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

AMD only released the chip initially for 10, but backed up support to 7 (the most used windows version), after 10 was not ready when it was released. People were very happy to see that support.

 

I think we will continue to see that support, at least for the near future. If not, I'm out of the Windows desktop scene until they pull their head out of their orfices.

Sent from my SM-G935T using Tapatalk

 

AMD Announces Enmotus FuzeDrive technology to Speed Up Ryzen-based Systems
"AMD today in a blog post announced the fruits of its partnership with Enmotus, a mainly enterprise-focused company that has made its name in creating performance-optimizing software solutions."

"L; DR: Essentially, the performance increase doesn't come from activating the FuzeDrive software on the Ryzen system; but from both activating it and adding a Samsung 950 Pro SSD to the system. How much of the performance improvement can be attributed to the added NVMe drive alone is a valid question, and we'd wager it's most of it. You can purchase a license to Enmotus' FuzeDrive for your AMD system for $19.99."

 

Ryzen Mobile Gaming, What Can You Play? [Part 1] PUBG, Fortnite & More

 

Windows 10 Meltdown-Spectre patch: New updates bring fix for unbootable AMD PCs
AMD PCs can now install Microsoft's Windows update with fixes for Meltdown and Spectre and the bug that caused boot problems.
By Liam Tung | January 18, 2018 -- 14:02 GMT (06:02 PST)
http://www.zdnet.com/article/window...new-updates-bring-fix-for-unbootable-amd-pcs/

More details in this thread:

CPU Vulnerabilities, Meltdown and Spectre, Kernel Page Table Isolation Patches, and more
http://forum.notebookreview.com/thr...patches-and-more.812424/page-63#post-10666767

 

The latest update for W10 brings back speed and stability across all CPUs. W10 is very fast or maybe it is a placebo.

 

Ryzen 7 Guest Build, An Old Face Returns

 

VRM temps shot through the roof.
I did see his Devil Canyon build and it had SLI Titan X(Maxwell).

 

Isn't last update most intended for Micro$haft's AMD screw up? I can smell placebo Check out ASSSD and CrystalDiskmark vs. what you got before. Especially the 4K score.

 

Before patch:

Sequential Read (Q= 32,T= 1) : 1509.939 MB/s
Sequential Write (Q= 32,T= 1) : 307.999 MB/s
Random Read 4KiB (Q= 8,T= 8) : 1004.784 MB/s [ 245308.6 IOPS]
Random Write 4KiB (Q= 8,T= 8) : 311.257 MB/s [ 75990.5 IOPS]
Random Read 4KiB (Q= 32,T= 1) : 499.305 MB/s [ 121900.6 IOPS]
Random Write 4KiB (Q= 32,T= 1) : 310.653 MB/s [ 75843.0 IOPS]
Random Read 4KiB (Q= 1,T= 1) : 44.874 MB/s [ 10955.6 IOPS]
Random Write 4KiB (Q= 1,T= 1) : 174.811 MB/s [ 42678.5 IOPS]

Tech specs:
Model: PM951 Interface: PCI Express Gen3 x4 Form Factor: M.2 Density: 256 GB
Sequential Read
Up to 1000 MB/s

Sequential Write
Up to 280 MB/s
Random Read
Up to 250 KIOPS

Random Write
Up to 74 KIOPS

After update:

 

That's the problem with an open test bench - IDK why they don't supply cooling for them, they need it as much as the closed case if not more. No air movement = high VRM temps with nowhere to dissipate.

 

Almost 20% decrease in read and 54% in write 4K speed. Nice

Spoiler: What I feel...

 

I'm using native ms driver and not samsung driver.

 

Native or Samsung driver doesn't matter if performance is less.

Spoiler

I got up only a pict from your link. Here is the real deal

 

How did you add it?

