Linus' "Driver improvements over time" video

Final8y · 27 Aug 2015 at 06:20

Ash0787 said:
ye I called him out on that too, I assumed it was just an oversight / lazy rather than an actual decision, maybe they just had more nvidia cards around, you could be right though, I never thought he might have been on the green team

All of his business is using freely provided GPUs from Nvidia, but his reviews have yet to bother me.

kylew_is_back · 27 Aug 2015 at 06:24

After this video I don't think I am going to watch a single Linus's video even again.

Final8y · 27 Aug 2015 at 06:39

thebennyboy said:
Funny how this video appeared shortly after the results of the first DX12 game. He also failed to mention how unstable the recent Nvidia drivers are (at least from what i've heard).

Do Nvidia's new drivers cause the cards to eat more electricity?

Well Hardwarecanucks has just done a driver now and then comparison as well.

GTX 780 Ti vs R9 290X; The Rematch
http://www.hardwarecanucks.com/foru...iews/70125-gtx-780-ti-vs-r9-290x-rematch.html

And its being picked apart here as the numbers in comparison to other reviews just dont added up.
http://forums.anandtech.com/showthread.php?t=2444347

pgi947 · 27 Aug 2015 at 07:13

Linus isn't biased, he is however, very misguided. He gives off the vibe of 'I know what I'm doing' most of the time, when often he doesn't.

His gsync vs freesync video was incredibly flawed.

andybird123 · 27 Aug 2015 at 09:21

The funny thing is that by posting on here and making an issue of it you're actually giving him free advertising and extra clicks he wouldn't otherwise have gotten

I never usually watch his videos as he's an annoying little grunt

And as pgi says, half the time he doesnt even know what he's doing, as in the sync video.

Seanspeed · 27 Aug 2015 at 09:34

What on earth are people complaining about?

He's not comparing Nvidia to AMD. Not the point of the video. It seems it's *you* guys who desperately need to make it some competition.

His conclusion is one that's been pretty obvious, too. People complain that Nvidia have terrible support for Kepler now, but it's just gotten to the point where an architecture built in 2012 simply doesn't have the room for improvement anymore, while Maxwell does still(though even Maxwell's gains are likely to taper off in the near future as well).

David Bisset · 27 Aug 2015 at 09:38

Kaapstad said:
If he is comparing older and newer NVidia drivers I hope he points out the drop in performance on the latest family of drivers.

8 Pack is not happy with the latest ones when he plays his favorite game, PSU Meltdown.

Hope you're ready to get attacked by the rabid hordes for insulting their god!

Seanspeed said:
What on earth are people complaining about?

He's not comparing Nvidia to AMD. Not the point of the video. It seems it's *you* guys who desperately need to make it some competition.

The most interesting thing about performance of drivers over time is comparing between different card generations (by architecture not number ofc!) & different vendors, otherwise it is a pretty pointless video.

Seanspeed · 27 Aug 2015 at 09:45

David Bisset said:
The most interesting thing about performance of drivers over time is comparing between different card generations (by architecture not number ofc!) & different vendors, otherwise it is a pretty pointless video.

It's not pointless at all. It just depends on what you're trying to show. Comparing different generations of architecture would be interesting, but not the point of this. Comparing different vendors would be interesting, but again, not the point of this video.

I think the main takeaway/point is just that driver improvements typically don't see much gain beyond the first 2-3 years for a given architecture. Of course, it depends somewhat on what sort of games developers are building, and also how good the Day 1 driver support is in the first place, but still, it's something worth knowing, because a lot of people seem to think that driver improvements will just keep their card performing better and better forever, like it's some magic, when in reality, there is only so much room for improvement.

WeSbO · 27 Aug 2015 at 11:15

It's funny as I didn't come to the same conclusion as he did on that video he just says "yes drivers improve performance over time", which is somewhat true and we already know that, but what I see is that, release drivers are bad (we already know that) and that NVidia pretty much stopped optimizing the 480 in 2012, gains after that are mostly marginal and down to the margin of error. Also it would have been good to look at other variables than FPS, frametimes for example, min-max FPS etc ...Which in itself wouldn't have added much more work, just filling in a bit more info on spreadsheets.

