AMD Releases Ryzen PRO Processors Worldwide, 8-Core Ryzen Threadripper 1900X (techradar.com)

Today, AMD announced the global release and broad adoption of AMD Ryzen Pro desktop processors. At its launch event in New York City, the company touted three main pillars that define these chipsets: reliability, security, and performance. They support features like Trusted Platform Module 2.0, which integrates secure microcontrollers into devices, GuardMI technology, which enables silicon-level security to help protect against threats, and SenseMI technology, which consists of a collection of smart features that aims to fine-tune performance for most responsive applications. For the first time, AMD has partnered with the top three PC OEMs: HP, Dell and Lenovo. Brad Chacos for PCWorld provides a "rundown of the commercial-focused Ryzen Pro systems that are coming down the pipeline, straight from AMD":

-Dell Optiplex 5055 desktop PCs are expected to ship in the coming weeks.
-HP EliteDesk 705 desktop PCs are expected to ship in the coming weeks.
-Lenovo ThinkCentre M715 desktop PCs are expected to ship in the coming weeks.
-Lenovo ThinkPad A475 and A275 notebook PCs are expected in Q4 2017.
-Ryzen PRO mobile processors are scheduled for launch in the first half of 2018.

The global launch of the Ryzen Pro processors is not the only bit of news AMD announced. The company also announced the release of a new budget Threadripper 1900X model. From a report via TechRadar: AMD has released its 8-core Ryzen Threadripper 1900X processor, offering people who were put off by high price of the flagship 16-core Threadripper 1950X a chance to build a PC with all of the advanced Threadripper features for almost half the cash. As we expected, the Threadripper 1900X will come with eight cores clocked at 3.8GHz, with a turbo that reaches 4.0GHz (and an XFR boost to 4.2GHz), and will cost $549 -- almost half the Threadripper 1950X's $999 asking price, and a fair bit cheaper than the mid-range Threadripper 1920X, which costs $799. In fact, the price is within touching distance of the AMD Ryzen 7 1800X, which comes with eight cores and 16 threads, and costs $499.

Comments

i7 8 core 3.07

[Hognoxious]

i7 8 core 3.07G throttled to 1.6 until I get a new heatsink. Still I/O bound before the CPU's maxing out.

I wonder what the point is.

Re:i7 8 core 3.07

[Archangel Michael]

I wonder what the point is

Fix your IO problems. Depending on how much money you want to spend maximizing it, there are some interesting technologies out there that can feed your CPU all the data it can churn through.

Re:i7 8 core 3.07

[Hognoxious]

If you use firefox you'll probably need to upgrade sooner than that...

Or eclipse, for that matter.

Re:i7 8 core 3.07

[smallfries]

So it really is a supercomputer on the desktop.

"A supercomputer is a device for turning compute-bound problems into I/O-bound problems." - Seymour Cray / Ken Batcher

More competition is good

[sinij]

More competition is good, Intel pretty much sat on their hands for the last long while. In my workstation I really don't care if it is 200W or 10W CPU, but I do care how long I have to wait for something to compute.

Re:More competition is good

[MBGMorden]

Yeah. While there were small gains here and there, for the most part a 1st Gen Intel i7 is nearly as capable as the ones they're still releasing now 9 years later.

Granted, they're not BAD, but then again since Intel has been the benchmark they're only not bad because even though they were making minimal improvements nothing was faster.

Heck in my gaming rig I was running a Core 2 Quad that I'd likely STILL be running if I didn't have to upgrade the CPU and motherboard to go beyond 4GB of RAM.

Re:More competition is good

[sinij]

Sure, I am also using laptop. It is plugged in 99% of the time. Even on the plane you now have power outlets to charge your electronics.

Re:More competition is good

[jellomizer]

I really miss the simple days of x86 where the bigger x is the newer the chip and the the number of megahertz where the higher the megahertz the faster the chip.

now we have these i3, i5, i7 and i9 but they span mutable generations so I need to know the stupid code name to figure out who is newer. The Ghz range peaked around 3ghz. Then if you ask which is faster you get the question what are you trying to do.

Re:More competition is good

[Anonymous Coward]

It's almost as if adding multiple cores multiplies the complexity of the chip! Or worse, programs might be written to take advantage of parallel processes instead of a single thread! The horror!

Awesome

[MBGMorden]

I've always been an AMD fan, but their offerings have been lackluster lately. On my gaming rig I finally went Intel and on my regular desktop machine I'm still running AMD but it's an ancient 7-8 year old Phenom II. I haven't had much reason to upgrade it until now but I'm thinking I may finally pull the trigger on a system built on the "budget" Threadripper.