 

Add .gif and copy and paste

 

https://wccftech.com/amd-2nd-generation-ryzen-5-2600-asus-crosshair-vii-hero-x470-leak/
So, compared to last gen, you have a 200MHz jump on the new process (not final). That means, 6% of the 15% is likely attributable to the increase in speed. That means 9% increase due to the process compared to last gen, from what is known. This also does not give an idea of ability to overclock. Also, this is NOT an X processor, so is expected to be lower just like the non-K variants from Intel. So before those comments roll in. A 15% on sisoft is pretty nice, and you cannot say it is JUST cinebench or a prime run, which I do want to know if prime performance was improved. Further, we cannot extrapolate from the 1.1GHz IMC the exact speed of the ram, as when I ran it with 3600 in Nov., I had 1.46GHz. That means maybe 3200MHz on ram, but possibly lower than that with lower latency or some other speed control or gearing mechanism. Because of that, of the remaining 9%, we do not know the amount attributed to the new process, the tweaks to memory, or the tweaks to the uarch. I am assuming that it was fully loaded on the multi-threading and that it settled at its stock or the amount for all core above stock, similar to 1600 scores. Lots of assumptions. But 15% in a single gen, and looking at least that much, if not more, with Zen 2, they should be able to hit 30% over Zen with Zen 2 7nm, if not the 40% previously mentioned. This is good news, especially with not getting hit as much on VMs and I/O as Intel. But, let's hear your thoughts....

 

I was expecting higher clock speeds based on what people here were saying, more along the lines of the 4.5Ghz of Intel. When does Zen 2 7nm come out?

 

So did I.
The 12nm LP is per technical specs designed for high clocks and high performing parts... the previous/existing 14nm LPP is not.
200 MhZ bump is not even a 10% increase... at best it's a 6.5% bump in clocks alone.

Then again, it WAS hinted that this could be an engineering sample... we will see what the final product ends up with at the official release (as we had a lot of speculations about other Ryzen parts and early benchmarks that turned out to be wrong or faked)... and besides, this is a non X version of the Ryzen 6 core cpu... plus, AMD might have revised the TDP specs for the refresh... but we hadn't heard anything about that yet.

Still, we were told at least 10% increase in comparison to 16nm process from TSMC (which is already designed for high performing parts)... and there's 15% higher chip density to take into account.
Something is definitely amiss... and this could be an ES (engineering sample), fake benches, or potentially really low yields (which would be surprising given what the technical specs implied).

I doubt AMD would be messing about with IPC increases for a simple refresh (as it would be out of character... but I guess it's not impossible if they were planning this for a while).

Zen 2 7nm seems to be slated for early next year.

 

First off, remember how much lower non-X variants clock. For example, the 1700 is a 3..0/3.7GHz chip, and the X variant is 3.4/3.8. The 1600 is a 3.2/3.6 chip and the 1600X is a 3.6/4.0GHz chip. This is a 3.4/3.8GHz chip for the 2600. [Edit: by an increase of 200, we are looking at 4.4GHz on the X series. Further, we don't know whether more will hit that amount other than golden chips or not. But going from 4.0 to 4.2, if it adds 200 regular on top and more chips can reach it, that is right near your 4.5GHz.]

Now, this is early silicon and we did see a difference last year.

Next, @Deks - reread my comment above, it shows 15% increased performance. NOT EVERYTHING IS SPEED. Also, the 200MHz is within the 65W TDP envelope. NONE OF THIS SPEAKS TO OVERCLOCKING ABILITY, TWEAKS TO THE POWER CURVE FOR THROTTLING AT DIFFERENT WORKLOADS due to changes in sense, or what the 95W variants will see.

Also, performance doesn't scale to chip density. Intel proved that multiple times and you can see that with Nvidia graphics cards also. More density, more heat, so efficiencies in refinement can be slightly retarded due to increases in heat. Use basic physics before pointing to such things.

