AMD Announces Ryzen 3000 Pro

DejaWiz

Fully [H]
Joined
Apr 15, 2005
Messages
21,819
Looks like AMD has kicked out the PRO series of Ryzen:

https://www.amd.com/en/press-releases/2019-09-30-amd-announces-worldwide-availability-amd-ryzen-pro-3000-series-processors



Ryzen3000Pro.png
 
I've never quite understood the purpose of the Pro editions of Ryzen. They seem pretty much the same as the non-Pros.

Personally I am hoping for 45w and below Ryzen 3000 parts at some point.
 
it just use less power I guess. so more energy saving for business when not in use?
 
I believe the Pros are lower clocked/lower TDP parts rated for high longevity/low temps under load. They are aimed at OEMs who value reliability over performance. At least to my understanding...
 
I believe the Pros are lower clocked/lower TDP parts rated for high longevity/low temps under load. They are aimed at OEMs who value reliability over performance. At least to my understanding...
What is this "longetivity" buzzword people throw around. Sounds like Pro iNtel Propaganda bovine stercus.
When was the last time anyone here killed a cpu running at stock speeds.
 
I believe the Pros are lower clocked/lower TDP parts rated for high longevity/low temps under load. They are aimed at OEMs who value reliability over performance. At least to my understanding...
I thought the PRO version has more security features for the enterprise setting.
 
More security as well as far less variance on given results .. my understanding

seems for the most part (look at Nv and recent debacle on very $$$$$ pro cards NOT being verbatim results as damn well should be)

one little example I can think of, GPU that use GDDR5 can have all out ECC (provided motherboard etc also supports) for the above outlined reason. i.e am pretty sure a financial institution, power companies and the like absolutely require as close as possible to prevent error from code crashing / destabilization of given results all that stuff.

just going by read knowledge

the few GPU that were true identical save for the GDDR5 all out ECC enabled or only the simple sanity checker that consumer stuff uses (when it does use at all)
the result for the GDDR5 ECC kicking in or not can be tricky, though to keep the explain simple, ECC kicks in to prevent crashing of the kernel memory (running code) it does so I believe by and/or control volt / speed tables .. i.e it (GPU /OS) detect the GPU in this case being pushed outside it's range (overclocking and the like) so it puts "halt" commands to "slow down" the errors.

all out security etc is great, but it does come at cost, often consumer vs non (Pro in this case) usually the PRO do NOT hit the same peak clocks etc (to maintain the ECC concept)

----------------

My understanding.. so far, I noticed beyond rare exception(s) that is usually the difference pro or non...more pricey, but you are paying for "works without fail" ... that being said, I have read on many Pro versions have just as many crash issues etc, which begs question. why pay more, potentially get less IF bulletproof stability is NOT #1 requirement.
 
PRO for this iteration appears to be lower power slightly fewer clocks.

If I can give my CAD workers and other similar developers a workstation with a 16 core CPU with 32 threads designed for lower TDP and top of the line performance with a RTX 6000 CPU then why wouldn't I? It's going to be performance parity for a LONG time to come. Easily 5 years before I'd need to replace the hardware again unless some new feature drops that will make everything better for my workers. And my cost of ownership is lower because of the year on year refresh cost is lower. Not to mention if performance on the CPU front gets a lot better I have a solid chance to have my vendor spend a weekend swapping CPU's and I'll have upgraded systems for minimum cost.

These are all win win and without needing to invest in a beefier more power hungry and more expensive Threadripper setup.


Now sure the bigger top of the line workstations will have the big 32 core threadrippers and dual Quattro cards... but that's to be expected. ESPECIALLY if it can increase the productivity of my resources.

I personally know a guy that does bridge design and architecting. He's specialized... and brings in over 500 bucks an hour as a consultant. If I can get him a 10, 20, or heck 30k workstation that improves his performance by 10% that 10% is worth 100k a year to me.

