Welcome back to Hardware Unboxed. The promised part three of the November Q&A is here, so you can all just rejoice in that knowledge, sit back, relax, and enjoy what's about to unfold. But before that, today's video is brought to you by Sapphire and their Pulse range of GPUs and motherboards. If you're a value-conscious builder looking for a great 1440p gaming experience in a compact system, this is the range for you. The Sapphire Pulse B850M Wi-Fi motherboard supports the latest Ryzen processors, PCIe 5.0, onboard Wi-Fi, and Bluetooth. The matching two-and-a-half slot Sapphire Pulse Radeon RX 9070 GRE 12 GB graphics card features Sapphire's Dual-X cooling, a dual-fan compact cooler that's perfect for smaller builds while still offering excellent performance and value. Despite its small size, the card runs very cool and quiet, which is important for more compact builds. Sapphire's value-oriented pairing is a great option for those looking at a mid-range gaming experience. So for more information, please check the links in the video description.
All right, we'll start with a Discord question here. So do you think NVIDIA may have to go with some light AI performance cards similar to how they did LHR during the mining era?
Um, well, they can't. They basically can't do this, right? Because there's now gaming features that use the AI processing capabilities.
So with like the light hash rate cards, they could nerf mining performance without really affecting gaming at all. So it still made those cards like super viable for gaming. I think people tested them and they found pretty much no difference between, you know, gaming performance on the non-LHR and the LHR cards. But if you nerfed the AI performance of a GeForce GPU, well then your upscaling would run worse because that, you know, uses the tensor cores, same with like generation, the models. And then obviously now Nvidia is offering like DLSS 5 technology that heavily utilizes all the AI processing capability. So if you made like a light AI card, maybe there's some people that don't need any of those features and they'd just be happy with, you know, the...
I think you could call it the Radeon version?
Well, yeah, effectively, right? Um, so people are going to be so mad at that. They're going to be so mad. Anyway, um, so yeah, I mean they basically can't do this. And whether they would want to, I'm not sure because I think with the LHR stuff, they made them because they weren't... were they making more money off like the full mining version? I thought they were just kind of annoyed that, you know, all the AIBs and distributors and stuff were scalping cards.
Yeah, Nvidia wasn't on the ground floor for that.
Adapted to their surroundings.
They weren't exactly circular financing the coin miners.
Like we didn't create that environment. We didn't create that gold rush, but we sure as hell are going to create the next one.
Yeah, and so they can't really limit AI when they're sort of all about AI.
No, they're sort of... they're limiting the amount of CUDA cores they give you because AI.
Yeah, and Nvidia likes... like let's be honest, Nvidia likes the situation where they can sell you an RTX 5090 for $9,000 and then discontinue it to make a $16,000 Pro GPU. Like that's a great situation for Nvidia. They would love to be able to sell every GPU for $16,000 if they could, but they haven't previously because there's only so much workstation demand historically, and so they have to sell gaming GPUs at a lower price. But right now they don't have to do that. They can just sell cards for absurd prices. Same with like 5080s being $2,000. They don't care that people use them for AI instead of gaming. That's a $2,000 sale for them. So they've got no incentive whatsoever to make like a cheaper AI limited version for customers.
They've helped create the perfect storm for themselves.
Yeah, they don't need to do that, so they won't. And apart from the fact that it wouldn't work, they also won't do it. So that's currently the situation that we're in.
This is a different question. I quite like this. What's your take when you benchmark a game just in time for release day and a couple of days later, sometimes even the same release day, there are new drivers or game fixes that invalidate your data? Like I said, interesting question. I actually don't think this scenario plays out all that often. It's actually very rare. What normally happens is AMD, for example, won't have a Radeon driver ready for the game and Radeon performance looks a bit subpar, and then when the game rolls out, either at the same time or a day later, they'll release a Radeon driver that everyone then says, 'Oh, your data's invalidated. You should have waited until AMD got the driver out.' And then we test that driver and it makes zero difference. That's what normally happens.
