AMD Responds to Lack of Ryzen Mobile Driver Updates, Claims OEMs are the Issue

[Valantar]

But, but my underdog can do no wrong!

But evil shintel and ngreedia!


Oh wait, i can still update Nvidia mGPU drivers from 2009 just fine(gt230m)

Shitty support gets you shitty reputation. RyZen was good. Meanwhile RTG has been nothing but a train wreck.

Underdogs can absolutely do wrong, but they do of course have less power to abuse. The very existence of this thread ought to tell you that nobody here is that blind. Also, RTG's dGPU driver effort in the last couple of years has been excellent. Which is a big part of why this is so baffling.

 

[FreedomEclipse]

What does legacy have to do with the discussion at hand?
(And yes, supported for old architectures is relegated to a legacy branch. It's still one driver, it just doesn't make sense to send the bits for old cards to everyone. But even that gets updated from time to time.)

Someone was moaning about AMD not supporting some 25year+ gpu

 

[bug]

Someone was moaning about AMD not supporting some 25year+ gpu

AMD tends to drop support rather quickly. It's way off topic, but it's true.
For example, Nvidia has only stopped support for their 6 series (that's 6000 series) in 2015. They supported that for 11 years, right up to Win7.

 

[Valantar]

AMD tends to drop support rather quickly. It's way off topic, but it's true.
For example, Nvidia has only stopped support for their 6 series (that's 6000 series) in 2015. They supported that for 11 years, right up to Win7.

The perks of being a bigger company, I guess? I remember they dropped support for my HD 6950 not too long after I sold it. Can't say I mind, considering that the GPUs losing support are usually old enough to not perform sufficiently in modern games anyhow. Though of course it sucks for those few games where performance is garbage for some non-hardware reason, where driver fixes deliver 50%+ increases.

 

[bug]

The perks of being a bigger company, I guess? I remember they dropped support for my HD 6950 not too long after I sold it. Can't say I mind, considering that the GPUs losing support are usually old enough to not perform sufficiently in modern games anyhow. Though of course it sucks for those few games where performance is garbage for some non-hardware reason, where driver fixes deliver 50%+ increases.

That's part of the reason Nvidia moves those to legacy. The support is more or less set in stone, past a certain point. All that's needed is security fixes and support for new Xorg releases on Linux or WDDM revisions on Windows. So it's not like it eats into your resources, if driver development is set up nicely. AMD's history wrt drivers is more rocky though and their support is shorter. Some find that annoying.

 

[xkm1948]

Underdogs can absolutely do wrong, but they do of course have less power to abuse. The very existence of this thread ought to tell you that nobody here is that blind. Also, RTG's dGPU driver effort in the last couple of years has been excellent. Which is a big part of why this is so baffling.

Oh sure. RTG actually works on improving dGPU drivers. If that were the case i would not have moved on from FuryX. Bugs upon bugs upon bugs. Every single driver after 18.5.1 broke VR for me l, submitted tons of reports and only email i got was “noted”

No, hell freaking no. I have given RTG way too many chances. They are no longer the ATi i used to like.

 

[Valantar]

That's part of the reason Nvidia moves those to legacy. The support is more or less set in stone, past a certain point. All that's needed is security fixes and support for new Xorg releases on Linux or WDDM revisions on Windows. So it's not like it eats into your resources, if driver development is set up nicely. AMD's history wrt drivers is more rocky though and their support is shorter. Some find that annoying.

That sounds like a very sensible solution. Keep providing sufficient support that it works, but no more or less.

Oh sure. RTG actually works on improving dGPU drivers. If that were the case i would not have moved on from FuryX. Bugs upon bugs upon bugs. Every single driver after 18.5.1 broke VR for me l, submitted tons of reports and only email i got was “noted”

No, hell freaking no. I have given RTG way too many chances. They are no longer the ATi i used to like.

Hm. VR must be a weird edge case, I haven't had a single bug or driver fault with my Fury X for years. I did have a series of driver crashes when I was undervolting it heavily (around -.1V IIRC), but that's my own fault.

 

[TheGuruStud]

Laptop = useless, so who needs a driver anyway? Lol. XD

99% of ppl with a laptop don't update or use anything, so you can guess why no one cares.

 

[Vya Domus]

Just because someone still issues driver updates for 10 years or more doesn't mean they are particularly useful. My Nvdia powered laptop used to receive countless updates for years yet the black screens on boot after almost each of them never disappeared. What use was it ?

Whoever seriously believes any product gets more than 2-3 years of actual serious support is deluding themselves.

