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

Ah yes, the infamous (and overblown) AMD driver issues on Reddit.
First off, A LOT of those problems were down to OEM's themselves producing sub-par GPU's... not AMD's fault specifically.
Asus for one thing even went so far to provide an option for people to send back most of those Navi GPU's so their cooling assemblies could be mounted with appropriate pressure (since Asus hadn't bothered to test those GPU's mounting pressure for adequacy in the first place).

Also, most people hadn't actually experienced issues with Navi GPU's... and a good portion of those who had did not perform driver installation correctly, they also came from previous NV gpu's without removing those drivers entirely from the OS for example - even though two different sets of drivers in the OS CAN cause problems for BOTH AMD or NV gpu's )... have been running outdated OS-es, non-updated chipsets or BIOS-es, and in a few cases (albeit this was a very small sample) didn't have adequate PSU's (it wasn't the 1000W Platinum ones they were using either, but rather ones of less than half that capacity and questionable quality).

In various cases a proper installation of drivers resolved a lot of people's issues with the black screen... with others, it was specific cables which caused problems certain specific GPU's, or was just a plain badly made GPU by OEM.

Yes, AMD did acknowledge the black screen issue, but only because people who had them reported them... and it was a relatively small % of the population to begin with. The updated drivers DID help but it didn't resolve the issue with some people's systems... but it also turned out that those people who did have problems ended up having them because they got them from OEM's like Asus (you can't blame AMD for OEM's mistakes, nor can you expect of AMD to solve a hw problem created by the OEM's using driver updates).

Higher power draw was indeed AMD's trademark (not so much anymore), but we all know that AMD overvolts their GPU's out of the factory to increase the number of functional dies and that manually undervolting those GPU's drops power consumption substantially (to NV counterpart GPU levels or even below them - since even Polaris/Vega, undervolting managed to DROP power consumption to around or below NV counterparts while simultaneously increasing performance because the boost clocks can be sustained better).

As for Catalyst drivers layout, well, I've used it and can't say I had a problem in navigating it... but I also used NV drivers and one could say the same for their layout which underwent some changes over time (however, more notably, NV still uses outdated user interface for their drivers).

Using one or two bad experiences which happened long ago and allowing them to color your perceptions of every future product is called 'making assumptions'.

Obviously, you're free to avoid AMD gpu's if you want, no ones' going to stop you, but a LOT of the 'driver issues' as reported on Reddit are riddled with out of context assumptions made by a bunch of people whose understanding of IT is minimal at best.
Even IT experts end up struggling with a lot of problems that arise on a day to day basis.

Shall we also use previous bad experiences people had about AMD CPU's to color people's perceptions of those CPU's today?
My word... look at what the OEM's are doing to this day for AMD based laptops alone... they allow the hw to reach near 100 degrees Celsius when fully stressed (which is not acceptable at all and is more likely to cause problems down the line for other hw in such an enclosed environment where space needs to be managed with adequate cooling - even though the said hw is more than capable of operating with much lower temperatures under maximum loads if it had appropriate cooling in place).
They also refuse to pair more powerful mobile GPU's by using the excuse that Renoir lacks the needed lanes (even though the existing amount is adequate and that even high end GPU's don't suffer a noticeable peformance impact - about 1% in a 'worst case scenario' - which is of course nothing).

Not saying that AMD is not supposed to shoulder some of the blame, but I do think people are shifting too much of the blame on AMD and expect too much of a company that's a proverbial underdog with much fewer resources.
Oh and last time I checked, even NV drivers are causing headaches for plenty of people on NV forums... but for some reason, they aren't as 'advertised'.

Drivers can cause problems... be it from AMD or NV... each GPU (or even a system) will come with its own 'quirks' that the user will need to adjust to.

People here are slightly more experienced so its not a huge deal for us, but for the 'average joe'... one small mishap (without understanding the context or troubleshooting correctly) results in them returning the unit while listing 'specific' cause (even though something else easily could have caused the problem).