Um, that's not to say that's what always happens. Sometimes the scenario that you've laid out does happen, so the game either gets like a day one patch that noticeably improves performance. In those instances, I don't think it does in one way invalidate the data, but not totally, because often the margins that we saw between various models and various brands stays true. So if GeForce is 20% faster, it often remains 20% faster. Not always. Sometimes a day one patch will correct like a bug with, say, Radeon in that particular scenario. But again, we don't see that happen too often. Like Borderlands 4, for example, it was like, was it six months before we got proper performance patches that solved issues with that game despite the fact that they swore that it was optimized as optimized could be. Um, so usually, I would say the majority of the time what happens is the data we show you pretty much holds true.
Um, it's very rare. People think that the new driver does magical things and sometimes even people get better performance because either there was a problem with their system previously or whatever. You know, it can happen and my thoughts when it does happen are, 'Oh bother, that's annoying.' But what do you do about that? So if we're given early access to a game, you know, AMD has a driver, Nvidia has a driver, and we test it because more often than not it's useful information and even if the game does get updated you sort of see where, you know, what it was coming from. And yeah, but it, you know, the data can get invalidated, but that's true of everything. It's like when we get a product to review that can happen. There have been instances where products have gotten better shortly after release because of updated drivers or, you know, firmware or whatever it may be. Even sometimes Windows updates can improve performance. So it's just a thing that you have to deal with. Sometimes it happens, well, a lot of the time it doesn't happen. That's probably the best way to put it. Sometimes it happens, more often than not it doesn't happen. It's like what you see is what you get. Um, but you know, we've even seen 12 months down the track that, you know, driver updates have improved things so...
Yeah, I think people potentially overestimate what the game ready drivers actually do in terms of performance for a lot of titles. You know, obviously Nvidia makes a list of all the games and AMD does this as well with like, 'this driver supports these games.' That doesn't mean that the driver prior to that version is broken in those games. It just means that they've sort of validated that, hey, the expected performance is, we've checked the game in this driver, we know it works, we're going to make it game ready. But with a lot of game development, you know, the developers are working to make these games work on prior driver versions, right? Because they're testing the games months before launch on some form of GPU driver. So if you go back across history and you actually start testing, you know, let's go back one driver and then two drivers and three drivers prior to the ready driver, you'll often find there's quite a lot of drivers that deliver basically the same performance, but then there'll be one driver that if too far back, the game is like fully broken or there's some major issue.
Performance degrades. More often than not, what we actually see in those situations is the latest driver. So it's more like a hot fix. So there'll be maybe some certain models or certain configurations that are unstable with the existing driver.
Usually it's a stability thing or a visual thing that it solves more so than a performance thing. We see that more often than not that that's what the newer driver is addressing rather than some big performance improvement. Because yeah, by the time you get to that stage, it's rare that there'll just be, oh, overnight we solved this thing that gives you heaps of performance. It's usually the game is quite... well, no, sometimes the game is quite well optimized out of the box. Usually it's more of a Borderlands 4 situation where, you know, months after release they start to improve performance and that's when you see the data become, you know, invalidated, but it was valid for all the people that bought the game straight away and wanted to play it.
Yeah, I think for some actual examples, I believe with Forza Horizon 6 there was an Nvidia GPU driver that actually did improve performance that was released maybe a day or two before the game was released, but you did include that data in the final version.
Yep. That happened with Doom as well, I'm pretty sure for AMD maybe.
Yeah, but then I think a game like Black Flag Resynced, there was a driver that was released shortly before the game came out again that didn't change the performance at all.
No, I had to go back and retest. Like I had to test entry level, mid-range, high-end, both brands to make sure.
Yeah, and as well, I think with this situation of like we might make a performance review of the game and then suddenly something has changed, you know, at launch or something like that. I think it does provide a good incentive for developers and all the driver manufacturers to have these things ready ahead of time. Because we've seen situations this year where AMD, for example, they don't have a quote-unquote game ready driver for a game when the game is actually launched. So you might have to wait like a week for the...
Was it the case of E Day, the game that they sup-
There was some silly AMD...
