB 3.1; miniDP; HDMI 2.0; Mic in; Headphone S/PDIF out; Line-in; Line out; RJ-45 / LAN
Security: Smartcard Reader, TPM 2.0; Fingerprint; BIOS Disk Encryption; Kengsington Lock; HDD BIOS Password

 

https://www.overclock3d.net/news/cp...a_10-15_ipc_boost_and_up_to_16_cores_on_am4/1

 

AMD AD;
https://segmentnext.com/2018/07/23/jab-intel-xeon-airport-ad-amd-epyc/
https://www.pcgamer.com/amd-disses-intel-xeon-in-ads-scattered-around-san-jose-airport/

 

AMD financial results Q218

Tldr; killing it

https://www.zdnet.com/article/amds-q2-results-illustrate-2018-momentum-in-servers-graphics-pcs/

 

Unrelated ( and, sorry for going off topic ), but nonetheless as interesting... Facebook is not killing it - https://www.usatoday.com/story/tech...-analytica-chill-user-growth-slows/837683002/

 

50 million Facebook profiles harvested for Cambridge Analytica in major data breach
http://forum.notebookreview.com/thr...ajor-data-breach.814675/page-13#post-10770742

 

AMD stock gains after Intel gives updates on 10-nanometer
Published: July 27, 2018 10:20 a.m. ET
https://www.marketwatch.com/story/a...ntel-gives-updates-on-10-nanometer-2018-07-27

"Shares of Advanced Micro Devices Inc. AMD, +2.78% are up 5.8% in Thursday trading after rival Intel Corp. INTC, -8.67% gave updates on its 10-nanometer progress in conjunction with its second-quarter earnings report. The updates were "not positive," according to Bernstein analyst Stacy Rasgon. "It is now crystal clear that 10-nanometer servers will not be here until sometime in 2020 (not a surprise to us, but we believe some investors had held out hope)," he wrote. "This will naturally fuel the AMD narrative, who will be delivering 7nm Rome products in 2019, competing against Intel's Cascade Lake (14nm) on a process that will not be just at parity, but superior." AMD reported its best quarterly profit in seven years on Wednesday afternoon. Its stock is up 86% so far this year, while the S&P 500 SPX, -0.69% has gained 6.1%."

AMD vs. Intel: Balance of chip maker power appears to shift
Expected product delays at Intel play right into AMD’s hands
Published: July 27, 2018 2:10 p.m. ET
https://www.marketwatch.com/story/a...ears-to-shift-2018-07-27?mod=pulse_full_story

"In the first round of chip maker earnings this season, Advanced Micro Devices Inc. emerged a winner over Intel Corp. in what appears to be a sea change in semiconductor manufacturing.

Spoiler: Long...

After both chip makers reported earnings this week, analysts said that AMD is fast gaining competitive parity, or even an edge, as Intel struggles with delays and management uncertainties. That could portend a sea change for a rivalry that has long been seen as a powerful champion against a plucky but typically hapless challenger.

After reporting earnings late Thursday, Intel INTC, -8.70% shares plummeted more than 8% Friday and are on pace for their largest percentage decrease since Jan.15, 2016, when shares fell 9.1%, according to Dow Jones data. Even though results for the quarter topped Wall Street estimates and the company boosted its outlook, data-center revenue growth didn’t reach what analysts had been expecting on average and Intel appeared to delay the rollout of new products.

On the other hand, shares of AMD AMD, +2.81% on Friday added to their highest levels in more than a decade at Intel’s expense, following the company’s best quarterly earnings in seven years. AMD shares were up nearly 2% Friday after jumping 14% Thursday, and hit an intraday high of $19.88, their highest price since January 2007, according to FactSet data.

For the year, shares of Intel are up 3% while shares of AMD are up nearly 83%, compared with a 2.8% advance on the Dow Jones Industrial Average DJIA, -0.32% a 5.3% gain on the S&P 500 index SPX, -0.70% a 12% gain on the tech-heavy Nasdaq Composite Index COMP, -1.49% and a 9.2% advance on the PHLX Semiconductor Index SOX, -0.59%

Bernstein analyst Stacy Rasgon, who recently downgraded Intel, said that Intel’s updates on 10nm products were not positive and that the boost in outlook “implies a Q4 gross margin collapse (and EPS below the Street).”

