AMD's Ryzen CPUs (Ryzen/TR/Epyc) & Vega/Polaris/Navi GPUs

Thinking about it, if AMD does not let TR run on x399 there may be another option. That is to wait for Zen3 and AM4 for the 16 core variant. This would maximize my current hardware investment and hopefully get another 8% or so IPC.

 

Yeah, I'm thinking, since my budget was spent, wait for 2021, but do not disagree with that idea.

What I'm hoping AMD does is to put transcoding as part of their hardware decoding and encoding, focusing at minimum to give more streams than Nvidia's 1660/Ti/super. Basically working with companies like Plex for low to midrange cards to accelerate on the fly streaming for media collections. This is coupled with them supporting emerging standards like VP9 and AV1, along with HEVC H.265 4:4:4 at multiple levels. But that is to buy at the low end for my media server. I'm hoping they have something at the high end for my gaming/workstation build.

So far, I'll in 2021 or 2022 upgrade my server and workstation systems, or at least try to do both (might be one one year and the other the next). Will probably upgrade the router around then, too. Waiting to see if PCIe 5.0 and Cat8 cables, which can carry 40Gbps over twisted copper for 30M (approx 90ft), will help to drop the costs of 10Gbps switches, which can be had at the low end for $360 while closer to $600-700 is more what people would likely look at for switches. Granted, this only effects your LAN because we are basically becoming a third world country by not laying fast fiber, which I don't mind so long as they have a plan in 2023, with PCIe 6.0, to lay shirt tons of fiber that can utilize all that bandwidth and build out more decentralized backbones, etc. At that point, a single cable should be able to get up to 600Gbps for fiber, meaning burying those bundles will allow for open throughput with little to no need for any throttling, etc. It would truly change how our infrastructure is managed. But that is my dream.

In any case, that may be the time for my router as well, which my new NAS raid 6 array currently can fill a 10Gbps connection, so it's time to start planning for that.

 

gonna be hard to justify buying zen 3 with pcie5/ddr5 support on zen 4 just the year after. not to mention possible golden cove with 25-30% IPC boost + optane dimm support comes around similar time as zen 4. interesting time indeed.

 

Every time you post, for as long as I can remember you are putting off purchasing for some reason or another. Now you are putting off buying Zen 4 - 2 years in advance.

Maybe buy something now, a nice 3700x or 3900x, or if you start shopping around right now, by the time you get the parts together you could get a 3950x - that should be a fun ride.

Although, I think for that 3950x 16c/32t there might be new motherboards coming out in unison to support it as well.

I guess wait another couple of weeks?

 

The optane Dimm isn't anything. See the X299X reviews next month.

Agree on waiting. But let's see Intel hit a new process node when they said they would! That is laughable. But seriously, if they miss on 7nm in 2021, then it is real bad for them. Also, let's wait and see on their frequency, which ice lake server chips will tell us all we need to know. Remember, ice lake mobile had IPC gains, but lost almost that much on frequency. So not worth talking about when AMD hitting 5nm in 2021 is practically a sure thing.

 

He has a 9900K. That means he could ride out to 2021. Just like how GN showed a 1920X for $200 was still a good buy this year and me waiting for Zen 4. If the 24-core is $1000, I'll be chaffed about spending the money already, but still will likely wait.

 

With all of the AMD activity happening right now, and given how often I've seen people upgrade their Intel gear - yearly or sooner - I wouldn't wait on supporting AMD at their best in decades - now's the time to buy AMD and buy often.

Heck, don't just buy AMD for yourself, buy AMD for your family and friends too. AMD for work, charity, and any special event you participate in. Get that R&D pool filled for AMD to continue innovating moving forward.

If you wait to buy you might miss it all; all of your Intel value.

Who wants a leaky Intel CPU now anyway, why wait for the hacks and incursions to become public enough that everyone will notice - then your investment in Intel will be lost - better to turn over the Intel crap to some other poor sap now, pass the hot potato right away before you get burned, and move to AMD while you can recover some of your Intel investment.