Further, AMD said they made changes to SenseMI, to the IMC and memory latency, die shrink, power efficiency, and clockspeed and overclocking. We have seen 15% on early silicon, which 6% is only attributed to the speed bump, with the other 9% to all other changes to the CPU, which would include IPC. It is a 65W non-x chip. And there may have been a slight slowdown with redesigns to help protect against Spectre and Meltdown, which won't be seen in coffeelake until this summer, and it suggests that AMD pushed off the launch of this to April from beginning of March to deal with the redesign slowing it down, meaning they started looking at that, likely, once they were made aware in June.

Yes, Zen 2 7nm will be a more drastic change, but nothing suggested 12nm, which is a less than half node refinement, would be that much different than 14nm. It seems to still be made on Samsung's design, which is a more power efficient design, whereas 7nm is licensed from IBM's design, which is known to target 5GHz, not to say AMD hits that. So you seem to be missing a LOT! The bench DIRECTLY shows a 15% increase between the generations. Do you not understand that you multiply Instructions per Cycle (IPC) by the Speed? They are NOT the same. One says how many instructions can be processed in a given cycle, the other, speed, dictates the number of cycles in a period.

 

Given the lack of OC headroom in the first release of Ryzen perhaps AMD is only going to extend the stock numbers a bit, leaving the rest for user OC headroom?

If AMD uses 50% of the headroom, that means another 200mhz for OC, or 400mhz overall for the 1600 - probably across the board for all the 2xxx CPU's.

That would be a conservative 4.4ghz, and pushing the better samples into the 4.5ghz+ range.

It's all guesses for now, we don't know if the new process really offers any more headroom for performance, or if it's just more power efficient with a little more headroom that AMD used up when bumping the stock numbers. There may be no additional OC headroom.

Looking forward to release day + reviews

 

Its still early days though.
The product is as of yet unreleased, so we don't know conclusively where the clocks will end up, or if what we see in those benchmarks isn't false.

 

It would be nice if they followed AMD type Phenom releases. 1st Gen sucked at overclocking, but the second gen was fabulous for overclocking.

 

I always felt there was something of a bottleneck in a small part of Zen 1 causing the 4.2ghz OC ceiling, it doesn't seem to be an obvious power, temp, or volts limitation. If AMD have found and fixed/tweaked that part of the chip, OC headroom may be beyond a linear extrapolation from a pure frequency increase due to process.

 

Clock speeds and overclocking tend to rely quite a bit on the manuf. process a CPU/GPU is made on.
Look at Pascal. It's nothing more than highly overclocked Maxwell using a 16nm process designed for high clocks to begin with.
AMD has been using a low power process suitable for mobile and lower clocks, which is why Ryzen couldn't really break 4GhZ reliably or at all.

Considering this 12nm LP was described per documentation as a high performing process, it seems to be a complete opposite to 14nm LPP, at least in that it should allow high clocks.

We'll see how it pans out.

 

Strange, the 2500u gets better FPS in some games for odd reasons, could be fixed with drivers and / or game tuning. First game shows how the 2700u should outperform the 2500u in all games...

AMD Ryzen 5 2500U Vega 8 vs. AMD Ryzen 7 2700U Vega 10 Gaming Review. HP vs. Acer. Gaming Test


So that means these results shown in the following games for the 2700u could improve with driver updates, and / or system tuning.

Battlefield 1 AMD Ryzen 7 2700U Vega 10 Gameplay Benchmark Test

Battlefield 1 Multiplayer 64 on AMD Ryzen 7 2700U Vega 10

Overwatch AMD Ryzen 7 2700U Vega 10 Gameplay Benchmark Test

The Witcher 3 AMD Ryzen 7 2700U Vega 10 Gameplay Benchmark Test

Rise of the Tomb Raider on AMD Ryzen 7 2700U Vega 10

 

Gaming on AMD Ryzen 2500U Vega 8 Part 1. 20 Games Test. AMD Ryzen 2500U Review. HP x360

Gaming on AMD Ryzen 2500U Vega 8 Part 2. 9 Games Test. AMD Ryzen 2500U Review. HP x360

Gaming on AMD Ryzen 2500U Vega 8 Part 3. 12 Games Test. AMD Ryzen 2500U Review. HP x360

 

I can not see upgrading for only 15%. The TR 1950x gets CB R15 at 3500 or so now, to just get up to 4000 does not seem enough reason to drop $1,000 on the upgrade. Now if Zen2 gets 25% and a CB R15 of 4,400 or so that may be doable. Of course I would like to see more but I somehow doubt it.