In a note, Rasgon said “it is now crystal clear that 10nm servers will not be here until sometime in 2020 (not a surprise to us, but we believe some investors had held out hope). This will naturally fuel the AMD narrative, who will be delivering 7nm Rome products in 2019, competing against Intel’s Cascade Lake (14nm) on a process that will not be just at parity, but superior.”

“7nm” and “10nm” — where “nm” means nanometers — refers to how small a chip maker can make the transistors that go on a computer chip, with the general rule being that smaller transistors are faster and more efficient in using power.

Susquehanna Financial analyst Christopher Rolland, who has a positive rating on Intel and a neutral rating on AMD, said Intel’s “moving parts were modestly disappointing” noting that the majority of the chip maker’s earnings beat was from “the Client Computing Group, which we view adversely.” Rolland lowered his price target on Intel to $64 from $58, and said Intel’s “stealth push,” or delay of 10nm products, “now narrows Intel’s time to market versus AMD’s 7nm in PC, and may even widen their competitor’s lead for server.”

Cowen analyst Matthew Ramsay, who has a market perform rating on Intel and an outperform rating on AMD, said Intel faces a wealth of competitive and manufacturing issues and new permanent CEO and this plays to AMD’s advantage.

Ramsay said “we are fast approaching a window (likely a widening one) where AMD will be on equal manufacturing footing as Intel in 2019, and actually have an advantage in the datacenter market, as 7nm EPYC ramps a full year earlier than Intel’s 10nm Xeon CPUs.”

MKM analyst Ruben Roy, who has a buy on Intel and a neutral rating on AMD, said he thinks the positives at Intel outweigh the negatives.

“While bears could continue to point to a slower than anticipated 10mn ramp and uncertainties related to a leadership change, we continue to expect INTC to maintain a significant manufacturing leadership position which, in our view, will continue to drive strong profitability metrics for the company.”

Of AMD, Roy said: “While we continue to appreciate AMD’s execution on its product road map (Rome on track for 2019 launch), we still see the shares as appropriately valued at current levels until more visibility into market share gains can be achieved.”

Of the 33 analysts who cover AMD, 14 raised their price targets after earnings for an average price target of $16.10, compared with an average target of $14.88 before the report, according to FactSet data. No analysts used changed their ratings: 13 have buy or overweight ratings, 14 have hold ratings and six have sell or underweight ratings.

Of the 42 analysts who cover Intel, 11 lowered their price targets after earnings but six raised targets, resulting in an average price target of $57.89, down from the average target of $59.17 before the earnings report. According to FactSet, 24 analysts have buy or overweight ratings on Intel, 14 have hold ratings and four have sell or underweight ratings."

Swan dive: Intel shares dip under interim CEO Bob as 10nm processor woes worry Wall Street
Chips not ready until 2H 2019, Epyc headache looms
http://forum.notebookreview.com/thr...-cannon-lake-cpus.817097/page-2#post-10771417

 

Der8auer testing overclocking the EPYC chip again, except this time he tested quad channel in every configuration and checked it against the 8-channel. The findings are very interesting and actually suggests that the performance, even with that latency, will be VERY close to the server chip with TR2.


@hmscott @TANWare @Deks @Scerate

 

More realistic now that he is using dual channel too the two CCX's. 12nm also tweaked the IF a bit so it probably will be even better as well.

 

Now that we know the Epyc overclocked wanted 500w or so it makes it a bit easier to see what x399 boards can handle with adequate cooling. I run at 1.375 so the chart below I highlighted that. I can tell you I can run P95 at close too a 400w for the VRM without issue but the chip fails at 4.1 GHz set to there I end up loosing workers. So it seems 66% VRM load is feasible with the 8 VRM CPU and 3 VRM SOC of most of our boards.

 

Do the tariff's yet effect Canada? Still less than a 7980XE though.

https://segmentnext.com/2018/07/30/buy-ryzen-threadripper-2990x/

and

https://wccftech.com/amd-ryzen-threadripper-2990x-2970x-2950x-specs-leaked-by-hwbot/

 

"However, the performance isn’t the representative of the real Threadripper 2990X due to memory restraints and the motherboards for AMD EPYC 7601 aren’t designed for overclocking."
https://segmentnext.com/2018/07/30/buy-ryzen-threadripper-2990x/

2x TR1 seems like a fair price for the 2x scale up in core count, but we'll need to wait for real benchmarks and testing results across a wide range of software and hardware implementations before we know for sure.