By the time AMD has Zen 3 / Zen 4 Intel will have slashed prices for everything - multiple times - so your Intel investments will be worthless.

Heck, rumors are that Intel is going to slash the desktop prices the same as the HEDT prices, and that could happen any time after the holidays.

Likely as not Intel has already filled the pipes for the holidays and are hoping to suck as many dollars as possible from unwary buyers wallets, and will wait to slash prices after the new year - but I guess Intel could slash desktop CPU prices sooner given AMD's market share gains.

The more performance records AMD takes, the less value the Intel CPU's will maintain, and eventually Intel is gonna have to drop prices across the board. As Intel looses it's performance lead Intel also loses it's justification for maintaining it's outrageously high prices.

AMD holds #1 and #2 highest R20 records; dual EPYC 50% ahead of quad Xeon
Submitted 8 hours ago by xcalibre
https://www.reddit.com/r/Amd/commen...and_2_highest_r20_records_dual_epyc/?sort=new

https://hwbot.org/benchmark/cinebench_-_r20/halloffame

IMHO, Now is the time to sell all of your Intel stuff before it loses all of it's value. Buying all AMD now is your best investment.

 

Benchmarks don't make what someone has and uses obsolete, nor worth, less. Not even 'world records' that change faster than I change my wardrobe.

Once you pay/buy a product it is a sunk (and should be a forgotten) cost. Nothing you do afterward makes it a better (or worse) decision. Except worse in your case with your great advice to all Intel users out there to start all over with a new platform that still isn't stable after how many years now? is it almost three years already?

The price slashing can only go on for so long, for either company. But AMD gets it, even if you don't want to admit it: it is charging a pretty penny where it can now. Just like the good little greedy corp it is.

Sigh, R20 'records'.

Yeah, I don't have breakfast or check my email on my Intel systems until I check that inadequate metric about whether the platform decisions I make and made are still applicable to mine and my clients' workloads and workflows.

Thanks for giving the worst advice ever on NBR. What an AMD shill.

See:
https://www.tomshardware.com/news/a...nue-since-2005-with-7nm-ryzen-radeon-and-epyc

 

3dxpoint from micron is my only hope to getting an AMD system. I could do away with just PCIe AIC optane SSDs but if im gonna spend that much might as well get the best and 3dxpoint dimms is where it's at.

i actually dont think intel will have issue with 10nm icelake server chips. they dont need high clocks unlike consumer desktop part. as long as yields are decent they are good.

 

Dude, 3Dxpoint DIMMs are useless except in scenarios where you run out of memory and rely on the page file system. That means they are worthless for 99.99999% of all users out there. There are very specific commercial uses that benefit, like 8K video rendering, CAD rendering, etc. Other than that, it is a huge expense with very marginal gains. But it is your wallet. But for the price, personally I'd go mega amounts of DDR5 ram at that point or do a massive storage array. I mean, once I'm done moving everything to my array, I'll have like 25-30TB of my 43.65TB raid 6 array full (before running data dedup). Then I'll have 3x8TB in a raid 5 for my workstation (so another 14.4TB, about). I get around over 840MBps read on my raid 6 currently, meaning I've now got to look into 10Gbps networking (1.25GBps approx.).

I just do not yet have any need for optane, except for adding a cache tier to my storage arrays (which I've considered their 128GB variants for my workstation, but would likely do a striped nvme 2x2TB for the raid 6 array).

So what is your need for optane?

Edit: my avg write speed moving stuff to the array was around 130-170MBps. This is with storage spaces on windows server 2019, BTW. That is where the cache drives are handy! But even my speeds are trouncing many out there, although non-sequential and small file sizes would drop it to 30-70MBps.

Edit 2: on ice lake server, that is their saving grace, although they are at a TDP of 270 or 280W at that slow of a speed with either 34 or 38C. That is still very power hungry! That is why, even with low frequency being normal for those parts, I'm assuming power will effect TCO and don't know if the IPC will fully make up for the drop in frequency when against Epyc. There is a reason they are planning on 14nm for high performance server parts next year alongside ice lake server.