Re:Awesome

[Anonymous Coward]

You wouldn't be disappointed with the ryzen 5 or 7 lineup either, but if you dont want to upgrade for another 10 years, threadripper is probably the right move. Id probably go with the 16 core version tho, unless i was keen on seeing how the zen2 cores shape up in a year.

i doubt the 8 core threadripper would have much resale value, i imagine the market will be flooded with them when it comes time to upgrade.

i was getting 3.7ghz on my 1700 with the stock cooler. figured id be on this platform for quite some time and set myself up with some watercooling. Does 3.9 now and is super quiet.

Re:Awesome

[gosand]

Prior to my current processor (Intel Q8400 quad core) I was an AMD fan as well. Obviously, I haven't built a new system in a while, and although 8GB of RAM is still plenty for me (Mint 18.2 XFCE, thank you very much), I do tax the processor from time to time. So an upgrade - scratch that, new system build - may be in order. I am looking forward to seeing where this all goes.

Re:Awesome

[Billly Gates]

The motherboard is expensive as it is designed with NUMA support and other enterprise level hardware. The threadripper is a remarked server grade Epyc CPU line which will compete with Xeons.

The 1900X is not worh the extra 5% improvement over an 1800x which will be much cheaper if you include the $50 CPU savings and the cheaper motherboard. However the x1900 will have guaranteed Linux support which some of the Ryzens are experiencing bugs though.

Re:Awesome

[Billly Gates]

That is Intel propaganda. They are not glued Ryzen cpus no more than Xeons are glued i7s ... well maybe the rushed i9 is :-D

The Threadripper is based on Ryzen but has NUMA memory and deeper stage pipelines and caching for things like Linux support which is not impacted by the bug of the cheaper counterparts. It has more cache and also supports quad memory channels which is the bandwidth you talked about.

Also the regular Ryzen does use less power than an i7. I just wanted to point this out as people still say this and think you blow something like $100 a month in power (laughable false but I see that in fanboy posts on youtube).

Re:Awesome

[CannonballHead]

$100 a month in power is a bit extreme.

However... if you shave 100W off your usage, and you leave your computer on 24/7, and you're in one of the slightly higher tiers in California, you actually can easily get $20+ per month in *savings*. A 300W thing running 24/7 at 30 cents per kWh is something around $60/mo. It does add up pretty quickly in states where power isn't relatively cheap.

Re:Awesome

[Anonymous Coward]

Even if the difference is 5 dollars a month that's 60 dollars a year over 5 years (most people replace a system somewhere around 5 years) that is a fair bit of change.

If the difference is 20 dollars a month that starts getting into whole new system territory over a 5 year period.

At 60 dollars if I lived someplace like that for sure I would be watching how much electricity everything in my house used.

That being said I live in one of the area's where the difference is probably like 5 dollars a month between the low end CPU's and the high end CPU's so I buy towards the middle (i5 Skylake) they play my games fine and when I need to get some work done they wont keep me waiting too long.

Re:Awesome

[Khashishi]

Who besides miners run 24/7?

Re:Awesome

[willy_me]

If you purchase a "budget" Threadripper then you also require the appropriate motherboard and memory - assuming you plan on making use of the extra memory bandwidth. All of this costs a significant amount - why not pay a little bit more and get the extra 50% or 100% thread count? Alternatively, pay significantly less and get the Ryzen 7 series motherboard, memory, and CPU.

This low end Threadripper does not appear to have much of a market. It is tightly squeezed by the higher end parts and the lower end Ryzen. I suppose it makes sense for AMD to sell Threadripper CPUs with some broken/disabled cores at a reduced rate - someone will buy them. But I see little value in purchasing such a product. It would have to sell for 30% less to make up for the increased motherboard costs.

Re:Awesome

[Kjella]

Threadripper de-lidded. Two are filler blanks, two are the 4+4 cores of the 1900X. Idiots and idiot mods, this place has really lost it.

Re:Awesome

[Anonymous Coward]

Those devices tend to be idle 99% of the time. TDP is almost completely irrelevant for their actual power use.

Re:Awesome

[smallfries]

Why do they do that way instead of scraping the bottom of the bin and putting out 2+2+2+2?

Re:Awesome

[kactusotp]

Wow it use to be (yes I'm that old) a case on slashdot that people would at lease do a cursory google before opening their ~~mouths~~ reply box http://lmgtfy.com/?q=threadrip...

Re:Awesome

[ncc74656]

Who besides miners run 24/7?

If you have a media server or similar gadget at home, you more than likely leave it running so it can download TV, serve up your address book to your devices, etc.