AMD Ryzen Threadripper 2990X listed for 2400 Canadian Dollars
Published: 28th Jul 2018, 18:41 | Comments
https://videocardz.com/newz/amd-ryzen-threadripper-2990x-listed-for-2400-canadian-dollars

" AMD Ryzen Threadripper 2990X for 1850 USD
A new listing of upcoming 32-core behemoth from AMD has emerged. At CanadaComptuers you will find Threadripper 2990X listed for 2399 CAD, which is around 1850 USD. This isn’t actually the first listing of this new processor. Almost exactly a month ago it was listed by Cybersport for 1509 EUR. Therefore, it may be safe to assume that the next flagship model will cost either exactly or slightly more than 1500 USD.

"Canada Computers" confirmed the codename and that the chip will be a 32-core and 64-thread part. The processor is listed with 250W TDP, 80 MB of total cache. The listing at Cybersport listed 180W TDP.

AMD Ryzen Threadripper 2990X is codenamed YD299XAZAFWOF and according to the picture, it may be available with the same type of packaging as 1st generation Threadripper.


AMD Threadripper 2990X, 2970X & 2950X Specs Leaked – 32, 24 & 16 Cores at 250W, 180W & 125W
By Khalid Moammer, 2 hours ago
https://wccftech.com/amd-ryzen-threadripper-2990x-2970x-2950x-specs-leaked-by-hwbot/

"AMD’s highly anticipated 2nd generation enthusiast Ryzen Threadripper processors are only weeks away from launch and as such we’re witnessing a noticeable uptick in leaks surfacing around these parts. In fact, three of AMD’s upcoming second generation Ryzen Threadripper 2000 series CPUs have been spotted on the HWBOT database.

The entries in the HWBOT database include clock speeds, power envelopes and core counts, confirming earlier leaks about the Ryzen Threadripper 2950X part as well as confirming the specifications of a new 2970X SKU.

AMD Ryzen Threadripper 2990X, 2970X & 2950X Specs Leaked by HWBOT
The three leaked SKUs are the already well known 32 core Ryzen Threadripper 2990X flagship part in addition to a new, 24 core Ryzen Threadripper 2970X and a previously leaked 16 core Ryzen Threadripper 2950X SKU:

AMD officially introduced the Ryzen Threadripper 2000 series at Computex, earlier this year, where the company debuted the world’s first 32 core desktop processor. The company revealed that the processors would launch in the second half of the year and earlier this month we managed to confirm the launch date to be the 13th of August, less than two weeks from today.

We haven’t confirmed yet whether all three of the aforementioned SKUs will debut on the 13th, but all three will be available on shelves before year’s end. Alongside these new parts will be a set of brand new X399 motherboards from AMD’s motherboard partners Asus, Asrock, Gigabyte & MSI.

Although, existing X399 motherboards AMD has assured the press will also support the new SKUs. If true, this means that first generation Threadripper users can double their core count without much hassle, a mighty tempting proposition for those productivity freaks out there.

AMD Ryzen Threadripper 2990WX, 2970WX, 2950X & 2920X Listed on AMD’s Website
A reddit user has spotted four Threadripper SKUs on AMD’s website listed below, although with zero info attached. AMD has always referred to the flagship as the 2990X, the added W to the 2990X and 2970X are peculiar. A theory floating around is that the W could stand for a water cooled version of the same chip, but that’s just speculation.

• 2990WX
• 2970WX
• 2950X
• 2920X

AMD Ryzen Threadripper 2990X – The World’s First 32 Core Desktop CPU
This particular monstrosity of a chip has been confirmed to feature a base clock speed of no less than 3.4GHz, which is downright unheard of for a CPU with this many cores. It’s also said to feature a Turbo clock speed 4.20 GHz with XFR2. A feat that’s attributed in part at least to Globalfoundries’ new 12nm manufacturing process that we’ve seen debut with the mainstream AM4 2nd gen Ryzen parts.

Quite uncharacteristic for a monster like this, the TDP comes in at surprisingly reasonable 250W for a 32 core, 64 thread part. In comparison, the company’s brand new 2700X 2nd gen Ryzen is rated at 105W and that’s just an 8 core part with clock speeds not much higher than those of the 2990X.

The 2990X is expected to launch on August 13th at $1500.