 

[FordGT90Concept]

Oh wait, i can still update Nvidia mGPU drivers from 2009 just fine(gt230m)

"Updating drivers on a Radeon Mobility HD 5850 was a problem" said no one, ever. ATI/NVIDIA drivers always worked for discreet GPUs because no AIBs/OEMs modify the hardware as far as drivers are concerned. That's not true with Intel Atom, AMD E-series, and apparently Ryzen U-series. The manufacturer has to take Intel/AMD's base drivers and do something (don't know what specifically) to make them work.

RyZen was good.

We're talking about a Ryzen. It's literally in the thread topic.

 

[bug]

"Updating drivers on a Radeon Mobility HD 5850 was a problem" said no one, ever. ATI/NVIDIA drivers always worked for discreet GPUs because no AIBs/OEMs modify the hardware as far as drivers are concerned. That's not true with Intel Atom, AMD E-series, and apparently Ryzen U-series. The manufacturer has to take Intel/AMD's base drivers and do something (don't know what specifically) to make them work.

That's a logical fracture. You say dGPU drivers aren't a problem because 3rd parties can't modify the hardware. And then proceed to tells that's not true wrt IGPs. How can a 3rd party modify an IGP?
If anything, it's a dGPU designs that can be customized.

 

[FordGT90Concept]

AMD tends to drop support rather quickly. It's way off topic, but it's true.
For example, Nvidia has only stopped support for their 6 series (that's 6000 series) in 2015. They supported that for 11 years, right up to Win7.

X series had their final driver released in 2010 (Windows Vista).
HD 2/3/4 series had driver support until 2013 (Windows 8.1).
HD 5/6 series had driver support until 2015 (Windows 10).
HD 7/8 (debuted 2012) and R# series have driver support through today. GCN definitely has everything NVIDIA ever released beat in terms of driver support.

That's a logical fracture. You say dGPU drivers aren't a problem because 3rd parties can't modify the hardware. And then proceed to tells that's not true wrt IGPs. How can a 3rd party modify an IGP?
If anything, it's a dGPU designs that can be customized.

It's not OEMs modifying integrated GPUs, it's the driver that allows communication with the CPU in general. Something, somewhere, prohibits vendor agnostic drivers from successfully working; they need vendor specific drivers.

As ACE76 suggested, it could easily be power/heat related. The Vega component of these Ryzen APUs may be virtually the same as their big brethren meaning given the power and thermals, they could consume 100+ watts. Vendor specific drivers set a limit so the machine isn't damaged. Yes, you'd think that would be in firmware/BIOS and not in drivers so that probably isn't it unless there's a cost cutting component to it (would require non-volatile memory) so they handle it in driver instead.

Another thought is that all of the overclocking components of the driver are excluded by OEMs as part of their licensing/contractual agreement with AMD to prevent damaging hardware and creating warranty complaints.

TL;DR: HP says "jump" and AMD asks "how high?"


Edit: I'm getting the impression RTG doesn't even know what the problem is. At least not yet. The outcry over this has made it a priority.

Edit: It is isolated to HP Envy X360 as per the Redditor that reported it which, coincidently, was the first Ryzen Mobility product ever launched. It's a special cookie.

 

[yeeeeman]

The biggest problem here is that some OEMs like Lenovo block the instalation of non-signed drivers, meaning that you can install only drivers from Lenovo website, which is any way you put it, stupid. I have got a HP pavilion laptop with 6700HQ and nvidia gtx 950m and I installed all my drivers from Intel/Nvidia website, no problem.

 

[ACE76]

What kind of a beta driver for Vega are you talking about?


Yet it's been possible in the past with for example the A-Series APUs for AMD to deliver drivers through their website and if users were dealing with instabilities, the OEM drivers didn't suck as hard as they do now. I think you're the one that doesn't understand the mobile space as much as you should to talk on a forum like this.

Gimme a break...Ryzen is a single platform...the mobile CPUs are just cut down CPUs from their desktop variants...yet AMD is magically providing drivers for the desktop CPUs just fine... maybe their just doing this to the mobile CPU out of spite...yeah that must be it...it couldn't possibly be that the OEMs are customizing the frequency and power delivery, thus requiring custom drivers.

 

[FordGT90Concept]

What I don't get is why now? It has been out for over a year now...

https://twitter.com/i/web/status/929885959918845952

 

[Darmok N Jalad]

The biggest problem here is that some OEMs like Lenovo block the instalation of non-signed drivers, meaning that you can install only drivers from Lenovo website, which is any way you put it, stupid. I have got a HP pavilion laptop with 6700HQ and nvidia gtx 950m and I installed all my drivers from Intel/Nvidia website, no problem.

I have the same issue with my iMac 27” on boot camp. Running the normal AMD installer, it won’t find any compatible hardware. I can only use the Radeon Pro drivers that Apple qualifies, which are about 6 months behind. There are third-party modded drivers available, and they require booting to safe mode and patching them to look like the Apple signed drivers.