 

Catalyst Control Center was a nightmare on OEM jokebooks. You couldn't update to newer drivers at all! I believe AMD Adrenalin changed the driver performance drastically on supported GPU SKUs.

 

They did.
Albeit, despite having desktop grade hw in my Acer PH517-61, my Adrenaline drivers do NOT detect new drivers when they are released... hence I cannot update them from inside the drivers themselves... I am forced to manually download newest drivers.
This could be described as a problem shared by both AMD and Acer, but I'm also blaming Acer a bit more because they stopped supporting that unit/monster mere 6 or 8 months after release (without even releasing new GPU drivers at all). I could inform AMD of this, but dunno how much help would they be.

Is it annoying?
A bit, yes... but to be fair, I actually find that manually downloading the drivers and just 'updating' them on top of existing ones takes up similar amount of time than just going through Adrenaline itself... and gives me same results.
I tend to clean install a driver if enough revisions happened to justify it.

I chalk it up to one of those 'quirks' when it comes to drivers (and something a person needs to live with).
These desktop replacements are probably implemented in a strange enough capacity that's confusing the software. Its desktop hw in a laptop... so its neither desktop, nor laptop specifically... or rather something in between.

 

Revising my earlier estimates regarding RDNA 2 performance potential by taking into account the fact that AMD will be using N7P process node (which only allows either 10% efficiency improvement, OR 7% performance improvement), and taking into account AMD's 50% increase in performance per watt, I arrive at the following:

40CU RDNA 2 = 2080ti (at 4K) at 225W. Due to N7P node enhancement, we can reduce 10% from stock TDP, and this drops it to 202.5. If we take also hypothesize that AMD did NOT gain any IPC increases, by dropping the frequencies about 20%, performance drops by about 10% and power consumption goes down by another 40%, arriving at the following:

40CU RDNA 2 = 10% slower than 2080ti (at 4K) at 112.5W
48CU RDNA 2 = 10% faster than 2080ti (at 4K) at 157.5W
60CU RDNA 2 = 40% faster than 2080ti (at 4K) at 225W
72CU RDNA 2 = 70% faster than 2080ti (at 4K) at 292.5W
80CU RDNA 2 = 90% faster than 2080ti (at 4K) at 337.5W

This is actually in line with what some websites like Tweaktown suggested in their rumors.

Obviously, if AMD gains 10% IPC increase, then above performance numbers go up accordingly without increasing power draw.

Note: the power estimates do NOT take into account additional voltage optimizations... and given AMD's history for releasing voltage-unoptimized GPU's, the above values seem more or less reasonable.

Obviously, the size of RDNA 2 would be double with 80CU's (considering AMD will not be using EUV process node).

EDIT: I think I missed something.
I forgot to take into account that 50% perf. watt would come from both node and uArch enhancements (so, 10% efficiency gain on node and 40% on uArch - similar to what happened with enhanced Vega).
That basically drops performance further by about 5% for above estimates at same power draw:
40CU RDNA 2 = 15% slower than 2080ti at 4k and 112.5-125W
48CU RDNA 2 = 5% faster than 2080ti at 4k at 157.5-160W
60CU RDNA 2 = 35% faster than 2080ti at 225W
72CU RDNA 2 = 65% faster than 2080ti at 292.5-300W
80CU RDNA 2 = 85% faster than 2080ti at 337.5-350W

Hm... this might be more accurate.

 

AMD says its OEM's issue when driver can't be in-place upgraded.
I always do a clean driver install anyway. I update driver occasionally.

 

Figures its an OEM problem.
But as I said, I do clean driver installs too... it gives me a peace of mind knowing I'm doing it like that.

 

time for back to school. My grandsons, now going to kindergarten and 2nd grade are in need. that is of new laptops. Since it will be home schooling with Google Classroom I selected a pair of new Acer Aspire 5 3200u laptops. I just wanted to let others know sines the nation will be facing the home school dilemma cheap laptops may become rare, so get ready.

 

Grandson?
Interesting. Didn't know you were old enough to have one...