 

Sounds interesting I wonder how that would affect shadow of war?

 

There is always something better and/or enticing around the corner. TR zen 2 was for me but that may be dead. TBH the 1950x is way more than I will need for a LONG time to come. I just want to keep my 16 cores but get that IPC gain.

 

What’s special about SoW, does it really hammer I/O or something?

 

I need help for my grandson, I need Zen 2 to support x399 so we can solve the puzzle below;

Spoiler

 

It has a function built into the game to deliberately take advantage of paging file

 

Interesting. Having you observed any performance increase from enabling Large Page Mode?

 

I never got around to it lol, might try this weekend or look for some data elsewhere on the net while working tomorrow.

 

Yup, Micron owns the 3dxpoint manufacturing facility and after the first of the year Intel doesn't have shared access to the manufacturing output - other than purchasing through Micron - at least that's how I read it:
https://www.anandtech.com/show/14183/intel-to-align-3rd-generation-optane-dcpmm-with-ddr5

It looks like you don't need to wait any longer? Micron announced the X100 NvME SSD:

Introducing the Micron X100 NVMe™ SSD
The Micron X100 NVMe™ SSD is the first solution in a family of products targeting memory and storage applications for the data center. This solution leverages the strengths of 3D XPoint™ technology, offering higher capacity and persistence over DRAM along with higher endurance and performance as compared to NAND.
https://www.micron.com/products/advanced-solutions/3d-xpoint-technology/x100

Introducing the Micron X100, the world's fastest SSD.
MicronTechnology
Published on Oct 29, 2019
The Micron X100 NVMe SSD is the first in a new class of high-performance products based on 3D XPoint technology.


Micron CEO on New Chip - Micron Ditches Intel for new Chip
Bloomberg Technology
Published on Oct 24, 2019
Oct.24 -- Sanjay Mehrotra, Micron Technology Inc. chief executive officer, discusses the company's new 3D XPoint memory technology, the tailwinds from the deployment of 5G and Micron's relationship with Huawei Technologies Co. He speaks with Bloomberg's Taylor Riggs on "Bloomberg Technology."


"3dxpoint dimms is where it's at..." IDK, it seems to me that the bird in the hand is already fast enough. You can stop waiting and get in on 3dxpoint right away - soon? When is it shipping?

 

it's only useless if the user and software can't take advantage of it. infact, it'll be less useless than zen many cores cpu simply because any files on it gets a boost, no matter how small the boost is. where as a single threaded software is simply just that, use single thread and have 0 benefit from many cores except maybe the cache.

to name a few:

- index search of entire C: drive with easily over 300k files.
- virus scan entire C: with many files would likely take half of 50 mins
- speed up small files read/write both sequentially and random
- UE3/4 level editing over 5-10k texture files are beneficial, then load said level when testing it
- run multiple vmware, easily stack up to higher QD random read/write
- great for torrent/vmware, much higher endurance than NAND flash
- faster reinstall window, boot, backup/restore, faster on every software

Ive been running server 2012 OS since my m18xR1 days like 7 yrs ago. server 2012 counter part is win8 so in terms of performance it is above win10/2016/2019. also I use fastcopy to bypass lots of storage/system caching, only files smaller than 1KB goes to ram then to disk.

we just have to wait and see how icelake does thing. intel could be tryign to play the investors yet again. saving grace or what not wasnt what i was talking about. I simply said if their yields are good then they would have no problem with icelake as server chips are much less dependent on higher frequency.

yep i saw that. thats just what im saying if i were to put so much into a new rig then i'd want the best. X100 goes through PCIe so it'll probably get me 400-500MB/s random read/write at best (which is still amazing). I may as well go for a smaller capacity optane dimm for cheaper and getting 3x random performance.

 

How are we gonna get back to the Moon now?

Maybe best to not wait for Zen 4?

