WEBVTT 00:00:00.000 --> 00:00:05.120 Two years ago, Graphics Giant NVIDIA caused a wave of outrage when they announced two 00:00:05.120 --> 00:00:10.440 gaming GPUs that had the same name, but seriously different performance. 00:00:10.440 --> 00:00:15.040 Thankfully, though, they learned their lesson from that response, and they never did it 00:00:15.040 --> 00:00:19.600 again. Sorry. They did do it again. 00:00:19.600 --> 00:00:23.640 They did it again less than a month later, and they're still doing it today. 00:00:23.640 --> 00:00:27.640 The only difference, the only difference, is that they're keeping this behavior out 00:00:27.640 --> 00:00:33.120 of the headlines by doing it on the budget GPUs that the media usually ignores. 00:00:33.120 --> 00:00:39.200 For example, the RTX 3050 6Gig, did you even notice when this thing was released? 00:00:39.200 --> 00:00:45.760 I don't blame you. Rather than creating a product guide and sending samples to reviewers for testing, NVIDIA announced 00:00:45.760 --> 00:00:50.160 this thing by writing a short blog post about it like it was 2009. 00:00:50.160 --> 00:00:54.600 As a result, very few reviews were made, and even fewer people watched them. 00:00:54.920 --> 00:01:00.320 You could say NVIDIA probably did that because they knew that no one was going to be excited 00:01:00.320 --> 00:01:06.080 about a relaunch of the same GPU from two years ago anyway, but I think they did it 00:01:06.080 --> 00:01:10.880 that way because they didn't want anyone to notice that it is not the same GPU from 00:01:10.880 --> 00:01:14.400 two years ago. How do they think that this is okay? 00:01:14.400 --> 00:01:20.520 It isn't okay. And to show you just how not okay it is, we are finally going to be looking, not just 00:01:20.600 --> 00:01:26.440 at the 3050 6Gig, but we're also going to be putting a spotlight on the many times in 00:01:26.440 --> 00:01:31.640 the past that NVIDIA has pulled this dirty move, stretching all the way back to the 00:01:31.640 --> 00:01:34.760 beginning. And we're also going to put a spotlight on our sponsor. 00:01:34.760 --> 00:01:39.120 If CUDA cores and memory buses don't mean much to you, don't worry, because all you 00:01:39.120 --> 00:01:45.760 really need to know is that when we compared the new 3050 6G to the original 8G model, 00:01:45.760 --> 00:01:51.200 we found that in both real games and in benchmark apps, there was a big enough difference that 00:01:51.200 --> 00:01:57.560 a more appropriate name for this new car would have been RTX 3040 or even 3030. 00:01:57.560 --> 00:02:03.160 This calls into question, what even makes a 3050 a 3050 in the first place? 00:02:03.160 --> 00:02:08.320 It clearly doesn't mean the specifications and it doesn't even mean the performance. 00:02:08.320 --> 00:02:13.880 And what's really frustrating for me is that by drawing this comparison to the original 00:02:14.000 --> 00:02:19.160 3050, it makes us think, oh, this is a really bad card, but it's not. 00:02:19.160 --> 00:02:23.800 It's actually the best GPU on the market that doesn't need extra power cables. 00:02:23.800 --> 00:02:29.240 So if I were, say, upgrading a Dell OptiPlex for some serious gaming business, it could 00:02:29.240 --> 00:02:38.800 actually be a really great option. It just isn't an RTX 3050 option, and calling it that is misleading. 00:02:38.800 --> 00:02:44.400 All NVIDIA had to do to avoid this video ever being made is give it an appropriate name. 00:02:44.400 --> 00:02:48.960 Let's talk about why they didn't, and fair warning, you might not like what you hear. 00:02:48.960 --> 00:02:54.160 The thing is, budget GPU shoppers are much less likely to read or watch a review before 00:02:54.160 --> 00:02:59.440 they click checkout. I know this from my experience working retail, but even if you don't believe me, look at 00:02:59.440 --> 00:03:06.600 the numbers. The original RTX 3050 got the full launch treatment from NVIDIA, and if we look at the Steam hardware 00:03:06.640 --> 00:03:12.680 survey, we can see that that card sold extremely well, more than the high-end RTX 3080. 00:03:12.680 --> 00:03:18.880 And yet, the 3050 got fewer views across the board for the media that covered both cards. 