The Acer Aspire 5 3200u laptop costs £640 on Amazon UK.
Though, other computer websites (such as e-buyer, etc. - at least here in UK) offer 3500U and 3700u for lower or same price.

Depends on where you're buying them from I suppose and what your overall budget is.

I do agree that laptop prices have gone up (partly due to the pandemic and people buying stuff en-mass).

Finding cheap laptops might become even more difficult...

How much did you get yours for?

 

3200U and 3500U based laptops are very good. 3200U almost beat my 6700HQ with peak performance with 2c/4T vs 4c/8t. Also it had better thermals than 6700HQ. Only issue is sustained workloads when it crosses 15W TDP very quickly and OEM must uplift to 28W and it will certainly beat 6700/7700HQ w/o much issue. We got Ideapad 15 API with 3200U for 350$ for basic office usage. Only upgrade was 8GB RAM which brought total RAM to 12GB with 2GB RAM dedicated to VRAM. The Vega 8 is much much better than Intel HD graphics.

 

Threw in the towel in on my 2500u. While plenty fast enough I injured my right eye so the 14 inch is hard to see. I went larger with a HP Omen and a 4800h for now.



Performance mode



Have to say while the 2500u was fast, this is just way snappier.

 

So my options were;
1.) Acer Aspire 5 4700u @ 729+40 for 16GB and 256gb ssd
2.) Acer Nitro 6400h @ 679+ 40 for 16GB and 512GB ssd
3.) Legion5 6400H @ 779+40 for 16GB and 256 GB ssd
4.) HP Omen 4800H @ 949 for 8GB and 512GB ssd
5.) HP Omen 4800H @ 1049 for 16GB and 1TB ssd

I went for option 5, the best of everything.

 

Tried with Realtime CPU priority? It should be a tad bit faster but not too much!

 

TBH I have no need for further power. I just wanted the better snap so even a 4600H would have been enough. mainly wanted the better screen and 1TB & 16GB helps too.

 

This new software called ClockTuner for Ryzen seems very promising for tuning Ryzen's CPU performance and should be released in the next few months:
https://wccftech.com/1usmus-unveils-clocktuner-performance-boosting-utility-for-amd-ryzen-3000-cpus/

 

It says its only for Zen 2... would it by any chance work on Zen+?

 

I think not. There is a video linked to Tom's Hardware site that describes how it works. I think they said only Zen2 and beyond.

 

From the looks of it again just 3000 series and maybe above.

 

Now that I have gotten to put the laptop through some paces, all I can say is wow. This is just about as powerful as a 3700x. what it needs though is a RTX 2070 or better and you can say goodbye to the competition.

 

Yeah, the OEM's aren't pairing AMD mobile CPU's with more powerful GPU's because of lower number of PCIe lanes... or at least, they 'use' this as an excuse not to pair them with more powerful GPU's such as RX 5700M and RTX 2070 (to be fair, going past those GPU's, the price to performance ratio simply doesn't match as there's a point of diminishing returns in performance for a really high increase in price).

 

According to the Wccftech article posted above:
"How the tool achieves this is quite complicated but 1usmus's CTR is fully automated so users don't have to worry about anything. It is explained that the utility will increase the overall performance of AMD's Zen 2 processors, specifically those that feature a chiplet based design, by undervolting each of the individual CCX modules. By performing an undervolt on each CCX, Zen 2 processors are shown to not only run faster but also cooler. This also drops the overall power consumption while the CPU can retain an active-state for all the energy-saving technologies."

Ryzen+ is also made up of CCX modules, albeit no chiplets.
Once the tool is released, I'll see if I can try it on my Ryzen 2700 (certainly wouldn't hurt if by any chance it gains some performance at a more optimized power consumption - not that it needs a lot of undervolting since its pretty efficient and cool in my Acer).

 

While Nvidia is prepping for 30 series graphics cards, their competitor AMD is silently launching The Radeon RX 5300 3 GB Graphics Card, Features Navi 14 GPU With 1408 Cores

It looks like AMD isn't quite done yet with its first generation RDNA architecture as it has introduced the Radeon RX 5300 graphics card, an entry-level product designed for budget gamers. The Radeon RX 5300 comes with the same graphics core as the Radeon RX 5500 XT graphics cards but will come with a lower price point and tweaked specifications.