 

Or, Micron could offer that X100 SSD as a PCIE 4.0 part and get that uplift in performance right now, instead of waiting for DDR5. 3dxpoint is likely too slow to know the difference between DDR5/4 DIMM slot and PCIE 4.0 / 3.0

AFAIK that 3dxpoint is still slower than fast DDR4/5 memory, and is supposed to act as a cache between memory and slower storage. 3dxpoint storage could fit in a tiny DDR4/5 DIMM sized part but easier to manage using current hardware technology as a large format SSD on PCIE as fast cache layer storage.

Better to save the DIMM slots for faster RAM and use the 3dxpoint in the X100 format using PCIE as replacement for existing fast SSD cache. You could fit 3dxpoint as an M.2 SSD, even more storage than an SODIMM sized format as a 2280 / 110 / 120 sized M.2 part for laptops.

An bunch of X100 3dxpoint using PCIE 3.0 / 4.0 would be perfect compliment to AMD EPYC / ThreadRipper with all of those PCIE lanes ready for huge storage expansion.

 

depends. when ram is so much faster than HDD and SSD it maybe better to have more ram as cache if one is willing to have data flushed down to storage each time they shutdown this process is extremely annoying when you're trying to troubleshoot an issue and requires a bunch of testing, restore, reboot etc. optane dimm storage will not have such a problem as files are retained.

example of a single channel on 2nd gen optane dimm. random read/write already faster than my ramdisk and compared to actual ram only 2-3x higher latency. optane dimm will replace OS drive + memory caching and not having to worry about everything else.

that x100 will probably cost upwards $10k. i'll stick to 3rd gen optane diimm 2x 128GB for probably under $1k.

 

I call BULL! This was tested with Optane drives already. Unless you have a program that leverages it, average tasks DO NOT TAKE ADVANTAGE OF IT! PERIOD! No one has taken as deep a look at optane DIMMs to be fair, but there is little reason to believe there is ANY major difference in outcome.

The only things you said which have been shown to have ANY benefit is the VMs IF you are low on memory, which once again invokes the page file. Also the editing texture files, although the only benefit that was seen on that was WHEN IT FILLED MEMORY AND USED THE PAGE FILE. The OS crap is crap. Only time that MAY come up is once again pointing to loading VMs. Other than that, on a regular OS, it REALLY does not matter.

Server 2012 lacks many features added since then. Reason I'm on 2019 is specifically Storage Spaces/Direct optimizations (and I got a good deal on the license). Win 8 SUX! ALWAYS HAS. The only difference is level of M$ Spyware, and yes, I am still learning optimizations for Server 2019 with desktop gui (specifically the desktop gui). You forget you are talking to a person that makes a custom iso with integrated updates for Win 10 MONTHLY! Unlike many people on these threads bemoaning the death of Win 7 (Win 8 and equivalent are EOL and will NOT receive the same extensions Win 7 did). I instead updated my customization skills I had for Win 7 to Win 10. Performance doesn't really have the uplift people want, but the spyware is gutted, which is MORE important.

Also, show me statistics and proof Server 2012 is better on performance than anything above it, including R2, which you left out. I'll wait.

And my use on it is specifically archival, although after I upgrade my server to something more powerful, I have more plans for VMs on it (standard is limited to two instances of Windows Server VM with the license, IIRC, whereas Datacenter has unlimited).

But with all that said, there are so few feature changes that unless you need SPECIFIC optimizations and changes in the software, there is no reason to upgrade to the newer versions of server. And in that regard, I feel ya and Server 2012/R2 (I'd prefer R2 from my readings, but...) are REALLY solid, and 2016 and 2019 both REMOVED certain functionality while adding small features, which if you need those, the later versions are non-starters (MS is going to tighten that noose when 2012 becomes EOL and they hear customers screaming about no upgrade path or more find ways to move over to Linux Distros, etc.). At least that is my understanding ATM.