AMD Ryzen Threadripper 2970X – The New 24 Core
Next up we have the company’s brand new ##70X Threadripper SKU. This chip will slot in right under the 2990X and will feature 24 cores and 48 threads. This chip runs at an even higher base clock speed of 3.5GHz. The maximum clock speed of this part however is not yet known, but it’s reasonable to expect that it will at least match that of the 2990X at 4.2 GHz.

Unfortunately we don’t have an MSRP for this part, but it’s reasonable to expect it to come in under the $1499 MSRP of the 2990X, potentially replacing the 1950X at $999-$1199.

AMD Ryzen Threadripper 2950X – The New Low TDP Threadripper
Next up we have the 2950X, this part will feature 16 cores and 32 threads, just like its 1950X predecessor. Unlike its 180W predecessor however, this part is rated at a frugal 125W watts and will run at a slightly lower base clock speed of 3.1 GHz.

This hints that this part may very well NOT be a replacement for the 1950X and could in fact come in at a lower price point to replace the 1920X instead at $599-$799, leaving the brand new 2970X to replace the 1950X.

AMD Ryzen Threadripper 2920X – Mystery Chip
Listed on AMD’s website is a very peculiar SKU that has yet to be leaked anywhere, except well on the company’s own website. We don’t know anything about this chip at the moment, but we could reasonably assume that it’s a 12 core replacement to the 1920X and probably another 125W SKU in the 2nd gen Threadripper lineup. It’s going to be very interesting to see how the company is going to choose to price this part.

CPU Base Clock Max Clock TDP Cores Threads
Ryzen Threadripper 2950X 3100MHz TBA 125w 16 32
Ryzen Threadripper 2970X 3500MHz TBA 180w 24 48
Ryzen Threadripper 2990X 3400MHz 4200MHz 250w 32 64

 

The problem on that pole may be the price, it may be higher than 1,500 for the 2990x.

 

Yeah, it's an... " on-going process of discovery..."

IDK if it's really valid to compare one made up price to another made up price, but yeah, it's all fun and games until someone get's quoted on it and their stock drops, and then it doesn't come true after all.

 

1,500 had me but 1,850 makes me think more than twice. Also while not a big break for Intel, this news is seemingly helping the stock slightly as while undercutting them it is not a cut throat price. The price of 1,500 or even under was having Intel literally shake in its boots.

 

AMD 2nd Gen Threadripper And Intel Core i9 8-Core Cinebench Throwdown Leak Is Drool-Worthy
http://forum.notebookreview.com/thr...e-lake-cpus-z390.811225/page-17#post-10773472

 

2990WX
https://videocardz.com/newz/amd-ryzen-threadripper-2990wx-cpu-z-screenshot-leaks-out

 

The image from that screenshot came a little after computex, IIRC. Kinda old news.

 

The powers that be most likely knew this was to be a date with huge amounts of positive news for AMD. This was the first shot at some positive Intel news. It seems though it never happened.

Not too sure if AMD just did not want to give great news just too have the stock price go down knowing of the sell off or what. I mean I could see the haters posting AMD looses share price despite whatever. Hopefully we will know what is up soon.

 

@Papusan @TANWare @hmscott @Deks - Was wondering if you guys would want me to do a consolidated update post on the Intel 10nm+ process (the one expected next holiday season), and the AMD 7nm process, along with the IPC information on AMD's vs Intel's current offerings, projected IPC, and approximate release dates for each? I know TANWare, hmscott, and Deks have been reading my analyses of each either here or in the AMD thread. But, I still cannot escape this Intel image on 10nm+ Ice Lake and what they said in 2017:


If this is still true, this fall's chips are going to be better on transistor performance than next year's chips, but at a tighter node, Intel can add extra transistors. Now, with the extra density, Intel still predicts either a 25% increase in performance or a 45% reduction in power usage compared to the 14nm process (Intel did not clarify which one, but the article was fairly good at citing the + on different process iterations, so it may be compared to the original Skylake and Broadwell process for 14nm). https://www.custompcreview.com/news...efficiency-2-7x-increased-transistor-density/

 

AdoredTV on Q2 AMD earnings and where things may go.

 

One thing for sure, Intel won't let AMD run away with performance increase and themself instead go for max reduction in power usage. The increased competition does not allow this in the same way as they did before.