00:03:18.880 --> 00:03:25.800 And that kind of makes sense. I mean, anybody would put the extra research into a $700 purchase, but wouldn't necessarily 00:03:25.800 --> 00:03:32.560 on a couple hundred dollar purchase. So mainstream shoppers tend to rely more on salespeople at big box stores for help choosing 00:03:32.560 --> 00:03:38.640 the right product. And if you've ever tried to do that, you know that having someone read the label on 00:03:38.640 --> 00:03:41.640 the shelf to you is often the best you can hope for. 00:03:41.640 --> 00:03:48.800 And I don't even blame those sales reps. They're paid to sell things, not to be experts on all of the stuff in their spare time, and 00:03:48.800 --> 00:03:57.800 there's a lot of stuff that they're expected to sell. So I think that by reusing the RTX 3050's name on an inferior product, NVIDIA has made 00:03:57.800 --> 00:04:06.520 it confusing on purpose. We've created a situation where even a tech-savvy sales rep that is trying could think that, 00:04:06.520 --> 00:04:12.040 well, hey, as long as you don't need that extra memory, say to game at a higher resolution, 00:04:12.040 --> 00:04:16.280 this version with less memory should perform the same and save you a buck. 00:04:16.280 --> 00:04:24.240 And it gets even worse. Even for the folks who attempt to do their own research, by reusing the same name, NVIDIA 00:04:24.240 --> 00:04:28.680 ends up confusing them too. Look at these results on YouTube. 00:04:28.680 --> 00:04:31.920 Sponsored Junk. This is fine. This is also the right card. 00:04:31.920 --> 00:04:36.520 Sponsored Junk. Wait. That's a video about the original 3050. 00:04:36.520 --> 00:04:41.520 So I could search for the exact correct product name, pick the first highly viewed video I 00:04:41.520 --> 00:04:46.640 see, and I could think the card I'm buying is, I don't know, basically that, but with 00:04:46.640 --> 00:04:53.920 less memory. Now, if this was the first time it happened, I might say, oh, NVIDIA, you made it silly. 00:04:53.920 --> 00:04:58.080 But as I've already said, this is part of a pattern of shady behavior. 00:04:58.080 --> 00:05:03.560 Let's pivot to the RTX 3060 8 gig, which is another example that happened less than 00:05:03.560 --> 00:05:07.000 a month after the 4080 12 gig scandal. 00:05:07.000 --> 00:05:10.600 On first glance, this one doesn't seem that bad. 00:05:10.600 --> 00:05:15.720 Then you notice that when they cut off a third of the memory, they also cut off a third of 00:05:15.720 --> 00:05:19.400 the memory bus, which slows all of it down. 00:05:19.400 --> 00:05:25.120 So in both games and benchmark apps, even ones that didn't need the extra video memory, 00:05:25.120 --> 00:05:32.320 we measured a big performance difference. Now the saying goes that there's no such thing as a bad product, only a bad price. 00:05:32.320 --> 00:05:37.600 So this one would have been more okay if NVIDIA had also provided a discount, except that 00:05:37.600 --> 00:05:45.040 they didn't. The cut down RTX 3060 launched at the same MSRP as the full fat card. 00:05:45.040 --> 00:05:49.480 Why? Well, because it was during the GPU shortage and NVIDIA thought they could get away with 00:05:49.480 --> 00:05:54.040 it. Unfortunately for them and good for everyone else, the shortage was ending around this 00:05:54.040 --> 00:05:59.160 time and regular RTX 3060s soon came back down to their normal pricing. 00:05:59.160 --> 00:06:03.400 So thankfully there aren't that many of these manufactured e-waste cards out there. 00:06:03.400 --> 00:06:07.160 Now to be clear, lots of other companies have misleading names for their products. 00:06:07.160 --> 00:06:11.560 We just finished calling out NVIDIA's rival AMD for a similar behavior. 00:06:11.560 --> 00:06:16.440 But as the leader in the market, I feel that NVIDIA has a responsibility to set a good 00:06:16.440 --> 00:06:21.520 example, to put the bar high and instead they've done the opposite. 00:06:21.520 --> 00:06:28.320 NVIDIA at every turn, basically forever. And this is a real problem with real world consequences for the consumer. 00:06:28.320 --> 00:06:32.000 I personally have been bamboozled once by NVIDIA. 00:06:32.000 --> 00:06:37.320 I bought an N-Force motherboard for my AMD CPU, this was back in the early 2000s. 00:06:37.320 --> 00:06:41.600 And a big part of the reason was that it had GeForce 4 graphics. 00:06:41.600 --> 00:06:48.840 Well, imagine my surprise when I learned that GeForce 4 MX is not in fact GeForce 4. 00:06:48.840 --> 00:06:54.440 GeForce 4 MX was more like a GeForce 2 than a GeForce 4. 00:06:54.440 --> 00:06:58.240 Now we're not going to have Labs actually benchmark a 20 year old motherboard to show 00:06:58.240 --> 00:07:03.200 you the exact differences, because why would we when we can pick out more recent examples? 