AMD, Intel and Nvidia driver issues and last recommended version

[NikosD]

He could easily be the fourth or fifth tech journalist writing the same garbage regarding Ampere's performance versus any other card.

I have really lost counting the nVidia pro journalists.

I'm talking about Anthony Garreffa and his hilarious article: https://www.tweaktown.com/news/75367/if-these-new-amd-big-navi-leaks-are-true-the-rtx-3080-will-destroy-it/index.html

I really have no idea if Big Navi21b exists and if it will be faster than 3080 eventually.

All that I'm saying - regarding the same mistake or "mistake" of tech journalists - is that according to the article if Big Navi21b has the specs that have already leaked - we all agree that the performance in TFLOPS metric will be 22.5 TFLOPS just like he says while 3080 we already know is 29.8 TFLOPS like he says too.

Then the result of this comparison of Big Navi21b at 22.5 TFLOPS vs 3080 at 29.8 TFLOPS means that Big Navi21b will be FASTER, clearly FASTER in GAMES than 3080!

Not the opposite!

What an idiot!

[butterw2]

I mean, you're just describing the normal evolution of tech, the first implementations of a new tech are always lower performance than the older standard but after a few generations nobody would want to go back to the old standard. Would you have have wanted to go back to 16-bit rendering in the mid-2000s?

Many shills for Nvidia marketing were adamant that Raytracing was a game changer when RTX released. I'm just saying there has been very little evidence of it so far. Also the cards are expensive, power hungry and low on VRAM.

Even VR might come good at some stage. That doesn't make those who bought into the early hype right.

[SamuriHL]

That is why you always buy a card for what it can do at the time you buy it rather than for some future promise. I never bought the RTX cards for RTX in gaming, although I do enjoy it where I've seen it. I bought them for madvr and don't regret that purchase one bit.

[glc650]

looks liek my repoprted bug is fixed: [Notebook]: Performance Power Mode cannot be set from the NVIDIA Control Panel.
[200657525]

the notebook part is odd but what ever.Have you been able to confirm if that's the case?

[huhn]

not yet

edit: the entire behaviour is totally different on my system. even with optimal is very aggressive but still different from the rest but it's for my system now totally save to use it's very aggressive compared to the old adaptive.

i even trigger a new bug never seen by me before:
https://abload.de/img/bugwdjiz.png
it even was at over 50% clock deviation.

[SamuriHL]

Holy hell that doesn't seem good. LOL I haven't had a chance to play with the new driver yet myself. On the list for tomorrow night hopefully.

[ryrynz]

I'm interested to know if with the 20/30 series and 20H2+ builds and latest driver I can get zero frame drops and repeats over the course of a couple of hours or so.

[VBB]

Latest Nvidia driver is good. Fixes at least one issue I had with the previous one (RGB@60Hz HDR bug). Timings are the same for me, power profiles as well (as far as I can tell).

[NikosD]

Latest nVidia driver tries to fix the "2.0 GHz Frequency Wall" issue of 3080/3090 cards but without success.
They managed only to cap GPU frequency at 2.0GHz, dialing back some performance of GPU boost, in order to avoid the Crash-To-Desktop result of a still unknown root cause.
Some AIBs like Colorful which was the first one to report officially the problem, have already released new BIOS to cap frequency at 2.0GHz.
After the low availability of 3080 cards and the extreme low availability of 3090 cards, this "Frequency Wall" issue was the last thing needed to make this release even worse.
Tough times for 3080/3090 buyers or potential buyers...

[ryrynz]

Buy a manufacturer that cheaps out on caps n that's what you get.

[huhn]

while it is true that there are now 2 revision out for card with different cap setup but that doesn't change that the 3080 FE is crashing like other cards too.

these kind of things sadly happen with most releases.
NAVI was black screening for 9 month this issue was just removed from the known issue list in the new driver.
polaris was breaking the PCIe spec.

we still don't know if the caps and the crashing issue are the same issue EVGA claims the "bad" revision where only send to reviewer.

and bad AIB cards are sadly pretty normal asus TUF 5700 XT does someone really want revision 1 or the strix where asus wasn't able to screw correctly?

https://youtu.be/OJU8jKIYtS4?t=256
don't blame this on AMD.