 

They have nowhere to go. You really do not understand that, do you? There is no 14nm+++ process and 10nm+ is not great. 25%, if referring to the original 14nm process, means a 10% increase over current offerings. AMD is rumored to have 10-15% IPC alone, then up to a 35-45% increase in frequency potential. Then, if you watch that video I just posted, you find out Lisa Su confirmed that the EPYC Rome chips currently being sampled to partners is being produced at TSMC currently. And, the gate pitch between TSMC and GloFo matches this next round for 7nm so AMD can move whatever design between the fabs as they see fit.

Then you look at this information on density calculations:
https://www.semiwiki.com/forum/content/7191-iedm-2017-intel-versus-globalfoundries-leading-edge.html
https://www.semiwiki.com/forum/content/6895-standard-node-trend.html


This means you are asserting Intel has something up their sleeve while you ignore their problems with 10nm designs at all. So tell me, what do they have up their sleeve? What am I missing? And that is before factoring into Intel missing 10nm chips for mobile and that hurting the ODMs and OEMs, including component mfrs such as touchpads and fingerprint sensors, them losing Apple mobile business last year, Apple now looking to design their main CPU for the desktop in-house, and those laptop manufacturers now putting out more Ryzen professional laptops this fall, the first time they ever did this where all three, HP, Dell, and Lenovo, have these offerings in Q3 and Q4. You have Intel saying they have to hold AMD, at best, to 15% of data center when AMD was 0.5% or so before Zen and is already at 1.5% and should be around 4-5% in Q4, while already having large centers qualify the early 7nm chips put out this Q2.

So, please, tell me what I am missing? What performance does Intel have held back? Because responding with Ice lake in Q3 or Q4 of 2019 and not having Ice Lake server chips until 2020 sounds like they are fairly boned. Can you respond back with sources, not just aspersions on what Intel will do. Give me facts, articles, showing what they have.

Edit: Here is from IEDM 2017 and the curve for GF products at that time:

https://fuse.wikichip.org/news/641/iedm-2017-globalfoundries-7nm-process-cobalt-euv/2/

 

Did AMD NDA the Italy meeting?
https://www.computerworld.com/artic...e-best-workstation-you-didn-t-know-about.html

 

I would not doubt that. If rumors are correct, Nvidia may see a hit on expected earnings for Q3 on their earnings release expected on Aug. 16, 2018. If you have bad news, you may want to embargo something good to muddy the waters.

On Nvidia's front, in addition to what you have seen on Gigabyte's earnings and the stuff on returning 300K for one vendor and Nvidia potentially sitting on more, I offer this for TSMC:

“Our second quarter business was mainly impacted by the mobile product seasonality, while the continuing strong demand from cryptocurrency mining and a more favorable currency exchange rate moderated the mobile softness,” said Lora Ho, SVP and Chief Financial Officer of TSMC. “Moving into third quarter 2018, we anticipate our business will benefit from new product launches using TSMC 7-nanometer technology while cryptocurrency mining demand will decrease from second quarter. " http://www.tsmc.com/uploadfile/ir/quarterly/2018/2y4ZO/E/2Q18EarningsRelease.pdf

"Moving into third quarter 2018. Our business is expected to benefit from new product launches using TSMC's industry-leading 7-nanometer technology, while cryptocurrency mining demand will decline due to weakening cryptocurrency prices. That being said, we do see slight improvement in smartphone demand in second half of this year as compared to our forecast 3 months ago. GPU demand for AI in the gaming continue to increase. For the full year of 2018, we forecast the overall semiconductor market, excluding memory, will grow by 5% while foundry is expected to grow by about 7%. We forecast TSMC's 2018 revenue in U.S. dollar will grow by a high single-digit rate rather than the previously stated about 10% due to general weakness in cryptocurrency mining demand." http://www.tsmc.com/uploadfile/ir/quarterly/2018/2R5bO/E/TSMC%202Q18%20transcript.pdf

"As we stated 3 months ago that we are optimistic about the development of the industry's mega trend, particularly AI and 5G communication. Recently, we have observed more promising development. For example, we see AI continues to fast proliferate from data center to edge server and to end client device." Id.

"In HPC, we expect the increasing workload in data center and complexity of AI will boost the demand for AI accelerator, GPU and CPU in server. We also expect the introduction of next-generation video gaming will add growth of HPC." Id.