As long as I'm not making them efficient to the point of sitting idle collecting a check I'm a happy camper.
 
I thought the PRO version has more security features for the enterprise setting.
https://www.amd.com/en/technologies/guardmi

Powerful Protection
Every AMD Ryzen PRO processor contains a powerful, built-in security co-processor that runs AMD GuardMI technology, which is dedicated to helping protect users’ PCs. AMD Memory Guard5 helps defend against cold boot attacks with full system memory encryption, and AMD technology supports and complements OEM security features and Windows Security, including Lenovo ThinkShield™ and HP Sure Start™, among others.
 
PRO for this iteration appears to be lower power slightly fewer clocks.

If I can give my CAD workers and other similar developers a workstation with a 16 core CPU with 32 threads designed for lower TDP and top of the line performance with a RTX 6000 CPU then why wouldn't I? It's going to be performance parity for a LONG time to come. Easily 5 years before I'd need to replace the hardware again unless some new feature drops that will make everything better for my workers. And my cost of ownership is lower because of the year on year refresh cost is lower. Not to mention if performance on the CPU front gets a lot better I have a solid chance to have my vendor spend a weekend swapping CPU's and I'll have upgraded systems for minimum cost.

These are all win win and without needing to invest in a beefier more power hungry and more expensive Threadripper setup.


Now sure the bigger top of the line workstations will have the big 32 core threadrippers and dual Quattro cards... but that's to be expected. ESPECIALLY if it can increase the productivity of my resources.

I personally know a guy that does bridge design and architecting. He's specialized... and brings in over 500 bucks an hour as a consultant. If I can get him a 10, 20, or heck 30k workstation that improves his performance by 10% that 10% is worth 100k a year to me.

As long as I'm not making them efficient to the point of sitting idle collecting a check I'm a happy camper.

Except Ryzen is absolute trash in the most popular 3D CAD program out there - SolidWorks.
 
That's odd, solid works does not say anything bad about Ryzen.
https://www.solidworks.com/partner-product/ryzen-processors

You can get the value if you're doing multi-core stuff like simulation and rendering, but general modeling they lose. Which sucks. Because that's 90% of what I do.

But even benchmarks be damned, it just feels smoother on my Intel rig at work vs my Ryzen rig at home. (granted I have a K620 at work)
 
You can get the value if you're doing multi-core stuff like simulation and rendering, but general modeling they lose. Which sucks. Because that's 90% of what I do.

But even benchmarks be damned, it just feels smoother on my Intel rig at work vs my Ryzen rig at home. (granted I have a K620 at work)

Yup. General modeling work is necessarily single threaded, so fastest core speed is king.
 
The massive performance gains of yesteryear are now a thing of the past for both AMD and Intel. If the performance gains were still there, a 9900K would be a quantum leap over a 2600k and it just isn't. Significantly faster, for sure, but not the "ZOMG WHOLE NEW BALL GAME!!" leaps like we used to get between chip generations (8086->286->386->486->Pentium->Pentium 2, etc).
 
What is this "longetivity" buzzword people throw around. Sounds like Pro iNtel Propaganda bovine stercus.
When was the last time anyone here killed a cpu running at stock speeds.

My hexacore 3930k has been hot and bothered at 4.6-4.8Ghz (depending on the cooling I had installed at the time) since 2011.

I'd argue that temperature based longevity really isn't an issue for CPU's. At least not in the time period before they are thoroughly obsolete.

Now, the power phase on the motherboard, however, is a different story.
 
I've never quite understood the purpose of the Pro editions of Ryzen. They seem pretty much the same as the non-Pros.

Personally I am hoping for 45w and below Ryzen 3000 parts at some point.

Well, technically the 3400ge and 3200ge are in your wants.
 
The chart says and ryzen 5 pro 3400ge

It does indeed, but it is nevertheless a Zen+ part. The 2000 'g' series APUs likewise use original Zen cores and not the Zen+ cores the rest of the 2000 series uses. It is confusing, but true. If I had to guess I'd suggest they do it this way to not make the APU's look old since they release so much later than the non-APU chips.
 