 

AMD should hv gone another route thru mobile gears than stationary desktops... That would draw more interests from those users who may own both systems... Going loggerhead, or thru traditional desktop gears, that's going to be a long journey man...

 

No matter wht way you look at it, Ryzen had to be developed to what it is today. If they had tarted with th 2500u the money for R&D would not have been there. Be thankful it is, what it is.

 

Yep. Just like having Polaris (low end) then Vega (trying to be high end but failing and being late), you have first gen Navi (low end), then second gen Navi.

To be fair, though, second gen Navi is a big die, something AMD has not done in a long while. So I have no idea if it will be worth something (depends on price and performance and drivers). Meanwhile, for Nvidia, still do not know if it will be worth it if prices go up. Even staying the same, eh... But that is for a different thread (basically arguing that the 2080 Ti is only about 24% faster than the 2070 super, so increasing performance to that level in a generation is underwhelming if costs increase by 20% for the 3070 (you are paying for most of the performance increase at that point)).

 

AMD goes new ways to promote their products. Or new ways to keep their fan base happy? What will be the next? Offer outdoor clothing or/and shoes that match?

Okay, you might think we're only joking there, but AMD truly is selling bikes now as a new item in the AMD fan shop.


From Big Navi to Big Wheels? AMD Sells Mountain Bikes Now tomshardware.com

Nvidia just announced its GeForce RTX 3000 series graphics cards, and it's safe to say, the promises are impressive. Many people are curious how AMD will compete, but as it turns out, AMD isn't just pedaling towards its Big Navi RX 6000 launch – it's also getting into the bike market.

 

So MSI is adding the RX 5600m to their Alpha 15 and 17. The GPU is no longer exclusive to the Dell G5 15 SE.
I wonder when other OEMs will use the RX 5700m as it's only present in the Alienware Area 51m R2.

 

Yowza... that's NICE.
The question remains how will MSI execute cooling for those units... because up until now (with 4800H and 5500m) its been... not as good as it could have been (MSI definitely needs to improve on this).

Yes, I'm actually curious whether OEM's will be using 5700m... its been available for AGES now and it should be comparable to RTX 2070 performance and efficiency-wise... but we may not see it used in other laptop designs until 2021 (by which point RNDA 2 mobile should be available for laptops - hopefully).

 

The 4th photo in the article has 8GB GDDR6. Maybe eventually they are going to use a 5700m? It also looks like the heatsink no longer obstructs the M.2 drives as well.

I will have 5700m benchmarks here in a couple of days. Hopefully the performance is on par with the RTX 2070 Refresh.

 

Let's hope they do a better job than the morons at Dell.

Although didn't they say the Dell G5 15 SE would be the only laptop with the 5600m this year? And by the time 2021 comes around won't we be getting laptops with Renoir + Ampere, making anything with a 5600m pointless (larger amount of power for worse performance)?

Let's face it, the G5 15 SE is only good if you got it ridiculously cheap when Dell were trying to get rid of them. The 5600M can't compete with Nvidia

 

Although speaking of the G5 15 SE, anyone installed this, and/or know what it is exactly (and what to be expected)?

AMD-Navi-10-Firmware_DNT55_WIN64_V1.0_A00.exe

Edit: apparently this fixes thermal issues.

 

I think AMD said the G5 15 SE, which Dell is selling again, would be the only laptop to feature SmartShift, which isn't exclusive to the RX 5600m. If MSI wanted, they could have implemented it into their Bravo and now refreshed Alpha series.

 

The 5600M CAN compete with NV RTX 2060 easily... if it was allowed to pull the amount of power it needs. In Dell Gg 15 SE, the 5600M wasn't drawing what it should have... its power consumption was by about 20W lower compared to where it should have been.

 

Did your 51m r2 arrive yet? I'm super curious to see how the 5700 performs within.