 

[Valantar]

It's not OEMs modifying integrated GPUs, it's the driver that allows communication with the CPU in general. Something, somewhere, prohibits vendor agnostic drivers from successfully working; they need vendor specific drivers.

I think the thing prohibiting it is OEMs terrified of being blamed for a buggy driver release from a third party. Period. Nothing else than them flat-out demanding absolute control over this, while being neither willing nor able to spend the required resources to ensure timely delivery of driver updates themselves. Why they trust Intel with the same thing is a very good question - likely due to more experience dealing with them, but hwo knows?

As ACE76 suggested, it could easily be power/heat related. The Vega component of these Ryzen APUs may be virtually the same as their big brethren meaning given the power and thermals, they could consume 100+ watts. Vendor specific drivers set a limit so the machine isn't damaged. Yes, you'd think that would be in firmware/BIOS and not in drivers so that probably isn't it unless there's a cost cutting component to it (would require non-volatile memory) so they handle it in driver instead.

That is not how hardware works. Each APU has a specification, which covers frequency and power, which it will not exceed unless explicitly told to do so by the BIOS. My A8-7600 HTPC actually lets me set the GPU clock speed directly in the Bios. If the option wasn't visible there, it would be locked down. No Vega APU iGPU for mobile boosts past 1300MHz, and in those frequency ranges Vega is quite efficient (not that far from Pascal in perf/W, but ofc far behind in absolute performance). From what you're saying here, a Vega 64 pulls... 640W, if a Vega 10 pulls 100W? That's dry ice/LN2 territory, and any functioning GPU would thermal throttle long before reaching this.

There are both clock, power and thermal limits that are not driver controlled. Of course the vendor is responsible for implementing the BIOS, but I'm quite sure it would be near impossible for them to mess that up so that a rogue driver could melt your APU - if so, that's a BIOS error, not a driver error.

Another thought is that all of the overclocking components of the driver are excluded by OEMs as part of their licensing/contractual agreement with AMD to prevent damaging hardware and creating warranty complaints.

TL;DR: HP says "jump" and AMD asks "how high?"

I think you're right about the latter. As for locking out overclocking, that ought to be trivial, as U-series chips don't generally allow OC'ing in the first place, so all that's needed for that is a check for "CPUID=2200U/2300U/2500U/2700U true/false" in Ryzen Master/Catalyst.

 

[FordGT90Concept]

Searching on the internet, I get the impression not many people actually have problems with the HP Envy x360. It appears to be one person that's driving the noise. Others might be jumping on the bandwagon just because they'd like to see APUs have a driver release cadence like GPUs do but that's not going to happen because of cost.

It might be that Envy's firmware requires HP-signed drivers to work. As pointed out previously in this thread, Lenovo has done that. Apple too.

 

[R0H1T]

Like Ford said ~ do the ULV chips allow IGP OCing? if so it's possible that the OEM may want to cover their collective arse, since the cooling would obviously not be sufficient in such cases. Allow generic AMD drivers would probably make it impossible for them to control this aspect, since there's also AB & other tools that could work, even if AMD shut it down in their software? Intel doesn't allow OCing anything that isn't explicitly unlocked on mobile chips.

 

[silentbogo]

Yes, when you include GPUs and similar generic PCI Express-based hardware. When looking specifically at APUs, that's not the case. People have tried AMD's generic driver on these machines (and E-#50 previously) and the generic drivers have lots of problems. OEM-tailored drivers are absolutely required because the hardware changes they made are extensive.

Those are usually set by hardware, not software. If you look at modern laptop motherboards - they only look different. Underneath is a generic cookie-cutter design for each platform with minimal modifications required to make a dozen models of the same laptop. All the new soldered shit is also straightforward(especially Ryzen and any intel older than 6th gen): it's just an SoC with barely any logic ICs on the board. That same HP Envy X360 only has SoC, KBC, LAN and audio.
If a mobile platform has provisions for dGPU, it usually boils down to following: Is there a GPU? Yes? Put a strap here, add this section to firmware.
Issues with display outputs? Also gone on modern platforms. As far as all the stuff that's been at my workshop, all new laptops have eDP output to LCD as default and use iGPU as a primary output device, and all have HPD (hot plug detect) enabled regardless of whether it's AMD or Intel. HPD is also wired directly to SoC, so as data lines for I/O.
Structurally all older laptop platforms had a structure of the christmas tree, with CPU, NB, SB (or CPU, PCH and SuperI/O) as a trunk. Now it's a soviet star with SoC at the center. The only modifications you can make, is route your PCB trace to the left, or to the right (figuratively speaking). All issues usually are a result of OEMs going out of spec and cutting corners.