I also have some Raspberry Pi 3s plugged into TVs for media access that I leave running all the time, but their power usage is low enough that I don't particularly care how much they add to the bill (some quick calculation based on 12.5W maximum power draw says they cost me less than a dollar each).

Re:Awesome

[CannonballHead]

People who don't like turning their computer off. Like me :) Not 100% sure why, but having to re-open stuff, boot VMs, etc., tends to factor into it. Plus, I run a plex server on my desktop. But I have a somewhat low-power setup, max power draw is something like 60W, if I remember correctly.

Re:Awesome

[Khashishi]

Just put it to sleep.

intel pci-e io sucks!

[Joe_Dragon]

intel pci-e io sucks!

Re:intel pci-e io sucks!

[K. S. Kyosuke]

Old MacIntosh had a card

P-C-I-e-IO...

Re:intel pci-e io sucks!

[Joe_Dragon]

enjoy your TB3 be-hide the DMI link that also has storage + usb + network on it as well.

8 core vs 4+HT?

[Gabest]

Need to see benchmarks of course, but I wonder how much better it is than a i7-x700, which costs $200 less and has better single core performance.

Re:8 core vs 4+HT?

[aliquis]

A lot better for multi-threaded tasks.

What will be more interesting though is how the i7 8700K will do vs the Ryzen 7 1700(OC) or 1700X though.

The i7 8700K will still be somewhat better in single-thread performance than the 7700K but will also have 6 cores so general multi-tasking performance will become closer to the Ryzen 7 though it doesn't seem to really catch up, it's very close to the Ryzen 7 1700 but that's a lower clocked low TDP variant, under similar settings as a 1700X or 1800X it can give similar performance so that's likely what it should be compared with anyway because I guess that's how those who need the performance will use it anyway.

The 7700K I would keep away from since the 8700K is so close and Ryzen offer so much better multi-core performance.

Re:8 core vs 4+HT?

[ravenshrike]

Bullshit. Single thread performance is equal to the 7700k when normalized for memory speed. So unless you're buying this as a workstation and thus not using higher ram profiles, there is no difference.

Re:I am a processor fan (the electromechanical kin

[K. S. Kyosuke]

Don't be silly, they're not selling IBM's entry level mainframes...

Wow!!! Trusted Platform Module 2.0 !

[Anonymous Coward]

And it's a FEATURE. Gosh, I'm so happy and thankful to relinquish control over how the thing works after paying you for the privilege. I'm glad that you spend time designing ways to deliberately cripple its functionality instead of making it stable and reliable. Who wants stable and reliable anyway? No, what I want is a chip that has a bunch of sneaky shit embedded inside it, and consumes power checking my decisions, and preventing me from doing things I might want to do.

Re:Does it boot Windows 7?

[K. S. Kyosuke]

I'm wondering...isn't it possible to run Windows 7 on "unsupported" hardware through virtualization, possibly even paravirtualization? I suspect the VM-level situation shouldn't change, except for maybe user-level instruction sets.

Is there enough market demand for workstation CPUs

[rsilvergun]

and I don't mean general purpose CPUs that do double duty but something like this that really is meant for the workstation. If you're just a gamer these CPUs are worse than useless. They sightly under perform the top general purpose CPUs at twice the price. I know time is money, but there's diminishing returns (anyone remember 52x CDR?)

Yes

[ArchieBunker]

Yeah you can even do video passthrough and use 7 inside a hypervisor like a regular desktop. From what I've read there is a slight performance hit so gaming would not be the optimal use.

Re:Yes

[BigDaveyL]

People like myself are gaming inside a VM with pretty good luck.

AMD's EPYC Spyware!

[Anonymous Coward]

GuardMi and SenseMi and TPM 2.0 are all scams to give you a false sense of security while they leave the feds a 'key under the mat' in the form of a crytographically secure 'Platform Support Module', analogous to Intel's ME/AMT technologies, that can access everything in your system with only a few very difficult methods of logging and discovering the activity. All of which require a second uncompromisable system sitting between your AMD/Intel system and the internet, sniffing over every packet and hoping they were sloppy enough to allow discovery.

Furthermore, unlike the Intel hardware the AMD PSP hasn't been proven to have a method to disable the firmware, which combined with the statement that neither minimal firmware for bootstrapping, nor auditable firmware for security analysis will be available makes them just as bad if not worse than Intel at this point in time.

Remember folks, if you're supporting either Intel or AMD in the desktop/notebook space, you're funding the permanent reduction of your digital rights. Just like with Cell Phones.

Remind your friends and ask them 'If you have nothing to hide, then you wouldn't mind me publishing those sexting pics to the internet, would you?'

Re:AMD's EPYC Spyware!