Although this is weird to count the next-gen gaming card as HPC growth, which is generally thought of as the server market. This is also after talking about the third quarter and full year, moving into long-term business growth factors. That is an important part of analyzing these statements as this is the fab for Nvidia, so their discussion on when they are doing what market segment determines releases more than rumors.

Q&A (Id.)
Randy Abrams - Crédit Suisse AG, Research Division - MD and Head of Taiwan Research in the Equity Research Department
The second question I have on cryptocurrency. In the past, it was on a lagging node like 28 and 16, where it was filling capacity as a second-wave application. How do you view crypto? It's slowing in second half? But how do you view devoting new capacity as crypto to get the best performance and power want to move to 7. So how do you see that market and also devoting capacity if it becomes more of a first-wave application?

C. C. Wei - Taiwan Semiconductor Manufacturing Company Limited - CEO & Vice Chairman Well, cryptocurrency all related to the hash rate. So naturally, they will move to very high-end leading-edge technologies to improve the performance and lower down the power consumption. Whether that will be that at the same time as a high loading when we move into the leading edge that I -- probably, we cannot say that. It's volatile in the business. You knew it, and it's continued to depend on the cryptocurrency's pricing, so we don't plan our capacity because of that.

Also, they have going from 1% to 10% of production on 7nm, but less than 10% growth and would not say which nodes are going to weaken while 7nm is ramping up (which also goes with the decline in crypto, which could be either ASICs or graphics cards). The known growth comes from Apple mobile chips and other phone manufacturers in communications, rather than the computer or consumer segments, which were high Q2 and made up for soft demand on communications during the normal seasonal slump. http://www.tsmc.com/uploadfile/ir/quarterly/2018/2bAoF/E/2Q18ManagementReport.pdf

Between this and Gigabytes earnings, along with the amount of GPUs Nvidia may be sitting on, this might be a paper launch similar to how no one could really get the V100 on release (which was partly due to limited HBM2, similar to AMD's problem there). But, just some food for thought. http://www.tsmc.com/tsmcdotcom/ExtIRListingQuarterlyAction.do?action=listByYearAndQuarter&year=2018&theQuarter=2&language=E#

 

Yeah but, look at what AMD said about 7nm process... 40% higher performance vs existing parts (though we don't know if this is with or without IPC improvements - if its without, we might see CPU's in the range of 5.5GhZ or more)...next add IPC increases and whatever AMD does to improve Infinity Fabric, and you end up with good hardware on your hands.
Also we don't know what will happen exactly on 7nm+.
AMD may have similar improvements in store for Zen, but I guess we will see.

Plus, Intel's 10nm is still being pushed ahead into the future.
We don't know what kind of enhancements are in store for Zen on 7nm+ for example... apart from probable use of EUV litography...
But, AMD is also ahead of Intel in regards to their ability to scale Zen with Infinity Fabric... Intel however still uses a monolithic approach, and that also costs money.
Ryzen 1 alone managed to upset Intel enough so they would scurry about release their i9 series... which is relatively slightly ahead of AMD but is otherwise much more expensive.
Then Ryzen+ came about and even though the node limitations prevent AMD from breaking through a certain clock barrier, they managed to relatively improve performance and will also be releasing TR 2 which will reportedly boost to 4GhZ across all cores on the new Wraith cooler (which remains to be seen if it turns out accurate).

 

@Papusan
https://www.extremetech.com/computing/274597-intels-entire-2018-2019-roadmap-may-have-just-leaked
https://www.anandtech.com/show/13126/intel-10nm-production-systems-for-holiday-2019
https://videocardz.com/76929/lastes...0k-i7-9700k-and-i5-9600k-to-launch-in-october


https://www.semiaccurate.com/2018/07/26/intel-roadmaps-confirm-semiaccurates-server-articles/

https://www.extremetech.com/computing/274597-intels-entire-2018-2019-roadmap-may-have-just-leaked

If this is true, that Intel is moving the 8-core to Q1, I change my recommendation of getting the Intel 8-core to waiting for the Zen 2/Ryzen 3000 series on 7nm. They are hoping to get the people to hop on them in Q1 and that MB mfrs will do like they did with the Kaby release and screw over AMD's Ryzen 1 release with crappy firmware, etc. I just do not see it, and with 3 months between them, it doesn't matter anymore what performance the 8-core has over the 2700X, because AMD gets to respond in March or April of that year, literally within 2-3 months.