[nevcairiel]

It appears that at least part of it is that the boost algorithm is too aggressive under certain workloads, which makes the clock jump up far higher then it usually does, and causing the crash in the worst case. Limiting those jumps should not impact performance noticeably - in fact some people are reporting higher speeds with the new drivers since the clock is more stable and thus a bit higher in average, instead of jumping up and then being corrected down for power limiting. But more thorough testing will surely follow.

[el Filou]

I'm afraid we're headed to more frequent issues like this as chip manufacturers rely on dynamic boosting to the absolute limit like GPUs have been doing for a few gens, and AMD started doing with Zen and Intel with TVB in a less extreme way. Margins are thin, feels like the very first P3 1,13 GHz. Thankfully clock controls are now very finely grained and they can fix it like this.

[brazen1]

I can confirm 456.55 modded with 388.59 audio driver is working pretty good using a GTX 960 4GB. I understand this is a well seasoned GPU but I honestly have no reason to upgrade because I passthrough to a Q90R. I know others still use this model as well. I notice great improvements in clock deviations making frame repeats a thing of the past now as has been reported by others. SDR and HDR. I did notice a 2160p 60fps title (Gemini Man) slowly improves but is improved nonetheless.

I do see when the player is windowed or when the seek bar is in view, HDR goes away. The switch is immediate back and forth. The panel does not flash in or out of mode. I tried setting nVidia and madVr to 8bit instead of 12bit. I also toggled FSE on/off in madVr. This did not help fwiw.

Despite this small problem, 456.55 will now become my daily driver because the trade-off of gaining a near perfect clock deviation is worth more to me. Good job nVidia!

On a side note, Gemini Man (2019) has been my go to video for displaying my system. The way video presents is superb imo. I recently obtained Aircraft Carrier Guardian of the Seas (2016). This too is a very high quality presenter imo.

[NikosD]

After reviewing some overclocking data regarding 3080/3090 cards, it is more than obvious that nVidia pushed the overclocking envelope to its limit.
They didn't leave almost any performance on the table, out of the box.
It's a very aggressively clocked product, extremely high-powered and rushed.
But why ?
Because as I have said before, nVidia is not simply afraid of Big Navi.
nVidia is terrified.
So, after the first leaks of Zen 3 performance in games, with the 5800X using 8C/16T being faster than the fastest Intel CPU for desktop the 10900K of 10C/20T, we can now be sure for one thing:

AMD is going to be the champion in both CPUs and GPUs at the same time for gaming workloads, probably for the first time in history.

It will be an epic, monumental milestone for tech.

The next month coming in a few days, will be described in history tech books as Red October

Historic moments.

[chros]

The next month coming in a few days, will be described in history tech books as Red October
:D Hope so!
Btw I'm curious whether they will have 45W 6C 12T CPU (for HTPC without fans).

[el Filou]

Do HTPCs really need 6 cores? Or are you thinking of resale value?
I'd be happy with a much lower TDP 4C/8T, I'm still not limited by my 12-year old actual 25-30W overclocked 2C/2T except for hypothetical 4K copyback which I don't need yet.

[huhn]

maybe there will be a mainboard with a 4800U/4700U that should be your CPU have your eyes on ali express you can find these things there from time to time.

current zen 2 desktops have the relative inefficient io die.

[pirlouy]

AMD is going to be the champion in both CPUs and GPUs at the same time for gaming workloads, probably for the first time in history.

It will be an epic, monumental milestone for tech.

The next month coming in a few days, will be described in history tech books as Red October

Historic moments.
:D

I quote it for the future (I know you don't say it seriously).

[butterw2]

:D

I quote it for the future (I know you don't say it seriously).

Aunt Lisa is coming for you Jensen !


Zen3 5000 series should be released in October.

4600G, 4600GE Renoir zen2 apu might become more widely available:
https://www.amd.com/en/products/apu/amd-ryzen-5-4600g

For low power use, power limits would be an option also, and ClockTuner for Ryzen (CTR) has just been released.

[NikosD]

Another day, another incompetent (shills) tech journalist(s) come up.

They blame in their article the drivers/gaming engines of not using the full FP32 (TFLOPS) potential of Ampere architecture, like VRAY, just as gaming engines have to be CUDA rendering apps.