[Anonymous Coward]

I totally agree with you here, but what options does your typical consumer really have? It's either "run old hardware" or "run slightly gimped, expensive hardware". Purism seems to be one of the only people trying to divorce the CPU from the management chip. The catch is, until they reach an economy of scale with their line of machines, they're over $1000 for the baseline models, and they don't appear to offer financing. I'd buy one in a heartbeat if I knew I was getting a "free as it gets" machine and could finance it with a reasonable interest rate. Hardware killswitches are a fantastic idea.

That's essentially what needs to happen: Purism needs competitors who care about the security and privacy of their customers. Most OEMs don't give a fuck.

Re:Is there enough market demand for workstation C

[jwhyche]

and I don't mean general purpose CPUs that do double duty but something like this that really is meant for the workstation. If you're just a gamer these CPUs are worse than useless. They sightly under perform the top general purpose CPUs at twice the price. I know time is money, but there's diminishing returns (anyone remember 52x CDR?)

Then again they are not being marketed for gamers. I believe they are being marketed for high end cpu intensive applications like cad and 3d animation software, such as blender. Still, I'm kind of scratching my head there. I'm not sure how big that market would be.

When I built my new i7 system I opted to step down from 8 to 4 cores. From what I have been reading most game engines are optimized for 4 core intel based systems any way. Most games I get over 70 fps on a 4K gaming rig so I think it was good investment.

As for using extra cores for blender, I really don't see the point there. I use blender and like most serous blender users that I know we use the cycles engine and pawn the rendering off to the gpu. To me, it was a better deal to step down the cores and spend the extra $$$ on a better graphics card.

Not the threads you were thinking of.

[Gravis Zero]

A little background here. "Threadripper" is a reference to AMD swiftly removing the threads that hold Intel's coin purse together. ;)

Re:Is there enough market demand for workstation C

[ckatko]

THIS JUST IN.

CPU marketed for amateur servers NOT PRICE EFFECTIVE FOR GAMES. ::mind explodes::

Re:Is there enough market demand for workstation C

[yodleboy]

Who doesn't want more performance at lower price? I can't be the only person that does more than game on their PC. I just got a Ryzen 7 1700x, maybe not a true workstation CPU, but 8c/16t would have been seen as such not that long ago I guess. I'm gladly giving up a small amount of gaming performance for massive gains in things like video encoding for far less than Intel would have cost me. Just being able to run handbrake on 8 threads and some game on the other 8 and have BOTH perform well at the same time is nice.

AMD desktop boards more pci-e then intel

[Joe_Dragon]

AMD desktop boards more pci-e from CPU then intel. How much did intel pay for that post and that raid key only works for DMI switched ones.

On intel desktop boards you have a lot of stuff stacked on the DMI bus. And the intel Enthusiash ones are min cost $1000 to get 44 lanes.

AMD Enthusiast chips have 64 and servers have 128.

Re: AMD desktop boards more pci-e then intel

[Joe_Dragon]

and you can also get an cheap pci-e x4 to m.2 slot board.
https://www.amazon.com/Rivo-PC...

https://www.newegg.com/Product...

For data science, not gamers

[werepants]

The 1900x is looking to be THE go-to processor for data science needs, or really anybody who is serious about massive multithreading or deep learning. You've got more PCIE lanes than anything else on the market, to support multiple GPUs with high-bandwidth x16 PCIE interfaces. Single-thread performance is almost never the bottleneck here, so really what you want to optimize on the maximum number of PCIE lanes per dollar, and this cheap Threadripper wins by a mile.

Time to start prepping the Newegg wish list. And convincing the wife that the bare minimum for this system is a Geforce 1080 Ti... or 4.

amd needs an server range TR chips that can do wit

[Joe_Dragon]

amd needs an server range TR chips that can do for people who don't need the full EPYC build out.

But the low end 1 cpu epyc system can make for a nice all flash ZFS node with some 10TB or higher nics.

Or even some CEPH nodes.

Will stay away until the "segfault" bug is settled

[ffkom]

As much as I appreciate the price/performance of Ryzen based CPUs and the competition it has sparked, I will stay away from any Zeppelin-Die-based CPUs until AMD provides a proper explanation and fix for the "gcc segmentation fault" bug that haunts Ryzen CPUs.

After many months of not admitting any bug existence to the affected users, AMD finally admitted there is one, yet they neither recalled affected CPUs nor do they tell us how to distinguish affected from unaffected CPUs - so even if you buy a Ryzen today, you can still buy one unsuitable for ordinary gcc compilation tasks.