Now, for people not following what is going on with AMD, we are looking potentially at 5GHz chips, 10-15% improvement on IPC, and potentially up to 12 or 16 cores on mainstream platform.
https://digiworthy.com/2018/03/10/amd-7nm-zen-2-matisse-cpus-5ghz/ (Comments from Gary Patton, CTO of GF in March of 2018)
https://www.anandtech.com/show/1243...h-dr-gary-patton-cto-of-globalfoundries#three (Gary Patton interview with Anandtech referenced in the other article)
https://segmentnext.com/2018/03/07/amd-ryzen-3000-series-cpus/ (AMD expected 5GHz)
https://www.overclock3d.net/news/cp...a_10-15_ipc_boost_and_up_to_16_cores_on_am4/1 (self explanatory)
https://www.pcgamer.com/msi-hints-at-amd-releasing-socket-am4-processors-with-more-than-8-cores/ (MSI ad saying 8 core and up for AM4)
https://www.techpowerup.com/246100/msi-drops-first-hint-of-amd-increasing-am4-cpu-core-counts
https://www.techpowerup.com/246282/no-16-core-amd-ryzen-am4-until-after-7nm-epyc-launch-2019 (the absurdity of the framing on this one: "What this effectively means is that the fabled 16-core die with 8 cores per CCX won't make it to the desktop platform any time soon (at least not in the next three quarters, certainly not within 2018). " It was expected March or April, not Q3 or Q4 2018, and that is end of Q1 to the beginning of Q2. EPYC is expected in December of 2018. So just bad writing overall. Plus, as mentioned, if Intel waits, it incentivizes to wait one more quarter for something good. As to the end of that where the person says no Zen+ with over 8 cores, well no duh! It was never expected to happen before Zen 2.)
https://www.extremetech.com/computi...fer-higher-core-counts-major-ipc-gains-report (Discussing the 10-15% IPC and 16-core mainstream chips)
https://www.techpowerup.com/246258/...0-15-ipc-uplift-revised-8-core-per-ccx-design (Same)

So the over 8 cores comes from two places, Chiphell forums (has lots of good leaks there) and an MSI markup for a future board saying 8 cores and up that MSI subsequently pulled the video down.

Many articles said the IPC of 10-15% is not that impressive, so I did some digging. Here is what I posted on the AMD page:
So, update with quality IPC comparison.
https://www.sweclockers.com/test/24701-intel-core-i9-7980xe-skylake-x/19#content

This shows all chips running at 2.8GHz compared. We can see that Intel's 8700K only had a 7% IPC performance (ignoring latency, cache speeds, etc., that cannot be controlled for) over the 1800X and 1700X. That shrunk slightly with the 2000 series. So, getting 10-15% IPC is going to create half of the current gap or the amount of the current gap between the 1800X and 8700K, except in AMD's favor, for the 8700K to the Zen 2/Ryzen 3000 series. When you see that the process should allow for around 5GHz speeds in addition to the IPC increase, thereby matching Intel on overclocked speeds, you have AMD fully caught up or exceeding Intel. That is a phenomenal feat, especially since we already know that AMD wins on multithreaded applications and will now also be at the same level on single core performance. I really am not seeing why people are saying they are unimpressed with that news, especially if they are talking 10-15% over Zen+ instead of Zen.



This shows that clock for clock in CB, the single threaded performance was only around 3.6% (he stated 4%, which is an accurate round up) ahead of Zen+ when comparing it to the 8700K. With the PCMark, it is pretty well matched in multithreaded, and also matched in Monte Carlo simulation. Little slower (like 3%) in Carona, and another render (handbrake) was not able to leverage the Ryzen chip as well. Blender was 4% slower. Just 1% slower in V-ray.

Now, when you get to gaming, the latency is the killer, which is why I hope the talk of interposers is true (along with the benefit for other use cases).

So I hope that adds some perspective here.