Even at that the i9-7980xe overclocked will still have a performance advantage. This not including 24 months of silicon release from Intel between now and then too.

http://www.legitreviews.com/intel-core-i9-7980xe-18-core-processor-review_197903/10

 

Big Whoop, it's a crappy security failed performance hobbled Intel CPU with 2 more cores and 4 more threads, that costs 2x-4x as much to build and run.

I hope Intel enjoy's sitting on a pile of them at the end of 2018, that's what's gonna happen.

 

Thanks for reply. (I don't think your questions at the end there are meant for me, but if they are will just be met with the silence).

 

That may be true. Still the numbers are there for AMD to match. Even if it is not the performance crown I still want that Zen2 TR if it can overclock for the 4,400 CB R15 score.

 

No, AMD doesn't have to match the numbers.

Unless AMD releases an 18c/36t ThreadRipper for $2000, then AMD needs to match the numbers.

Otherwise AMD can keep doing what they are doing, with revision improvements - and keep their pricing as low as 1/2 that of Intel, and AMD will keep winning.

 

+Rep headed your way when I can give it to you again

Intel can enjoy the view of the performance crown while sitting on top of a mountain of them. Meanwhile they can also enjoy the multitude of class action court dates they will endure in 2018.
If the Ryzen 2700x will hit somewhere around 4.5, 4.6 ish with overclock, I will be happy.

 

I have to agree with @hmscott here. You are paying the price of an Intel 10-core and getting the performance of a CPU roughly with 2 extra cores. Also, 25% compared to what, Ryzen or Ryzen 2. Mathematically, 30% over Ryzen seems achievable, which is just under 15% of Ryzen 2 (15% of 115% is like 132%). Just looking at the process improvements, not to mention the uarch refinements, could theoretically get there. The design of it is finished, so it is minor tweaks from here out, and AMD should have a good idea if they will reach 30 or 40%. Considering them showing the beat of the 7-8% generational spread, and it growing, I'd assume they have potentially 30-35% performance on Zen 2 7nm, which for us with TR already would make for a good performance boost (although, you could always wait for the 7nm EUV lithography a year later which could be good, and that may be the last chip supported on the socket, although, at that point, you likely want a board upgrade for PCIe 5.0 and DDR5 which may have made it to consumers by 2020).

Also, Intel will release a weak 10nm+ product with small improvements, most likely, but on another socket in 2019. Cascade will be the pennacle of 14nm++, but they even said they will not get much improvement until 10nm++. As such, I think AMD will be just fine. Also, we do not know the slowdown of fixing meltdown in Intel chips, needless to say hardware changes to help with spectre. As such, Cascade could theoretically have small performance gains and just give performance with security, for a change. With that said, we do not know what dealing with spectre did to AMD chips as well, or if this 15% is with those fixes integrated considering it may be early silicon (but it could also be a market sample to finish products for market).

Either way, we'll know more soon enough.

 

Even if you go by the 7960x variant it still needs to get the performance in the range in order to not be a "Also Ran". AMD needs to become a performance leader to pull business away from Intel. They have made great strides with Ryzen, do not get me wrong here. It is just improvements need to still be made. I use the 7980xe as that is the mark that has been set.

The 15% seems to me more realistic, I in fact think we will be likely only to see 10% real world. As the process matures maybe much more down the line. Zen2 though has me much more excited. Intel 10nm is a dead issue unless massive amounts of time and process are invested to bet too 10nm++. This is because Intel took so much time and got so much from 14nm and AMD has not. Nor is it looking to stagnate at 12nm either.