Even with that, there is more involved with the server chips than just them using low boost clocks or base clocks. There are a number of large players that ignore the boost and may even lock them to base clock speeds, which would make Ice Lake untenable. With that said, there are specific changes in certain engines under the hood which WILL be nice (those changes will NOT be in Cooper Lake SP). As such, they are trying to create products to satisfy all layers, whereas I just do not see how Ice Lake will have enough to effect the TCO evaluations of these companies. Intel labeled the 9900K a 95W TDP when it pulls like 127-185W. They labeled Cascade SP and Skylake SP in the low 200W range when it could pull up to 550W. Now they label these new chips 275W and I'm scared of what the actual power draw could be. Reason they increased it is to try to get the base clock, which many companies do lock it to, higher so that the IPC increase may show some usefulness aside from the architectural changes I lauded a couple sentences ago.

Because of those factors, I really think it isn't going to move the needle much. It does a little bit for certain workloads (and I am interested to see the AVX512 improvements since AMDs 64 core basically negated that benefit by being able to keep up with the 28-core using AVX2 for the same tasks already and these chips will be going against Zen 3 with 8% more IPC). But, if they can pull off the extreme core count dies, it does speak to solving the problems on the process node, and to that I will give credit, even if so long after it should have been released.

And even with that, you miss the point on WHY Ice Lake cannot hit higher frequencies (hint: that stuff is hot, yo). That means that they will have to reduce frequencies EVEN WITH AN ALREADY LOW FREQUENCY PRODUCT. So your argument DOES NOT COMPUTE. With that said, I am wondering if the over 28 core variants employ a multi-die/node solution. Microsoft did, after all, fix the problems AMD was having with thread thrashing and NUMA awareness once Intel needed it this year for the 48 and 56-core variants. Makes me wonder if their 30 some core Ice-Lake is actually two dies with core counts in the teens. But we'll find out more as time goes on.

Also, IPC*frequency=IPS (instructions per second) which is what we commonly think of for performance. Ice Lake has the IPC, but dropped, at least on mobile, 1GHZ of boost, which was about 20% of the frequency, thereby greatly reducing the benefits of the IPC except in specific workloads which were accelerated, like CB R20 able to use the AVX enhancements to go faster than otherwise. There are some really good changes with Sunny Cove. I have not changed my opinion on that. But the process kneecaps it. Thems the facts and I haven't seen anything yet to contradict that understanding.

 

Ok, so it all already exists, what are you waiting for? Or, do you already have that in your 9900k system? Where did that Crystalmark run come from / on?

 

looks like you never owned an optane SSD to test individual workload to see any benefit. only vmware benefit from it once memory is filled and use the page file? sounds like load of crap if I have to wait till my memory all filled. my pagefile is disabled since day 1 of my installation of server 2012 7 years ago and anything i do on optane SSD is faster than it is on NVMe SSD. the list of things i mentioned are the ones I can remember off the top of my head that i specifically run to note down noticeable difference that is WORTH paying for optane.

you will have to present to me that server 2019 is faster than 2012 by default of it's storage driver. I have tested since 2008 days. 2008 & vista -> 2008 r2 & win 7 -> 2012 & win8 -> 2012 r2 & 8.1 -> 2016 & win 10. with each windows revision, storage performance gets worse. (not only apply to storage as well, @Mr. Fox does other benchmark he knows)

though I have not tested 2019 I can go with my experience and make good assumption about it simply because i tested each one of those OS and compared. Jon from tweaktown also did the test yrs ago of different consumer OS, server OS etc. gotta show some results then maybe i'll believe you.

honestly though who cares what intel labels TDP? it affects clevo's laptop design and that is as much as i'll worry about it. other than that im not gonna buy server chips and im only talking about how t hey would have no problem making icelake CPU if they have no yield problem. this is so straight forward that it applies to AMD as well. if yields are bad, CPU might get very expensive and not be made as much, even with its chiplet advantages the same rules apply, just to a lesser extent.

 

Another interesting interview with Lisa Su...

How Lisa Su Turned Around AMD
Fortune Magazine
Published on Oct 22, 2019
We talk to Su about her leadership at Advanced Micro Devices.

Spoiler: About FORTUNE...