Given that they cannot (or do not want to) say which CPUs are affected, and given that AMD did never explain a root cause of this bug and how it is fixed, I do not believe they actually have fixed it or know how to fix it. Even CPUs that were manufactured in calender week 25 of 2017 have turned out to be affected!

Why should I believe they fixed this for "Threadripper"?
Sure, they know by now how to test individual CPU exemplars for the bug, and might deliver unaffected ones to the press for reviews. Does that tell me they will do the same testing for the exemplars delivered to the mass market? No.

More information on this bug via https://forum.level1techs.com/... and https://community.amd.com/thre...

Re:Will stay away until the "segfault" bug is sett

[Gravis Zero]

I will stay away from any Zeppelin-Die-based CPUs until AMD provides a proper explanation and fix for the "gcc segmentation fault" bug that haunts Ryzen CPUs.

Umm... they already did both of those.

I do not believe they actually have fixed it or know how to fix it.
Why should I believe they fixed this for "Threadripper"?

Well, there's no cure for being an asshole, so you are shit out of luck.

Re:Will stay away until the "segfault" bug is sett

[ffkom]

Their production going forward has solved the problem.

Can you provide a link to any credible source stating this - or the nature of the change to the production process?
And if this was true, then why has AMD returned still-defective CPUs even to people who RMA'd because of exactly this bug in the last few weeks?

They aren't doing a recall because it would bankrupt them.

That does seem plausible, but personally I don't want to be involved in an RMA-until-you-get-lucky game.

Re:Will stay away until the "segfault" bug is sett

[ffkom]

I will stay away from any Zeppelin-Die-based CPUs until AMD provides a proper explanation and fix for the "gcc segmentation fault" bug that haunts Ryzen CPUs.

Umm... they already did both of those.

No, they did neither. Please provide a link to that proper explanation and description of the fix you say exists.

What AMD currently does is rather "have some poor guy test N CPUs to find some non-buggy exemplars amongst them and ship those to RMA demanding customers". And as one can see from the reports of affected people, even recently delivered RMA-exchange CPUs (manufactured in CW 30) were still affected: https://docs.google.com/spread...

Re:Will stay away until the "segfault" bug is sett

[Billly Gates]

Thread ripper is not effected as it was tested with Linux as it is based on the Xeon competitior based Epyc. Ryzen I heard was fixed but I do not know. AMD stated it did little testing with linux as it assumed only Windows users would use Ryzen but Ryzen2 would have thorough testing like the threadripper ones which are server/workstation based.

I too would avoid AMD this round as much as I want to cheer them. I will cut them some slack as the Ryzen architecture is brand new and a redesign from scratch. Remember the Core2Dous were blasted by Theo from OpenBSD who listed a large errata with them. If you send an assembler instruction to store data and use graphics the CPU would crash so you would have to write the kernel around the bugs.

I may consider AMD still in the future. The question is was the bug fixed and is hardware related or manufacturer related. Can a microcode update fix it?

Re:Will stay away until the "segfault" bug is sett

[Billly Gates]

That is not being an asshole. If I blew $900 for a new motherboard/CPU/DDR 4 ram combo you bet I would want blood if what I bought turned out to be bugged and useless.

The customer is always right.

For that kind of money it is reasonable to want thorough QA and a quick RMA by AMD if any last minute bugs did make it.

Re:Will stay away until the "segfault" bug is sett

[ShoulderOfOrion]

Same here. As Gentoo is my primary desktop I'm staying away from these latest AMD parts until the bugs are sorted out.

Re:Will stay away until the "segfault" bug is sett

[Gravis Zero]

Then you are in luck because they have been doing RMAs for anyone that has run into this extreme corner case.

Re:Make this simple for me.

[Archangel Michael]

Look for IOPS and sustained transfers.

That's why people have completely ditched spindle drives for CPU intensive work, and why all the SSD technology is working towards expanding the bus and higher IOPS. Fast Spindle drives are in the range of 300-800 IOPS, and SSDs are past 200,000 (upto 1,000,000 read IOPS in extreme cases).

Re:Is there enough market demand for workstation C

[jwhyche]

LOL at "investment". How much money does that 70 fps gaming rig earn you?

Not as much as it usually does. Only $95 the month before last, and a little over a $100 and some change last month. Not even enough to cover the cost on renting my homestead. But that isn't surprising marketplace sales are kind of slow this time of year.

I could have picked up some slack doing some Dj work or picking up some custom scripting contracts but I really didn't feel like messing with it. Things should pick up as fall starts, but to be honest, its going to be hard to clear $300 bucks. I have to get some new products up on the marketplace.

Wait. You where being a smart ass wasn't you? I bet it never dawned on you that you can use a kickass gaming rig to earn real money inside a game?