Now I think this video was asked by NVidia (Nvidia is a LTT official partner cf. Linusmediagroup website, the guy can get as many Titan x's he wants sent to him, that in itself is remuneration) because of the bad publicity they are getting by people saying they are nerfing their cards with the drivers. I think they also asked hardwarecanucks to do the same, they bought out a similar (but different) article about NVidia drivers just a couple of days befor linus, it's a very good article worth the read and interesting (yes the results are very interesting...)

My opinion on Nvidia drivers, is that they bring out too many of them, and performance is everywhere, sometimes you loose fps, then gain fps, and so on, bugs disappear only to reappear afterwards things like that, their drivers aren't consistent in quality, and some bugs stay around for a long long time.

AMD gets a lot of heat concerning their drivers, which is not deserved, they always introduce perf gains (except for launch drivers, which is to be expected) and when bugs are sorted they don't come back. And imo the Win10 AMD drivers are far superior to the NVidia ones, but I think AMD have been working on them a lot longer.

I would have liked to see the same thing with and AMD card as I think the perf gains would be more consistent, as they have to support their cards a lot longer, due to them rebranding, they've just bought out a new card on a 4 year old architechture.

Oh and i'm an NVidia and AMD user, and it will probably stay that way.

And I think the choice of that GPU was not anodyne... It may have been way different on a 780 for example, and it's a shame that we don't know the full options used for each game...

But this video has made me want to do a test of my own, testing each driver since release on a 970 (that's all I have available).

thebennyboy · 27 Aug 2015 at 16:25

Gregster said:
As much as Linus's voice cuts through me and he spends more time on adverts than reviews lately, I found the vid interesting. I don't see where the fanboy part comes from but I suppose if you want to see it that way (as quite a lot of people here do), then you can see anything said/done as fanboyism.

Interesting that performance gains and then tapers off in the last driver tested but some real nice performance gains for the 480 in the games tested.

Good work Linus.

tbh, the bias Linus has towards Nvidia was confirmed when he said on a Freesync monitor review that "I am not giving this freesync monitor an editors choice because it doesn't have gsync", have since unsubscribed. Would rather base my opinion on benchmark threads and discussion from this forum as nobody here is being paid to favour one vendor over another (apart from the vendor reps).

My issues is that he tried to portray the video as showing GPU driver improvements in general, showing only Nvidia driver improvements without showing AMD driver improvements is not "showing GPU driver improvments".

His response to the complaints was pretty poor too. Basically saying AMD's only got 20% of the market so he cba.

drunkenmaster · 27 Aug 2015 at 17:16

pgi947 said:
Linus isn't biased, he is however, very misguided. He gives off the vibe of 'I know what I'm doing' most of the time, when often he doesn't.

His gsync vs freesync video was incredibly flawed.

I don't know what you mean, if you make youtube videos you are automatically an expert in whatever you're making the videos on. Unfortunately that is how most youtube reviews think of themselves and most people who read or watch reviews think they are getting accurate information.

So many of the bigger sites that do podcasts or video reviews spout so much rubbish it's embarrassing. even one of the most respectable people sold his site recently to the company that owns there often biased competition making it even more a dodgy industry. Where one big media owner can and does own loads of magazines, review sites, where they can push the same story but in a different way to different people just by owning/buying every site that gets decent traffic.

andybird123 · 27 Aug 2015 at 17:25

If you mean anandtech why not just say anandtech?

I started reading toms back when it was owned by tom, after it was sold and I earned a holiday from their forum for pointing out one of their articles was three months out of date and the problem resolved (they deleted my comment and "updated" the article), I stopped reading

Seanspeed · 27 Aug 2015 at 17:37

WeSbO said:
It's funny as I didn't come to the same conclusion as he did on that video he just says "yes drivers improve performance over time", which is somewhat true and we already know that, but what I see is that, release drivers are bad (we already know that) and that NVidia pretty much stopped optimizing the 480 in 2012,

See, this is what I'm talking about. This is what the video was basically trying to talk about. You seem to believe that driver optimization is some limitless process that can create infinite performance improvements if you just keep at it.

It's like game code optimization. There's only so much you can do.