FORTUNE is a global leader in business journalism with a worldwide circulation of more than 1 million and a readership of nearly 5 million, with major franchises including the FORTUNE 500 and the FORTUNE 100 Best Companies to Work For. FORTUNE Live Media extends the brand's mission into live settings, hosting a wide range of annual conferences, including the FORTUNE Global Forum.
https://en.wikipedia.org/wiki/Fortune_(magazine)
First issue: September 1929; 90 years ago

 

I just want to see where she said about the delay that for Threadripper fans " it will be worth the wait" because without x399 support for current fans it will not be.

 

Yeah, I really think they borked things on HEDT this gen, and next, which is a shame. Especially with rumors the entry level boards will be a couple hundred higher in cost for TRX40, which means like $500+ for the entry boards and closer to $750-$1K for the flagship boards, if the rumor was true (like Asus prime vs Zenith Extreme line). At those prices, even the $1K price for a 24-core isn't necessarily going to fix the issue on competitiveness.
https://wccftech.com/amd-3rd-gen-threadripper-cpu-trx40-motherboard-prices-leak/

 

Now a 24 core at $1,000 and x399 would have been a home run. as it is looking it is just like a runner on base with two outs. Threatening a score but just not there yet.

 

Yuri @1usmus (the Ryzen RAM calculator guy) dug in to the AMD Windows Power Plan and thinks he has improved on AMD's settings to give a 3900x stable 4.6ghz single core with less power and heat generation, he's going to release it on the 4th and he thinks those changes will apply to the 3950x:

Юрий @1usmus 11:50 AM · Nov 2, 2019
Release date - 4 November
First of all, this applies to owners of 3900X and future owners of 3950X. I sent the appropriate recommendation to AMD, I really hope that in the near future it will be implemented at an official level.
https://twitter.com/1usmus/status/1190702781121867777

Юрий @1usmus 11:47 AM · Nov 2, 2019
My dears, I have prepared my own power profile for Ryzen, it will fix all problems with boost and also reduce the temperature in idle and loads where all the cores are not involved. On average, I get a difference of 200-250 MHz
https://twitter.com/1usmus/status/1190701909474258944

Юрий @1usmus Nov 2
All processors will get an improvement

Юрий @1usmus
Only zen 2

Leading up to the Windows Power Settings he was working on the AGESA 1004B testing - see the full thread(s) for details - then he dug into the power settings:

Юрий @1usmus 1:43 AM · Nov 1, 2019
AGESA 1004B,positive moments (part 2)
* Noticed a slight decrease in the temperature of the chipset 59 -> 55 degrees (I don’t think it’s getting colder in my room lol) * Improved compatibility for new memory based on Hynix and Micron chips * Improved training
https://twitter.com/1usmus/status/1190187605855473664

Head Up! 1usmus did it. Bios 1.0.04 at 4.6 Ghz single core boost on 3900x
Submitted 4 hours ago by Tik_US
https://www.reddit.com/r/Amd/comments/dr2lv0/head_up_1usmus_did_it_bios_1004_at_46_ghz_single/

1usmus update on new power profile for Ryzen to fix boost issues
Submitted 23 hours ago by Mograine_Lefay
https://www.reddit.com/r/Amd/comments/dqqq5o/1usmus_update_on_new_power_profile_for_ryzen_to/

 

helluva guy, yuri, evolving into a kind of tweaker wiz for the ryzen community

Sent from my Xiaomi Mi Max 2 (Oxygen) using Tapatalk

 

Yup, it just take's engineering resources with time to apply to exploration. AMD needs him or someone as experienced in house to keep beating on the OS / Power Settings to improve Windows performance. Here he's hit 4.7ghz stable enough on his 3900x for a best CB15 run. I think he's continuing to tweak the options for improving the Core 1-4 performance on top of all core:

Юрий @1usmus 12h
Replying to @Cat_Merc and @VideoCardz
220 on 4.7

 

Just waiting for tomorrow and some real Threadripper information.