 

Yeah. In the Firestrike / Timespy it is basically on par with a 5700 desktop card (maybe 5% less) even though its TDP hovers around 120W. This means it is slightly worse than the 2070 in Timespy but slightly ahead of it in Firestrike. It is amazing how cool and quiet the system runs even while under load.

http://forum.notebookreview.com/posts/11045240/
http://forum.notebookreview.com/posts/11045402/

In Linux it works fine with the open source drivers but needs kernel 5.8 (at a minimum 5.5 I think). In MacOS Catalina it works fine with Apple's drivers as well but there is an eDP link training bug so the built in panel does not work. I may try the Big Sur beta to see if it fixes the issue.

 

RTX 2070 refresh has a TDP of 115W... so 5700m pulls only 5W more than that... which is pretty close.

Are the scores representative of 5700m vs RTX 2070 Refresh or earlier versions?

Both 5600m and 5700m have identical specs to their desktop counterparts with slightly reduced core and memory clocks.
That would easily put them about 5% behind their desktop counterparts.

It would be nice if AMD and OEM's would at least enable mobile gpu undervolting.

Have you tried using MSI Afterburner with 5700m to try and undervolt it?

 

Afterburner does not work on the 5700m from what I have tried so far.

 

Hm... I think you need to activate voltage control in MSI Afterburner (which comes with a restart of the software). I was able to unlock clock and voltage controls.

 

If I ever went 51m R2, it would be AMD primarily for the OSX support. 2070 performance and quiet load performance is icing on the cake. One thing missing from my P870TM1 is any type of AMD support.

 

I used to have an MSI WT75 which was basically fully supported under High Sierra. Then Mojave came out and Nvidia driver support was dropped. The Precision 7740 is ok for Catalina, although using the UHD630 with multiple external monitors can get pretty sluggish.

 

AMD is going for a more traditional, non-blower style of cooler for their reference RX 6000 series GPUs. And it'll have USB C port too.

 

So basically, copying Turing reference cards with the open air cooler and VirtualLink port.

 

Sadly on the Alpha 17, it appears that MSI has chosen to go with a 45% NTSC 144hz panel and a smaller 51w battery. It is a shame as the 17 appears to have an extra vent for the 5600m compared to the Alpha 15. Given that the primary competition is the Tongfang with 4800H and 2060 you would think they would not skimp on the display and battery here. No idea what they are thinking.

 

That's abhorrent (but hardly unexpected).
I was hoping MSI would do better than that.
What is the deal with OEM's who put in lower quality displays on these AMD units (and batteries)? They are saving literal peanuts while charging customers premiums.

The Eluktronics RP-15 easily outclasses the MSI Alpha 17 with a 4-Cell 62.36WHr Polymer Battery battery, 72% NTSC color accurate display (and the baseline configuration of Eluktronics with 4800H and RTX 2060, 16GB RAM and 512GB SSD costs £1064 - whereas the MSI Alpha 15 starts at $1300 - which will come to roughly £1300 in UK).

On top of that, the RP-15 probably has better cooling than MSI.

I could probably 'forgive' MSI the smaller battery, but the display is pushing it.

At this point, I'd prefer if Eluktronics created the same laptop but with 5600m and 5700m as options... and they would likely cost the same (or at least the 5700m would probably be a bit more expensive, but much faster as well -in line with RTX 2070 - especially with that cooling inside).

It goes to show OEM's are STILL stingy on all AMD laptops when it comes to certain things... however, at least they are not giving them single channel ram and just a HDD like they used to - but to be fair, if they don't improve on the display and battery, then how do they expect to entice people to get those machines in the first place?

And of course, people will then somehow find a way to blame AMD itself for how their laptops have poor battery life and inferior displays (completely unaware that its' the OEM's who make these decisions).

 

What really bothers me about contemporary laptops (and phones, but that's another subject) is the optimization for thin and sleek at the expense of performance, storage capacity, usable screen, and everything else. I don't care if my laptop weighs 7-8 lb (or even a little more); I want a big, high quality screen (no, I don't consider 72% NTSC to be "high quality"; I want a wider gamut for photography), lots of storage slots (2-3 2.5" bays and a couple of M.2 slots), fast processor, and lots of RAM. And I don't care if it looks like something built in the Soviet era.