Don't shoot the messenger here. Intel set the performance mark with the 7980xe, it is now up to AMD to defeat it, even if they went to the rumored 24c/48t TR's. TBH I would like to see a Zen2 Eypc core for the TR just running the 4.2 present speeds

 

Matt's New PC, Is The Bitfenix Enso a Toaster?

 

Ryzen Mobile Gaming [Part 2] GTA V, The Witcher 3, Battlefront II & More

 

THREADRIPPER is not EPYC - or is it? (en)

 

This might explain where the rumored Zen2 TR of 24 cores comes from. The TR could be using 3 CCX cores overall then

 

That is some good work he did there! I still would have loved to see a consumer or OCable 8-channel 128 PCIe lane chip or one overclockable chip out of the Epyc lineup, but more cores and cache @7nm would be good with me!

 

Thinking about it, they would compete with Epyc if there were a 32c/64t TR. It would make a lot of sense to go with a 24C/48t 3 CCX chip with Zen2. We could still see CB R15 scores stock at well over 4,500 and where 3,500 overclocked now maybe even over 5,300 points. Now that has me excited for a future upgrade!

 

I think now that the markets have differentiated, with the 128PCIe vs 64PCIe and ram channels, you could do a 32c/64t TR without cannibalizing sales of either. One has squared itself as a monster workstation (TR), while the other has established itself for cloud and other server uses in a big way. I think this gen, it would have caused more issues and people would have balked at the lower speeds with 32 cores, whereas, if HPC on GloFo is targeting 5GHz, getting a 32C to 4GHz+ suddenly is something the workstation would be interested in, rather than in the 3GHz range. So, I do think it was to prevent confusion and people trying to cross-platform buy, etc. If the 7nm brings out that, plus they have 64 PCIe and support PCIe 4/5, then we are looking at a BEAST! Intel may or may not add lanes, as 4.0 is double 3.0, and 5.0 is quadruple 3.0. So, 64 lanes on 5.0 for TR would be the bandwidth of 256 PCIe 3.0 lanes and would only need around 4xPCIe5.0 lanes to accomplish the same thing. That means, even if you did 4 graphics card in a PCIe 5.0x8 configuration, you'd have so much left for storage throughput that it is insane!

But this is why AMD has had me excited: their 7nm chips! Intel just has not got me there about their 10nm chips. Sure, they will have DDR5 and PCIe 4/5 as well, but there are other things...

 

http://www.pcgamer.com/amd-hires-fo...cs-execs-to-lead-its-radeon-technology-group/
"Rayfield will serve as senior vice president and general manager of AMD's Radeon Technologies Group, the company's dedicated graphics division from which Koduri departed, and Wang will hold the role of senior vice president of engineering for the same department. Both will report to AMD CEO Dr. Lisa Su.

"Mike and David are industry leaders who bring proven track records of delivering profitable business growth and leadership product roadmaps," said Dr. Lisa Su. "We enter 2018 with incredible momentum for our graphics business based on the full set of GPU products we introduced last year for the consumer, professional, and machine learning markets. Under Mike and David's leadership, I am confident we will continue to grow the footprint of Radeon across the gaming, immersive, and GPU compute markets."

Rayfield is a 30-year veteran of the technology industry, having most recently served an executive role at Micron. Prior to that, Rayfield was the general manager of the mobile division at Nvidia, where he led the team that created Tegra.

As for Wang, this is not his first stint with AMD. He previously worked on AMD's system-on-chip (SoC) efforts, and prior to that, he held various technical and management positions at ATI, ArtX, SGI, Axil Workstations, and LSI Logic. His most recent position before returning to AMD was at Synaptics, where he served as senior vice president of Systems Silicon Engineering."