If AMD made a generic driver, the only issue might be a TDP lock, but I believe it should be set in UEFI firmware and not relate on driver alone to do the "limiting". Other than that there is absolutely no reason why AMD hasn't released one already. Though, I have a suspicion that they want OEMs to do the bidding, cause if something goes wrong - they'll be the ones to blame.

 

[Valantar]

Like Ford said ~ do the ULV chips allow IGP OCing? if so it's possible that the OEM may want to cover their collective arse, since the cooling would obviously not be sufficient in such cases. Allow generic AMD drivers would probably make it impossible for them to control this aspect, since there's also AB & other tools that could work, even if AMD shut it down in their software? Intel doesn't allow OCing anything that isn't explicitly unlocked on mobile chips.

There is absolutely nothing saying a generic driver couldn't detect U-series mobile chips and disable OC functionality. I don't know how many lines of code it would require, but I'd guess you could count them on your fingers and toes, at the very worst. Also, there is this phenomenon called thermal throttling, which OC settings can't change. One would almost suspect that it's a safety feature?

 

[maxli86]

Being a AMD Notebook user since AMD Phenom II mobile era, I can share this.
Worse experience with AMD driver was back with AMD RIchland notebook that support AMD Dual Graphics/Crossfire. (AMD A10-5750M with Radeon 8670M)
The OEM AMD Driver was so bad that often crash or had issues especially when Dual Graphics/Crossfire was turned on.
Tried to feedback or request for updated driver from the manufacturer, they quote that was the only available driver for that product.
Even tried to reach out to AMD about this issue, they also mention I should contact the manufacturer or try generic driver from AMD.
However since it support Dual Graphics/Crossfire, generic driver can't even be installed.
After this NB died, now I switch to Intel than getting another AMD NB fear of disappointment especially the driver.
My point is being the longest AMD mobile user, manufacturer don't always release new or updated driver.
Even if you try to feedback or request for new driver, usually they will quote that is the only available or best driver for that product.
With Windows 10, the annoying thing is it tend to update driver that it deem outdated which may cause problem or worse stability issue.
It raises a good question if there is issue or somehow when Windows 10 deem the driver that is outdated.
Can user install generic driver from AMD website? Or how can user obtain latest driver especially when there is issue?
Just a suggestion, maybe AMD can consider providing latest driver that is certified for use with certain notebook as a first step on AMD support site. (Maybe a unified site for mobile users)
Whether or not manufacturer will release their own driver or not at least it's up to them but at least there is a updated driver for use while waiting.
As well as gather feedback and issues from AMD notebook users on what are the issues so there will be fixes and etc.
With Windows 10, if possible there should be driver update every three months or so thus need not to be very frequent unless there is a major issue like OS crash and etc.
If AMD want to push in mobile segment especially mobile gaming, specifications is one thing while driver is the second key component.

 

[FordGT90Concept]

Honestly, until AMD or HP says something more, we're just speculating. Since Envy x360 has been out over a year and they're still selling it, not many people have complained. Reason why we're even talking about it now is because a Reddit post went viral (tech websites published articles about it). Now AMD is obligated to figure it out (or compel HP to) in order to save face. It will be fixed one way or another, just a matter of when at this point.

 

[yotano211]

hmmm not sure if Intel is paying OEMs to stop development of AMD OEM drivers

Intel doesnt have to pay any company for that. I rarely see that laptops with any kind of AMD cpu. The ones that I see are mostly lower power processors. The real money in laptops is in the higher power processor and only intel plays in that field.

 

[Valantar]

Honestly, until AMD or HP says something more, we're just speculating. Since Envy x360 has been out over a year and they're still selling it, not many people have complained. Reason why we're even talking about it now is because a Reddit post went viral (tech websites published articles about it). Now AMD is obligated to figure it out (or compel HP to) in order to save face. It will be fixed one way or another, just a matter of when at this point.

I don't think many people are having issues either, but not updating the GPU driver for over a year for a product with the iGPU as a major selling point is... rather embarrassing no matter who is responsible. Time to tackle this, maybe spend some of that Ryzen/EPYC cash on a few more people for RTG's driver team? Though I wouldn't be too surprised if this ends up being a consequence of the reportedly significant lack of communication between RTG and the rest of AMD under Koduri. Might be they say "nah, that's Ryzen, it's not ours, we've done our part, we gave you a driver a year ago, you update it."

Intel doesnt have to pay any company for that. I rarely see that laptops with any kind of AMD cpu. The ones that I see are mostly lower power processors. The real money in laptops is in the higher power processor and only intel plays in that field.

Real money? ULV (~15W) laptops likely outsell anything more powerful by a factor of 10:1, including in the massively lucrative enterprise sector. Even with smaller margins on ULV than higher wattage parts, there's no doubt there's more money in the thin-and-light mainstream. If there was more money to be had in higher-end segments, AMD would have focused on getting 6-8-core Ryzens into more laptops.