00:07:03.200 --> 00:07:06.720 Let's look at the GTX 660 from 12 years ago. 00:07:06.720 --> 00:07:12.000 This one was a mess, or rather these two are messes. 00:07:12.000 --> 00:07:18.720 Both of those cards are called GTX 660, but under their coolers are completely different 00:07:18.720 --> 00:07:26.040 chips. Plot twist. Their performance is not nearly as different this time, less than 10% in Tomb Raider 2013, 00:07:26.040 --> 00:07:33.560 and we see the same in our benchmark apps. But here's the thing, if I expect 10 donuts and I get 9 donuts and a half eaten donut, 00:07:33.560 --> 00:07:41.040 I'm still pretty ticked off. And depending on which GTX 660 you get, it could be even further off than that. 00:07:41.040 --> 00:07:46.920 On tech power-ups list, we found a whopping four versions of this GPU, but then none 00:07:46.920 --> 00:07:51.960 of them matched this card, which has more memory than any of them. 00:07:51.960 --> 00:08:00.080 And it's hopeless to find a pattern in this. Sometimes NVIDIA seems to make an honest effort to match the performance between their dissimilar, 00:08:00.080 --> 00:08:07.240 but same named cards. As we were writing this video, an early sample of an RTX 4070 with just 10 gigs of memory 00:08:07.240 --> 00:08:12.400 was found, and in this case NVIDIA boosted other specs in order to make up for the lack 00:08:12.400 --> 00:08:15.160 of memory, and they've done that sort of thing before. 00:08:16.160 --> 00:08:22.240 As we're finishing this video, they did it again, and this time it's both better and 00:08:22.240 --> 00:08:29.840 worse. A new RTX 4070 just came out with a slightly slower and cheaper GDDR6 memory instead of 00:08:29.840 --> 00:08:33.720 the GDDR6X the 4070 is supposed to have. 00:08:33.720 --> 00:08:40.600 Now that's about 95% as fast as the original, so to be fair it's not too far off, but what 00:08:40.600 --> 00:08:45.800 makes it bad is that the cards aren't going to have a different label to let you know, 00:08:45.800 --> 00:08:48.800 hey, you're buying a slower card. 00:08:48.800 --> 00:08:51.960 Why do these alternate versions of cards exist? 00:08:51.960 --> 00:08:57.120 Now the obvious answer is corporate greed, and sometimes Occam's Razer does hold true 00:08:57.120 --> 00:09:02.400 here. But other times their product strategy is actually pretty reasonable, even if I don't 00:09:02.400 --> 00:09:10.400 agree with the way they present it. The 660 chip change, for example, was done to salvage higher end dies that didn't make 00:09:10.400 --> 00:09:15.200 the cut and otherwise would have been wasted, so they were cut down a bit more and turned 00:09:15.200 --> 00:09:22.240 into a lower end product. Alright, no problem, but hey, if they were all reasonable and clearly labeled, we wouldn't 00:09:22.240 --> 00:09:25.880 have had to make this video, and this is where things get ugly. 00:09:25.880 --> 00:09:30.120 We've complained before about the way NVIDIA sells their laptop GPUs. 00:09:30.120 --> 00:09:35.240 Not only do they use the same names as their desktop counterparts, even though the mobile 00:09:35.240 --> 00:09:42.200 RTX 4090 is more like a 4080, but there's also the problems of them using much slower 00:09:42.200 --> 00:09:48.360 memory and a performance limiter based on the cooling and power, a limiter that varies 00:09:48.360 --> 00:09:52.520 from one laptop design to the next depending on the manufacturer. 00:09:52.520 --> 00:09:57.920 As we've seen, this limiter can drop performance on these chips by an entire product tier, 00:09:57.920 --> 00:10:00.960 all of which leaves us with the most important question of all. 00:10:00.960 --> 00:10:06.600 If the name of the product doesn't mean anything about the GPU you're buying, including the 00:10:06.600 --> 00:10:12.120 performance, then what the ever-loving bum does it mean? 00:10:12.120 --> 00:10:18.840 As of 2024, it seems like it just means you pay us this much, and if you're not obsessed 00:10:18.840 --> 00:10:23.160 with tech news, you'll get, well, whatever you get. 00:10:23.160 --> 00:10:26.400 Oh hey, I got a message from our sponsor. 00:10:26.400 --> 00:10:30.400 If you enjoyed this video, go check out the one we did recently about the confusing nature 00:10:30.400 --> 00:10:34.880 of these names in the first place, something that the whole industry has been doing a really 00:10:34.880 --> 00:10:39.760 bad job of because bigger numbers sell better even if it's not actually better.