Kaapstad · 27 Aug 2015 at 18:03

drunkenmaster said:
I don't know what you mean, if you make youtube videos you are automatically an expert in whatever you're making the videos on. Unfortunately that is how most youtube reviews think of themselves and most people who read or watch reviews think they are getting accurate information.

So many of the bigger sites that do podcasts or video reviews spout so much rubbish it's embarrassing. even one of the most respectable people sold his site recently to the company that owns there often biased competition making it even more a dodgy industry. Where one big media owner can and does own loads of magazines, review sites, where they can push the same story but in a different way to different people just by owning/buying every site that gets decent traffic.

I remember Linus doing a video on multi GPU SLI setups, it was total garbage and I have never taken anything he has done seriously since. The worst part about it is most people can not check out what he is putting in his videos.

WeSbO · 27 Aug 2015 at 22:39

Seanspeed said:
See, this is what I'm talking about. This is what the video was basically trying to talk about. You seem to believe that driver optimization is some limitless process that can create infinite performance improvements if you just keep at it.

It's like game code optimization. There's only so much you can do.

Well he definitely didn't say that optimization more or less stopped after 2012 on that gpu, and he could have said it, but that would not have sounded good, that is definitely not the message he wanted to put through on that video. Optimisations don't stop 18 months after release on amds side (mostly due to a lot of rebrands in the line up) and I do believe that drivers should be optimized for new releases 2 years after the gpu is released, these are multiple hundred pounds gpus not cans of baked beans

WeSbO · 27 Aug 2015 at 22:46

Kaapstad said:
The worst part about it is most people can not check out what he is putting in his videos.

Yes there are never reference points, full game options too look at, they always also say " We use these at maximum overclocked speeds that can be found in the spread sheet, link in the description", yeah that spreadsheet that is never there !

bloodkil · 27 Aug 2015 at 22:53

WeSbO said:
Yes there are never reference points, full game options too look at, they always also say " We use these at maximum overclocked speeds that can be found in the spread sheet, link in the description", yeah that spreadsheet that is never there !

Some of the overclocks in that chart seem kinda crap, doesn't seem like they put effort into it.

Seanspeed · 27 Aug 2015 at 23:00

WeSbO said:
Well he definitely didn't say that optimization more or less stopped after 2012 on that gpu, and he could have said it, but that would not have sounded good, that is definitely not the message he wanted to put through on that video. Optimisations don't stop 18 months after release on amds side (mostly due to a lot of rebrands in the line up) and I do believe that drivers should be optimized for new releases 2 years after the gpu is released, these are multiple hundred pounds gpus not cans of baked beans

I'm still not sure you're getting the point that general driver improvements just don't continue on forever. There is not some limitless potential.

You could spend 2 years optimizing code for a 20% performance gain. You could then spend the next 10 years after that and only find 5%.

Most of the time, even Game-ready drivers offer *potential* performance enhancements for more than just one game. It is mainly general driver improvements, not just very specific enhancements, although that still takes place, especially with games that use specific Nvidia enhancements(which are often based around the capabilities of newer Nvidia hardware, like the heavy tessellation-based effects of The Witcher 3's Hairworks that work best with Maxwell architecture).

Point is, you shouldn't expect performance improvements for years and years and years. Maybe it's not impossible to get it, but if you do, it's entirely possible that it's not because of 'great effort' on the part of the GPU manufacturer, it's because their initial support was just crap and they had a long way to go to make up.

That said, I'd be interested in seeing whether AMD cards from 2010 are still seeing dramatic gains all round, as you seem to imply they would be.

MyBrainz · 27 Aug 2015 at 23:13

Pretty obvious fanboys (both of them) if u check out the '00:50:08 - DX12 will not end Mantle says AMD' section of this video..

https://www.youtube.com/watch?v=c1scTmhAd74

Luke quote on Mantle, "Even then, it really didnt do anything".

Was The WAN Show from a couple of weeks ago.

I just cba with either of them anymore.

thesmokingman · 27 Aug 2015 at 23:14

Linus = the lolz.

Look up his fan review video where he brings a bunch of fans to Corsair's facility to use Corsair's test equipment. Guess which fans win the contest? Surprise, nevermind that he has no clue how the equipment is relevant and that he never tested them in the conditions that is the subject, on radiators!

He should stick to unboxing videos.