Was it the Witcher 3 remaster?
I think it... yes, yes, it was the Witcher 3 because that was absurd, that situation, because a week before the game came out I had the working driver and again it was less performance and more graphical glitches.
AMD provided me with a flawless driver, no crashing, no graphical issues, and when the game came to market it was like 24 hours the existing driver had like half the graphics missing.
How does that situation happen?
So yeah, I think it provides a good incentive because sure, we want our video to accurately show the performance that you guys will be getting when you play the game, but it is also sort of, especially when we get early access to games, which this is what this is really applying to, right? Is the performance testing we do with the early access review copies of games. It is also in some ways a review of the performance of the game. Are we seeing good numbers, so good FPS across a wide range of cards or poor numbers? And if the result is that one particular GPU brand is underperforming or the performance in general is bad, then that is giving an incentive to the developers to sort those things out so that reviewers... because it's not just us doing the performance testing, a lot of game reviews these days are done on PC copies, not like in the old days where everything was a console review and then the PC version would just come out and it would be all sorts of ways of broken and no one had seen that beforehand because they just didn't offer PC copies for review. These days PC copies are quite readily available. Um, a lot of these game developers are willing to give them for either performance testing or just reviewing purposes. And so it's in their best interest to make sure that there isn't some game fixing patch on day one that avoids the review process because then reviewers are going to say stuff like, well, this is broken, that's broken, the performance sucked, it was unoptimized, which is not in their best interest. And then similarly, if there was like Nvidia had their driver but AMD didn't and so AMD underperformed in the charts, that's a signal to AMD to not release their drivers late and so they don't have this situation where it appears as though in the performance review of the game that GeForce is way better. So there is some incentive there for them to sort of solve those problems.
With AMD, they'll actually fix the driver but then not release it.
Yeah, which is just unbelievable stuff. So yeah, it's obviously not ideal if things do get invalidated because of patches and things like that. And it has happened where across a review period there are things that fix glitches and things like that. So occasionally we do try and caveat some of the videos by saying, hey, we noticed this issue but we were testing the review copy of the game. We don't know whether this will be a solved issue. And some developers are quite good in giving you as part of a review guide a list of known issues. So if you do come across, let's say there is a specific graphical glitch, you can check it against the issues that they're saying they know about and they will fix for the day one patch. They might not fix it of course because it's sort of a future promise thing, but if they're aware of it then they might fix it essentially.
Yeah, that's a good point to bring up because we often caveat the reviews by saying, or that the game reviews by saying, look, this is typically how a 9070 XT will compare to a 5070 Ti but it's miles ahead or behind in this game suggesting that one brand, you know, needs to optimize. And we can see that by comparing when we usually do those sort of tests. It's like I think 40 to 50 GPUs. So you get a good idea of what generations need work, you know, relative to where you'd expect them to perform. So it's not like we just go, yep, you know, GeForce is 40% faster in this game. That's obviously going to be a thing that stands for months to come. And like there's clearly some work there that AMD need to do or vice versa, however it plays out. So there are also things that we take note of because we know roughly how everything should perform based on extensive past testing.
Yeah, and as well, there have been game releases where Nvidia, AMD, or both of them have had like preview drivers available for the testing period. So they'll have like basically, like you saw with the Witcher 3 remaster situation, reviewers can download an early version of the driver that will go live on day one so that we're testing essentially the final version of that driver. So I think with Gears of War E Day and especially with Nvidia titles where they've got maybe a specific feature that's coming to that game, um, they will have like an early access for that driver version. We see that quite commonly. And so even if you're sort of seeing, hey, you've released your video on that game and then the driver comes out the next day, it might actually be the case we've already tested using that driver. Um, and I think that's a really good way for these companies to go about things basically is, you know, if they are going to have some day one driver patches, just make it available during the review period of the game for reviewers. Um, that's really beneficial to everyone, I think.
Mhm. All right, this will probably be a bit of a quick one. Do you think it's still worth buying RDNA 2 GPUs after AMD gave RX 6000 series FSR 4 support?