While the CPU is more important than the GPU to me, a lot of people who would want that kind of laptop would want options for a fast GPU. If the laptop is thick, with plenty of cooling, that would allow for fast CPUs and GPUs.

 

Well... if ur memory on car history is good enough... A rebadged FORD Laser/Telstar commands a little more $$$ because of that blue badge. Yet the original build, Mazda 323/626 was cheaper in most markets where they were sold together with the FORD variants.
Is this an unethical business trick... it's up to individuals to evaluate. Back to todays computing products, TongFang's build is sold in different market under different badges and little difference in component specs. Here in Singapore, the AfterShock Apex15R lite, relative cousin to RP15 in America gets smaller battery, yet default RAM of 3200mhz, and 120HZ FHD Wide View Angle Display, with GTX1650. At the price of U.S. Specs with RP15 better...
https://www.aftershockpc.com/welcome/ProductconfigBaseic/apex-15r-lite

I would rather spend on a Lenovo Legion 5 with same CPU, same screen, yet doubles of RAM and fatter battery...

 

And to touch on the ugliness of the OEMs trying to put on AMD system...

...what a coincidence in time!​

 

72% is much higher quality than 45% though... but yes, nothing prevents OEM's from installing 100% NTSC panels... and quite frankly, I don't think the differences (money-wise) for them are huge or relevant.

As for the thickness... to be fair, OEM's could design thin and sleek ones that run cool, but it would require a total redesign of the cooling assembly inside laptops... something which OEM's are apparently reluctant to do (because if you may have noticed, the underlying basic cooling assembly in laptops hadn't really changed in decades).

But yes, I agree that if they want to stick with conventional cooling designs, they need to create better chassis with more effective cooling (as Eluktronics demonstrated, its more than doable).
Plus, my Acer PH517-61 with desktop grade Ryzen 2700 and Vega 56 are running cool and quiet - and while yes, that IS a 17" unit which is slightly thicker and heavier than usual, its not a huge difference (certainly nothing that I can't handle and still eminently more portable than a desktop).

It goes to show there's room for improvement with even existing designs... but OEM's are too lazy.
Leave it to Acer to create an excellent desktop replacement which runs cool and quiet, but then completely drop the ball on the BIOS and never release new updates which would allow you to put in a Zen 2 and Zen 3 into the thing.

 

I am even more confused about the specs now. I can see that they have a SKU of the Bravo 17 with a 144hz 3ms 72% NTSC panel, paired with the 5500m. There is really no reason why they should not offer it as an option on the Alpha 17. Maybe it was a mistake in the spec sheet? Maybe the larger battery needs extra time for certifications but the display is a no-brainer.

It would be great if the 5700m eventually comes to the MSI Alpha series. I still think that photo in the press release with 8GB GDDR6 means it is in the pipeline. Especially if the Alpha series bios can be unlocked and it has a MUX like the first version did. Who knows when we will see RDNA2 mobile parts. Probably not announced in January if I had to guess.

 

I hadn't seen any plans for ANY OEM's to use 5700m as of yet (except for DELL).
MSI using 5600m was a surprise too since Frank Azor (from AMD) said no other laptops (except DELL) will come out this year which have 'Smartshift' enabled (which by default is a hw feature of Zen 2 and therefore cannot be switched on or off on demand... its always active if the system is using both an AMD cpu and dgpu - this implied we won't be seeing any other all AMD laptops this year... so, MSI using 5600m was a surprise).

Like I said, at least I hope MSI decided to use better cooling. The 5500M in their Bravo series is heating up to really unacceptable levels (there is no real reason for this except MSI skimping on the cooling).

 

Maybe some good AMD news coming down the line.
https://arstechnica.com/gadgets/2020/10/amds-new-zen-3-ryzen-desktop-cpus-arrive-november-5/