Well, technically the 3400ge and 3200ge are in your wants.

I believe those are not Zen2 based. They use the older Zen+ arch, despite their 3xxx naming.

Besides, I am not looking for an APU part. Ideally it would be a slightly lower clocked hexacore part that fits inside 35-45w with no GPU.

I built my fiance a passively cooled system in a Streacom case with a 2400G, but I'm having nothing but trouble with that APU. (Intermittent hard freezes, only able to support two monitors, etc)

The plan is to replace the APU as soon as lower power Zen2 chips are available and instead use a passively cooled GT1030 for video output.

The case claims to handle 65W TDP CPU's, but it already gets pretty hot with the 2400G in 35W TDP mode (set in BIOS) so I don't think I'd want to go over 45W at most. I could get a Ryzen 5 3600 and try to manually underclock/undervolt it, but I'd rather have an official AMD product.
 
When you wait that long for Ryzen and it's single threaded performance is still 2nd best? Yeah. Stuck on an i7 6700. I'm bitter. !!

It's real hard to justify a $500 cpu that'll need a new PS, case, memory, etc to the boss for +25% performance. Over a 4 year old rig.

Why do you think you'll need new PS, Case, memory over the 6700? Both systems support DDR4.
I can make the same arguments against the 9900K over the 6700 as well.

25% performance (or 30-35% if you go Intel) should pay for itself pretty quickly. Either way, it doesn't justify your vitriol.
 
It does indeed, but it is nevertheless a Zen+ part. The 2000 'g' series APUs likewise use original Zen cores and not the Zen+ cores the rest of the 2000 series uses. It is confusing, but true. If I had to guess I'd suggest they do it this way to not make the APU's look old since they release so much later than the non-APU chips.

In no part of my post or the post I quoted stated anything about zen, zen + or zen 2.
 
In no part of my post or the post I quoted stated anything about zen, zen + or zen 2.

Your original quote stated he was looking for low power Ryzen 3000 parts. When most people say they are looking for Ryzen 3000 parts, they explicitly mean the Zen 2 parts (as he confirmed in a later post in this thread). While you are * technically * correct in that the new APUs are both low power and labeled "3000" it was fairly obvious that is not what the OP was looking for at all.
 
When you wait that long for Ryzen and it's single threaded performance is still 2nd best? Yeah. Stuck on an i7 6700. I'm bitter. !!

It's real hard to justify a $500 cpu that'll need a new PS, case, memory, etc to the boss for +25% performance. Over a 4 year old rig.

Flawed argument.

No system ever needs to be "the best". It's not an epeen competition.

It just needs to meet the requirements for how you use it.
 
What is this "longetivity" buzzword people throw around. Sounds like Pro iNtel Propaganda bovine stercus.
When was the last time anyone here killed a cpu running at stock speeds.
I don't think the CPU would get killed quickly by that, but the capacitors and VRMs on cheaper/lower-cost OEM motherboards might, so a slower clocked version for businesses and enterprises makes sense, at least in terms of longevity.
Also, it would cut down on power and heat, which when running hundreds to thousands of laptops and/or desktops, would make a massive difference in power consumption and HVAC costs.
 
When you wait that long for Ryzen and it's single threaded performance is still 2nd best? Yeah. Stuck on an i7 6700. I'm bitter. !!

It's real hard to justify a $500 cpu that'll need a new PS, case, memory, etc to the boss for +25% performance. Over a 4 year old rig.
The 2700X running at stock that replaced my 6700K runs circles around it in every category.
Even pre-patched, the 6700K would have barely been half the performance, and post-patches, it was about 1/3 as powerful.

Even a 3600X would be a massive boost in performance, in both single and multi-threaded applications alike.
 