It's certainly more worth it, but you know, RX 6600, I guess if you're getting that dirt cheap and you want to play really basic games, they might still be worth buying. Uh, the higher end models can still be quite capable. We've seen 6800 XTs, which is basically a 7800 XT. They still perform quite well. So adding FSR 4 support's great, but yeah, it's going to come down to the pricing and I guess what's available in your region. So you'll have to work that out. Does a 6800 XT at this price make sense relative to what else you can buy? But yeah, obviously adding in important features like FSR 4 upscaling to those GPUs certainly adds to their viability.
Yeah, I mean this is quite an old architecture now, so how willing are you to rush out and buy a used version of those cards is, you know, it's questionable and especially because the used market, a lot of things it does often normalize to performance, so it may not be in your, you know, your example of the 6800 XT being similar to a 7800 XT, it may not cost that much more to get a newer architecture. So again in the past, you know, for a while now, well how many months? A couple of months, you know, 7800 XT had access to FSR 4.1, the 6800 XT didn't. So did that make it much more worth it to buy a 7800 XT probably during that time? Now it makes a 6800 XT a bit more viable as well. But again on the used market it really comes down to pricing and that sort of thing, but yeah, it makes it more viable, right? Like prior to this release I wouldn't have been rushing out to buy like a 6950 XT for example.
Again it comes down to price though, like if it was dirt cheap...
Like I could forgo the newer upscaling model for that kind of discount. So that's why the best way you can answer the question is yeah, it's made them more viable, but depending on the model and the price of the model whether or not you should buy relative to what else is there is difficult to say.
Yeah, of course and like a 6600 is probably pretty slow in a lot of games, so even though it has FSR 4.1 and yeah, you can do upscaling at 1080p and it doesn't look like absolute garbage like it used to, um, it's still not going to be, you know, there's other issues with that card that make it not as viable.
Yep, yeah, come down to the price.
Despite AMD being very competitive with the past few generations, there's still a large amount of gamers that think they're terrible at making GPUs. How long until mindshare starts to change?
Yeah, good question. I think mindshare this generation has definitely improved in AMD's favor. Previous generations it might have gone more the other way. I think what we've said, and I think this is true, is you need consecutive generations to be a great success. And AMD's not been able to do that. Whereas they did it with Ryzen and they did that, you know, they found avenues to success, you know, different avenues. So when they weren't really competitive on at least gaming performance, they still managed to appeal to gamers en masse by making them just dirt cheap. Like, you know, throwing 2700X's out the door.
Cheap, platform support, unlocked overclocking, all that stuff.
Yeah. So they just got all their ducks in a row. Everything you needed to make it a success, they basically ticked all those boxes. Whereas with Radeon, they just... they're not able to do that. They just can't do it for whatever reason. Obviously they have, you know, very stiff competition with Nvidia, which is a big problem. Whereas Intel was floundering about and have been really ever since. Maybe Nova Lake will change that for the first time in a long time. But Intel's been on struggle street and AMD just did everything right that they needed to do.
Yeah, much more powerful foe in Nvidia. Yet despite that, they've still just managed to trip over themselves every few months. They seem to make a boneheaded decision um, on the regular. And we've talked about stuff like a 9080 XT, whether that was a missed opportunity. I think going back now and seeing how everything played out, they probably would release that product. And that's more of a, again, a mindshare halo type product. So, you know, if you look at benchmark charts and try and put yourself in the shoes of a normie type person, someone, you know, that's not a massive tech enthusiast, you look at most charts where you can find charts and you'll see RTX 5090 and 4090 and 5080 and maybe 5070 Ti and then oh there's a Radeon GPU and that just makes them look pretty crap. It's like if you're looking for a sports car, you've finally got sports car money and you're looking at maybe towards a supercar and you feel like brand here and then brand here, like option one, option two, option three, and you get down to like option seven, you're like, hmm, those last three options are pretty similar money and this brand has a lot of the top options. I think I'll go with one of them. They seem to be the competent supercar maker. And I guess that's the kind of thing that plays out with GPUs when you see Nvidia GeForce dominating the top like five, six spots, that's your mindshare right there and AMD just for whatever reason hasn't been able to crack into that. And then there's all the other problems we haven't talked about like, you know, not supporting products long enough or well enough or, you know, you can argue till you're black and blue in the face about what the maintenance mode actually meant. Doesn't matter. Like it's just an idiotic conversation to be had whether you're like, that's the normal way that things happen. It's just not a good look. Just it can be... most normal people are not going to interpret that the way that you think it's meant to be interpreted if you're defending AMD on that one. You don't put something into maintenance mode that you want to continue supporting at the forefront. It's not a good look, not a good way of going about it, and having the split branch thing's not something you can... your competitor who's in an advantageous position, they're not doing split branch mode for their older products. Just don't do dumb stuff like that. But that's just one of countless examples of the way that AMD Radeon continue to fumble the bag.