Instead of "blaming" the Compute-oriented (not gaming) Ampere's architecture and its maximum TFLOPS number advertised by the-leather-jacket man, completely unrelated with gaming workloads which execute 36 INT commands after 100 FP commands on average.

Gaming workloads are not interested in maximum advertised TFLOPS, achieved under specific non gaming conditions.

https://wccftech.com/nvidia-rtx-30-fine-wine-investigating-the-curious-case-of-missing-gaming-performance/

[Klaus1189]

I added it, but I must admit, it is still hard for me to follow the intel update and naming sheme. I also don't understand the supported grafics, like older integrated grafics will require older or other drivers, isn't it.

Please let me know of future releases and I add it, same for issues of it :devil:

[el Filou]

There you go: https://www.intel.com/content/www/us/en/support/articles/000005654/graphics.html ;)

iGPUs before Gen9 are out of performance/quality support now, they just release new drivers for older hardware when there's a security issue.

[SamuriHL]

Geeze Intel. Gen 9?? Really?

[el Filou]

That's what I'm understanding from the driver I'm using for my HD 4600 that covers Gen7 & Gen8, the latest "quality update" was in september 2015, all the ones released since then just mention "security updates" in the release notes. They're only adding features/correcting bugs in the newer 100.x driver that only supports Gen9+
(Edit: just to be sure to avoid any confusion, I mean Gen9 graphics (Skylake), not 9th gen Core CPUs.)

[VBB]

I have been using the 100-series drivers with my 6th gen CPU from the beginning. There is a cut-off with all Intel drivers, but in the case of the 100-series, the cut-off is 6th gen, not 9th gen. What's always been confusing about Intel drivers is overlap between their major revisions. With my 6th gen, I could also use the 15-series driver, for example.

Just saw your edit. That clears things up.

[SamuriHL]

Yea that makes more sense.

[NikosD]

According to the latest speculative video of MLID, the leaks of reverse-engineered MacOS drivers of Big Navi Radeon Pro card were right, although he says that AMD is very cryptic and intensionally is misleading specs, names inside drivers and performance even to AIBs!

The prosumer Big Navi21 will be 80CUs@2.05-2.2GHz with a TDP of around 250W and 32GB GDDR6 using 256bit bus, according to the video.

BUT the Top GAMING Big Navi21 will be even faster, if it sticks to 80CUs and not less.
So, the top gaming Big Navi21 will be a 72CU-80CU at 2.15-2.3GHz with a TDP of 250W to 300W and 16GB of GDDR6 using 256bit bus.
IF AMD actually releases such a monster of 80CUs at 2.3GHz, it is impossible to not be faster than 3090 in games.

Yes, AMD can kill nVidia this time.

His opinion is that Top Gaming Big Navi21 just trade-blows 3080, which doesn't make sense with such a configuration.

[NikosD]

I just saw a message from Tapatalk that @videoh quoted a post of me.

But where is his post ?

[chros]

I saw that video as well, interesting times, similar performance with 100W less?
The price will be the real question.

[Klaus1189]

He/She asked for the link of the video of MLID, but it seems that the post is gone...

[NikosD]

I was thinking about it the last few minutes...

Why He/She ?

[Klaus1189]

I don‘t know if @videoh is a boy or a girl :) The post was a quote of your first sentence of post 1934. He (assuming now) is asking for a link to the video of MLID. Way to less girls around ...

[NikosD]

I think videoh is a very well known member of this forum and has been in the past a member of the moderation group of doom9, if I'm not mistaken.

[Klaus1189]

Probably, I don‘t know. What is MLID? Maybe obvious but I can not think of anything that matches.

I am really looking forward for Zen3 and RDNA2. Very exciting October.

[NikosD]

MLID is just the acronym of the youtube channel Moore's Law Is Dead.

According to the latest leaks, Zen 3 is ~25% faster in single-threaded performance than Zen 2.

Yes, all signs point to a Red October

[Klaus1189]

My cute RX 5700 XT needs a bigger younger brother, since I wanted a 3080, also to get new information by testing it by myself and I didn‘t get one I am kinda pissed but hey AMD is going strong, I just wait for them.

[huhn]

after your death prediction for nvidia by vega i can't wait for how this will turn out.

[ryrynz]

You and me both. There's more hype on this thread than just about anywhere else on the internet.