 

I haven't read it yet, but here is 1usmus's power plan tweak's published through TechPowerUp:

1usmus Custom Power Plan for Ryzen 3000 Zen 2 Processors
by 1usmus, on Nov 4th, 2019
https://www.techpowerup.com/review/1usmus-custom-power-plan-for-ryzen-3000-zen-2-processors/

Юрий @1usmus 4h
1usmus Ryzen Power Plan - fix problems with boost , idle state and more smoother frame rate. Enjoy. The owners of processors with two or more CCDs will notice the maximum effect. Only for Zen 2 ( sTRX4 included ).
techpowerup.com/review/1usmus-custom-power-plan-for-ryzen-3000-zen-2-processors/…
https://twitter.com/1usmus/status/1191385264645652480

Download the 1usmus Custom Power Plan
TechPowerUp is hosting the 1usmus Custom Power Plan. Find it in the download link below.
https://www.techpowerup.com/review/1usmus-custom-power-plan-for-ryzen-3000-zen-2-processors/2.html

That's the link to the 2nd page where the download / installation info resides, but I'd read the whole thing a couple of times before risking install - maybe even wait for some feedback from others that have installed it. I wouldn't be surprised for development to continue evolving.

 

I installed the power plan on my 2550u and can tell you running user benchmark I could only see before 3.4 GHz on CPUz, with the new plan I see 3.6 GHz on occasion now. Desktop went from a best run of 46% to 49%.

 

Well November 5th is upon us, hopefully some news today.
https://www.guru3d.com/news-story/amd-ryzen-threadripper-3960x3970x-and-3990x-launch-dates-leak.html

But

https://www.anandtech.com/show/13538/amd-investor-relations-next-horizon-november-6th

 

From the 1st article, it's a rumor not an official AMD date for release:

"News reaches the web today that AMD will release three Threadripper 3000 SKUs: 3960X, 3970X and 3990X. The first two would be unveiled on November 5th, the embargo on sales and reviews would be lifted on November 19th.

The news reaches us through videocardz who claims to have confidential documents from 'a source at AMD'. They cannot guarantee that the dates presented mentioned are still valid, but do confirm that they were just that a few weeks ago..."

So, that's a rumor of a specific date...not an official AMD date.

The 2nd article you quoted is from / for 2018...it's already too old to be relevant.

The only "official" news we have from AMD says "November", with no other specifics as to the exact date in November:

An Update on 3rd Gen AMD Ryzen Processors Availability
Blog Post created by amd.corporate on Sep 20, 2019
https://community.amd.com/community...n-update-of-amd-ryzen-processors-availability

"Hi Everyone,

We are focusing on meeting the strong demand for our 3rd generation AMD Ryzen processors in the market and now plan to launch both the AMD Ryzen 9 3950X and initial members of the 3rd Gen AMD Ryzen Threadripper processor family in volume this November. We are confident that when enthusiasts get their hands on the world’s first 16-core mainstream desktop processor and our next-generation of high-end desktop processors, the wait will be well worth it.

Thank you

AMD Corporate"

So, no news until AMD actually makes some news by saying something specific. And, that AMD announcement can happen any time in November.

 

who wants to bet that launch will happen on Nov 29th? like, the very last possible day in Nov hahaha

Sent from my Xiaomi Mi Max 2 (Oxygen) using Tapatalk

 

I am so tired of waiting on this. If they wait till the 29th they may miss the holiday season.

Edit; waiting on bad news is worst than waiting on good news!

Edit;
https://www.eteknix.com/amd-to-announce-3rd-gen-threadripper-on-november-7th/
https://videocardz.com/newz/amd-to-announce-3rd-gen-ryzen-threadripper-on-november-7th

 

New Zen 2 based APU's in early 2020, new laptops at CES? Maybe we'll see some 3700x / 3600x / 3900x? in laptops at CES 2020 too; it could happen...