Yeah, and I think part of this question, it almost comes across as if like maybe gamers were wrong to think that previous Radeon generations weren't great and therefore, you know, when are we going to see the mindshare happening because they've been executing for a while. But I think if a gamer is sort of looking at the past couple of Radeon generations prior to RDNA 4, I don't think it's really all that outlandish to say that they made some products that weren't very good. And so it would be fair for a gamer to sort of say, well, GeForce was better. I mean, and it comes down to things like, in addition to all the stuff you mentioned, things like the 7900 XT having a really terrible price upon release relative to its performance. You know, the 7900 XTX underwhelming in terms of where it should have been in its performance, where the expectations were and what it actually delivered. And all the stuff you covered in the review about how that advertised a certain level of performance and it did underperform that level. The 7700 XT had issues. The 7600 was just basically a re-release of a fairly underwhelming product at a not very good price. And that's not even including all the features that they didn't support that Nvidia had at the time, things like DLSS upscaling. At the time, AMD was still on, you know, FSR 3 level technology, broken frame generation in terms of frame pacing and stuff like that with their technology, a lack of a ray reconstruction competitor. And so I think it's definitely fair to say that during that generation, some people would have looked at AMD versus Nvidia and come to the conclusion that the AMD product was not very good. And so...
Or at the very least inferior.
Yeah, and so you can't build mindshare when you legitimately have products that are not worth buying. Like mindshare begins from people actually wanting to purchase their products. And if you have a $900 7900 XT with lackluster features, it's not going to move the needle no matter how much of an AMD fanboy you are thinking that, oh, that was actually a good product and it's aged well because it's now got FSR 4 and all that sort of thing. At the time, you could understand perfectly well why GeForce was able to extend their market share. And so as you said, they need to, once they start executing on, you know, not stuffing out the price, actually having the features, not just depreciating the cards within a generation because they feel like it. As soon as they stop doing that stuff, and they have this generation where people want to buy them, they've executed well, customers feel like they've been supported with additional features and stuff like that, and they do that for a few generations in a row, that's when mindshare builds. You know, in a few generations time, if they keep executing, no one's going to care that it took them five years to sort out upscaling on RDNA 2. It's not going to matter anymore. And so people aren't going to be like, oh yeah, but you know, how bad was AMD in 2020? Like it'd be irrelevant in 2030 if they keep executing. So that's how mindshare is built. That's what they did with Ryzen, but as you've pointed out, they haven't been able to do that. It's only been one generation here and there that sort of has captured people's attention. Um, hopefully they're in a better position now, but again, it's AMD, so we have no idea. They could just do AMD stuff as we've said.
All right, interesting question here, Tim. Do you think manufacturer exclusive features are a thing of the past with AI helping reverse engineering features? Like we have DLSS on AMD now for an example.
This has been a really interesting development because one way of looking at it, right, is Nvidia has sold cards, like graphics cards, for the last, I don't know, seven or eight years now based on exclusive features, ray tracing, DLSS, frame generation, etc., right?
At the same time, Nvidia has been pushing AI super hard.
Data centers, workstation products, everything's AI now.
And it's come to the point where the thing that they were using to sell their gaming graphics cards has been undermined by their own technology.