 

Asus ROG Zenith Extreme Threadripper Gaming Motherboard Review + Linux Test

Beginning with Wendall's tepid recommendation - "if you need 10gbit ethernet, you get a card worth $100...":

cspkhm 9 hours ago
"I hate this board with a passion my board is complete crap and runs like ****. One week it will run fine then the next week because of crappy uefi code everything will become corrupted and nothing will work correctly. It takes a few weeks for the Uefi to go through garbage collect and fix it self. Because of the way the uefi memory is stored you can not downgrade to different versions. The numbers will change but the corrupted status will stay the same. AiSuit also overides the uefi set settings and the settings won't match. Its also fun when the company won't release drivers on there own site. Employees release them on overclock.net then random internet people post them on the asus forums."

James Mastroianni 13 hours ago
"the 10gig networking isnt value here, if you want 10gig value you get the asrock x399 fatal1ty board (which is a better board all around, asus is just overpriced)"

thegreatga 14 hours ago
"I have this motherboard and have really enjoyed it. I do however have the foxconn socket, and struggled with installing the cpu. It took multiple attempts and ultimately required extreme force to get it in. The included asus oc software sucked, and I completely uninstalled it because of so many problems (read forums)(very unstable and resulted in crazy voltage spikes). I easily oc'd my 1950X cpu to 4ghz with 1.125V, but could't get my gskill tridentz 32GB ram to the rated 3200, only stable at 2934mhz. The only complaint I have had is the lack of bios updates, while other asus boards and other board manufactures are getting updates more often this board seems to be forgotten. The lights and all the hardware do feel premium, but I really don't use any of the bells and whistles, I just wanted the 10Gbe nic that has been working flawlessly."

Krazie316 17 hours ago
"The motherboard and Threadripper had some issues on windows as well. The worst part is how The Ai suite software enables usesystemclock in bcdedit. I couldn't understand who a platform as expensive as this with as much processing power as it has could be responding so poorly in everyday normal usage. I had anything from glitches in the browsers to strange audio and video glitches in games. Immediately after finding out about the HPET issue and disabling usesystemclock in bcdedit, the system was working beautifully. Night and day difference, including a much faster boot time. There were a lot of people that either returned, or were about to return their Mobo's and/or Processors over the HPET issue. I'm glad someone figured it out although I don't think the information is widespread enough. I also got a Zenith Extreme with the Foxconn socket but my experience getting the screws to catch were not nearly as bad as other people made it out to be. I'd say all in all my initial experience was very disappointing due to the price these parts all cost. However it didn't take long to figure out how to iron out all the bugs and I'm very happy with my system now."
(Note: I think he meant useplatformclock )

To enable HPET as the only timer run the command:

bcdedit /set useplatformclock true

To disable HPET in Windows run the command:

bcdedit /deletevalue useplatformclock

From:

Try changing HPET settings to improve your PC's performance
by Martin Brinkmann on April 18, 2013
https://www.ghacks.net/2013/04/18/try-changing-hpet-settings-to-improve-your-pcs-performance/

 

@BeastsForever.TheDragon - this is the thread that talks more specifically about the tuning of TR. Somewhere in here I mentioned the TDP setting on the Asrock X399 Taichi. I pull over 400W, TANWare pulls just over 300. We talked a bit about it in the desktop thread as well. But, since you like AMD, here is for Ryzen news (TR = Ryzen).

 

So, EK is updating the baseplate for the waterblock and is making changes to their monoblocks as well, with a new base incoming. Asrock monoblock incoming for X399 in a week or two.

 

For the TR4 block, are my eyes broken, or am i really seeing that it is bent?

 

Just had the major performance issue with W10 and had to restore with Macrium Reflect. I just wanted to give a heads up that where .1709 allows for Iomuu to be enabled and still boot if it is enabled Macrium Reflect emergency boot recovery will still hang up on the x399 ASROCK.

 

I don't know if bent, but the fins are done differently. This is trying to make it right where they can send out replacement bases to customers because the original design was basically the same fin coverage as the x299 and tests showed 10+C hotter than the competition. So the left is the old, the right is the new.

Sent from my SM-G900P using Tapatalk

 

Is EK sending these to purchasers at a not cost fix?