Lisa Su confirms AMD Ryzen 4000 chips will be “coming in early 2020”
https://www.pcgamesn.com/amd-ryzen-4000-early-2020-renoir-apu

"AMD CEO, Dr. Lisa Su, has confirmed that the red team’s next generation Ryzen 4000 Renoir APUs will be launching early next year. We expect that means January’s annual Vega techgasm, CES, will be featuring a whole load of AMD-based laptops to kick off the new year in style.

The new Ryzen 4000 7nm mobile processors will be based on the current Zen 2 processor architecture, but it’s the choice of graphics component that we still don’t 100% know about within the new APUs."

 

I thought it was the 7th of November or so from some other announcement made in September or early October.

That is announcement, not release, though.

 

And here we go;
https://hothardware.com/news/amd-ryzen-threadripper-3960x-3970x-zen-2-cpus-listed-early-online

Edit;
And AMD, I waited but am not happy. It should have been let known about x399 long ago. I can understand why;

1.) If a x399 drop in sales would have been heavy from the get go, supply shortages.
2.) Since TR is only about 2-3% of CPU sales try and convince board makers to make a new board that would be a hard sell.
3,) if you let on early try and sell x399 boards and older TR chips.

So because of marketing user base takes the hit.

 

Some more;
https://www.tweaktown.com/news/6860...t-gen-ryzen-threadripper-packaging/index.html
https://www.neowin.net/news/amds-threadripper-3000-packaging-leaked-allegedly-launching-tomorrow

 

Nothing to be happy about if the ODM's prefer do it as this...


Acer throttles Ryzen 5 very strong notebookcheck.com

With the Aspire 5 A515-43-R057 Acer has a very affordable notebook with Zen + CPU in the range. The fact that the actually good performance of the Ryzen 5 3500U does not make it noticeable is very disappointing....

 

"But, do I *need* 64c/128t?" - Everyone

AMD Ryzen Threadripper 3990X: 64C/128T launches January 2020
AMD will reportedly unveil third-gen Ryzen Threadripper CPUs on November 7
By Anthony Garreffa
https://www.tweaktown.com/news/6859...90x-64c-128t-launches-january-2020/index.html

"AMD will reportedly be unveiling and detailing the Ryzen Threadripper 3960X and 3970X processors, while just reportedly teasing the higher-end and flagship Ryzen Threadripper 3990X processor.

We should expect the name and possibly core/thread count (64C/128T fingers crossed) but clock speeds and pricing will be held. The reviews of the AMD Ryzen Threadripper 3990X are set to hit in January 2020, while the Ryzen Threadripper 3960X and 3970X processors will find reviews ready for November 19 which isn't too far away from now.

According to the previous rumors on AMD's new Ryzen Threadripper 3000 series, this is what we should expect:

AMD Ryzen Threadripper 3960X - 24C/48T
AMD Ryzen Threadripper 3970X - 32C/64T
AMD Ryzen Threadripper 3980X - 48C/96T
AMD Ryzen Threadripper 3990X - 64C/128T"

I'll wait for reviews, benchmarks, user testing - to see how much and what the extra memory channels improve performance...for now, still waiting for the 3950x, that should be "enough" for me.

Sure is exciting...

 

TBH going 32 core and above with the lowered clocks is more for specialty use. HEDT says specialty use already but even more so with the high core counts. Even the current 24 core is so now.

The 3960x will be the new 2950x, the performance champ level of high end consumer HEDT. The 3970x will be also but with concessions.

 

agreed. nothing to add

Sent from my Xiaomi Mi Max 2 (Oxygen) using Tapatalk

 

And here we go again as it is the 7th;
https://www.eteknix.com/msi-trx40-pro-threadripper-motherboard-images-leak/
https://www.overclock3d.net/news/cp...ses_next-gen_aorus_threadripper_motherboard/1

https://www.eteknix.com/amd-3rd-gen-threadripper-prices-are-possibly-revealed/

 

That chipset fan looks like it means business...nice to have the multiple M.2 card and 10G ethernet included, but that means it's gonna be pricey. I wonder if there will be a sku without any extra expensive addon's? Probably not at first.

 

The 12 stage VRM too.