I think the HP B550 leak, Pro cpus, and something else just = boring reliable OEM boxes.
It’s not a bad thing for AMD to compete in this space.
I think it’d be amusing to see a 16core + pcie4 nvme drive packing “creation” box if the OEMs mandated a passively cooled x570 mobo.
 
Ryzen Pro would be comparable to Intel processors with vPro. While Intel doesn't make a specific model designating vPro compatibility, you have to do your research as an individual to know if you have it. However, if you are going for vPro, you are simply purchasing complete systems from HP or Dell, and you tell them you want vPro, and they'll ship you systems with the compatible processors. AMD is making their version of security and management a separate CPU line, so it doesn't get conflated with the consumer line.

In addition, the Ryzen Pro processors have a longer guaranteed support cycle, so customers who buy in won't have to worry about no parts/support in 2-3 years as they may with consumer line electronics.

I think (just going off my memory) that the platform for Ryzen Pro must be ECC compatible. While regular Ryzen processors are ECC compatible, the platform they are used in does not have to be.
 
Yup. General modeling work is necessarily single threaded, so fastest core speed is king.

A what? I'm not sure what kind of graphics software you are using sir. Its definitely late in development. 3dsmax, particle illusion, poser already supported preview, and rendering on multiple cpu's since Pentium 2 came out.
Dual CPU's were game changer especially on preview screen.


@ontopic
The pro series cpu's are only being sold in specific computers sold by vendors like HP. Its not being sold alone. The Ryzen 1700 pro and 2700 pro had much better turbo clocks, and could oc further. (you can view them occasionally in boinc projects, and wonder why they are killing your non-pro overclocked part.) Owners of 2700 pro, please delid and check if there's actually extra chip/chiplet there. I doubt.
 
Last edited:
I built my fiance a passively cooled system in a Streacom case with a 2400G, but I'm having nothing but trouble with that APU. (Intermittent hard freezes, only able to support two monitors, etc)

The plan is to replace the APU as soon as lower power Zen2 chips are available and instead use a passively cooled GT1030 for video output.

The case claims to handle 65W TDP CPU's, but it already gets pretty hot with the 2400G in 35W TDP mode (set in BIOS) so I don't think I'd want to go over 45W at most. I could get a Ryzen 5 3600 and try to manually underclock/undervolt it, but I'd rather have an official AMD product.

or maybe just get a descent but quiet air cooler (like a zalman cnps series) or like you were saying, a dedicated gpu to cut down on the heat output from your cpu. and like they were saying above that 2400G is actually using regular zen1 cores not zen+. so in that case try loosening your memory timings / speed because zen1 wasn't a very good memory clocker. you shouldn't be getting hard locks, a lot of times it's a memory issue.
 
it just use less power I guess. so more energy saving for business when not in use?

well less watts = less heat = better overclock

3700 pro has same clocks as 3700X but uses less watts to get there. i am curious to see how well these things will OC!!!! :nailbiting:
 
well less watts = less heat = better overclock

3700 pro has same clocks as 3700X but uses less watts to get there. i am curious to see how well these things will OC!!!! :nailbiting:

Traditionally I would agree, but it would seem like as we are getting smaller and smaller process sizes we are more likely to hit a hard process wall on clock speeds, and less likely to be thermally limited like in the past.

At least that seems to be the case with Zen2
 
A what? I'm not sure what kind of graphics software you are using sir. Its definitely late in development. 3dsmax, particle illusion, poser already supported preview, and rendering on multiple cpu's since Pentium 2 came out.
Dual CPU's were game changer especially on preview screen.

We were talking SolidWorks 3D CAD. It really doesn't care how many cores there are when you're doing 3D modeling/open/rebuilds. It's mostly single threaded.
 
The 2700X running at stock that replaced my 6700K runs circles around it in every category.
Even pre-patched, the 6700K would have barely been half the performance, and post-patches, it was about 1/3 as powerful.

In SolidWorks? Or general?
 
Back
Top