[NikosD]

No hype...Just leaks of the performance of Zen 3 CPUs and leaks of the RDNA2 specs.

And much of debanking tech journalists and users of this forum laughing of nVidia's fear regarding RDNA2...

[SamuriHL]

I'm not sure what's happening with my system but I THINK it's related to them messing with the power options and I'm *NOT* happy. I had my machine tuned very well and all of a sudden yesterday I'm getting massive dropped frames when upscaling 1080p->2160p. The decoder queue is all over the place whereas usually it's quite stable. I haven't yet tried a clean driver install but this completely blows. My machine was working fine previously.

[NikosD]

Please, take a look of the comments of this "tech journalist" that I previously posted with his hilarious assumptions regarding Big Navi performance versus Ampere 3080 and the TFLOPS numbers in gaming.
I just read the official review of 3080 from Anthony Garreffa an Aussie gamer who thinks he is a tech journalist at TweakTown.com.

Just quotes of the official review of 3080, if you want to see what hype means in official "tech journalism" of well respected sites.
Now, a sensible person would wait until AMD drops Big Navi -- but there are two types of people here, so let's break it down.

Long time NVIDIA fans: Just like a certain website said back with the GeForce RTX 2080 Ti launch: Just Buy It. If you were happy to spend $1199 on the GeForce RTX 2080 Ti at launch (which was fine, because it was a beast at the time and still was, until today) then you'll be ecstatic to spend considerably less than that and get a huge chunk in performance with the RTX 3080.
Performance enthusiasts: You too, guys -- Just Buy It. Seriously, the GeForce RTX 3080 is a graphics card that enthusiasts like me having been waiting for, for a very long time. Almost no restraints total grunt, 4K 60FPS easy and 4K 120FPS in most games.

And another, even more disturbing:
At this point I don't know why you would wait for AMD... I don't see Big Navi being able to topple the GeForce RTX 3080 although I think it will come close. It should launch with 16GB of RAM (GDDR6 or HBM2, rumors of both are floating around) and see NVIDIA retaliate with 20GB versions of the RTX 3080.

You don't need 20GB of VRAM, so if you wanted to bite the bullet now it's probably your absolute best time to get a graphics card before they sell out and the price increases. If you can buy a GeForce RTX 3080 Founders Edition (especially the FE card, it is a monster) for $699, do it.

**a flaming animated gif inserted in his official "review" so to speak**

https://media1.giphy.com/media/qDPg6HNz2NfAk/giphy.gif

[chros]

You don't have to quote idiots/mercenaries, the internet is full of them.

[NikosD]

Probably you didn't read carefully my post. It happens sometimes, especially in this forum.
It's not just an idiot.
He is a tech journalist, officially the main GPU card reviewer of TweakTown.com
Can you see the difference ?

[el Filou]

Could we at least wait until stuff is released and reviewed before going on and on about performance ?
These last few weeks this thread had more off-topic speculations and rants than actual information on driver issues.

[Klaus1189]

@huhn: Were you successfully in finding the issue"s" which you have with 456.55? I am thinking of the clock deviation, but you wrote that there was more (https://forum.doom9.org/showpost.php?p=1924776&postcount=60266)?
@SamuriHL: Were you able to find the root of the issue? DDUed already?

[DMU]

Recently I noticed in the OS HDR mode the metadata became correct.
Pic (https://drive.google.com/file/d/16lAiIuyXnsS338Y7PSftadfIyB_Ffhen/view?usp=sharing)

[SamuriHL]

No not yet. Will work on it tonight.

Sent from my SM-G975U using Tapatalk

[NikosD]

nVidia officially cancelled the launch of 3070 on 15th of October and pushed it back, just one day after the D-Day, the announcement of Big Navi.

The Earth has already started shaking for the upcoming release...

[huhn]

@huhn: Were you successfully in finding the issue"s" which you have with 456.55? I am thinking of the clock deviation, but you wrote that there was more (https://forum.doom9.org/showpost.php?p=1924776&postcount=60266)?
@SamuriHL: Were you able to find the root of the issue? DDUed already?

can't reproduce it and even if i have a problem it doesn't mean it's the driver.

456.55 makes HDR stop working for me i have no clue why it's not the only thing i updated.