Please, @Papusan @Mr. Fox @Johnksss - give me articles to show what Intel is hiding up their sleeve, other than delaying chips and having no response until the end of 2019, potentially delaying the 14nm++ 9th gen until Q1, depending which leaked roadmap is accurate, etc. Please show me a newer update than Intel saying 25% performance or 50% power efficiency on 10nm (which may be over the 14nm process, not the more advanced 14nm+ Kaby and Skylake-X process or the 14nm++ Coffee and whiskey lake CPUs (8700K and 9900K), while their own work in 2017 showed that transistor performance is worse on 10nm+ than 14nm++, although they still expect the 25% in performance. (repeat from last page so in spoiler)

Spoiler: Intel performance

If this is still true, this fall's chips are going to be better on transistor performance than next year's chips, but at a tighter node, Intel can add extra transistors. Now, with the extra density, Intel still predicts either a 25% increase in performance or a 45% reduction in power usage compared to the 14nm process (Intel did not clarify which one, but the article was fairly good at citing the + on different process iterations, so it may be compared to the original Skylake and Broadwell process for 14nm). https://www.custompcreview.com/news...efficiency-2-7x-increased-transistor-density/

So, what is it that you know about the transistor market that I am missing?
https://segmentnext.com/2018/06/28/intel-10nm-delay-resulting/
https://www.semiaccurate.com/2018/0...down-is-crushing-a-20b-market-cap-tech-giant/

Please, tell me what Intel has up its sleeve? I really want to know if I'm missing something. If you need me to post about the density and process and my theory that due to the single dummy gate in the Intel 10nm design that Intel is having too many defects in those gates, which are the only separation between two compartments, all due to it being originally designed for EUV lithography, but that being delayed for about 3-4 years by the time Intel gets a working 10nm chip on the market, how it would reduce the steps from around 35 to 9 steps for fabbing by using EUV, that TSMC will be in volume EUV production by 1st half of next year IIRC, so the idea that Intel is waiting for volume for when they can get EUV up and running isn't that crazy. Then you have the Intel architect that left that suggested he told management to move Icelake to 14nm instead of only having a design for 10nm, etc. Please, fill in the gaps for me. I do honestly want to know.

Edit: Also, this analyst's sources turned out to be right here:


Which were the same ones that fed him this:

AMD research from 2014 and 2015
http://www.eecg.toronto.edu/~enright/micro14-interposer.pdf
http://www.eecg.toronto.edu/~enright/Kannan_MICRO48.pdf

 

I have put up a few post on coming 8 cores here in this thread Intel Core i9-9900k 8c/16t, i7-9700K 8c/8t, i7-9600k 6c/6t 2nd Gen Coffee Lake CPU's + Z390 Maybe read the last 2-3pages if you haven't. But if it's true, 8 cores, 4,7GHz on all cores and 5GHz on single/dual core is damn good. As well keep it on to 95w . And Yeah, I know the power consumption will be exceeded (If you think I don't know).

 

That is just 14nm++, which is the same as coffeelake. What I posted is some newer leaked roadmaps suggest that may be pushed back from Q3 2018 all the way to Q1 2019. Now, if it stays Q3, I will recommend it for the time being. If the Q1 2019 rumor is true, then IPC being the same as coffee lake due to it being 14nm++ isn't impressive at all. That would mean the 10-15% IPC information says Zen 2 will have more IPC than Intel. With GF saying directly that AMD will likely have 5GHz chips or over, where is the value? Reread all that I posted.

I'm asking for new information. I shared the 8c/16t in my post. Is that it? You really don't have anything more than that?

 

None know the future. Thing can change over nigh or next couple of months as you know... TSMC can cut the entire AMD processor production at 7 nanometers-Sweclockers.com

For laptops, there ain't any other options than Intel. There is nothing out there. And it seems it will continue. BGA is out of sight, out of mind (Intel and AMD). Same for Ryzen LGA machines. What you have left without Intel... NOTHING!

 

So, you are pointing to a misrepresentation of a statement by Su? You do realize she said that they are currently at TSMC because their process is more mature. Did you read this?
https://www.extremetech.com/computi...7nm-capacity-than-globalfoundries-can-provide
https://www.eetimes.com/document.asp?doc_id=1333326&page_number=2

Funny how articles from May show you don't follow the industry closely.

Edit: and on laptops, other companies in the industry rely on Intel's systematic releases. Them no longer being reliable is the problem, as that is harming their partners, which likely contributed to the big 3 OEMs putting out business class ryzen laptops because otherwise they have no new product to release during the busiest time of the year.

 

I'm most interested in real hardware. Not what's still on paper (paper tiger) and not released. I jump on new tech when I have read what they can provide me. Not before or same day it'will be released. But as you know, with laptops you haven't many choices. Either buy or skip it. Just use what you have. For the records.... Not me who writes the articles or pick what should be in them.