1 00:00:00,000 --> 00:00:03,720 Picture this, you're gaming on your very average gaming PC, 2 00:00:03,720 --> 00:00:06,840 and it's all right. But then, 3 00:00:06,840 --> 00:00:10,520 Inheritance Strikes, your long lost Uncle Milburn, 4 00:00:10,520 --> 00:00:15,600 who you know well enough to make small talk with, but not well enough to be that sad about, 5 00:00:15,600 --> 00:00:18,960 left you 3,000 US dollars. 6 00:00:22,320 --> 00:00:26,560 Well, this is perfect. Everyone knows that if you're on a limited budget, 7 00:00:26,560 --> 00:00:29,560 the best thing is to get a good enough platform, 8 00:00:29,560 --> 00:00:32,760 and then dump the rest of your money into your GPU. 9 00:00:32,760 --> 00:00:36,280 But here's the thing about stuff that everyone knows. 10 00:00:36,280 --> 00:00:41,320 Sometimes, everyone is wrong. It's been a while since we've explored the effects 11 00:00:41,320 --> 00:00:46,840 of bottlenecks on the latest and greatest GPUs, and now that we've got a new, latest, and greatest, 12 00:00:46,840 --> 00:00:52,520 I think it's high time for us to ask, is it really worth spending this kind of money, 13 00:00:52,520 --> 00:00:56,920 or could we get most of the benefit for a fraction of the cost? 14 00:00:56,920 --> 00:00:59,080 Something, something, our sponsor. 15 00:01:09,160 --> 00:01:12,920 If you haven't seen our 2025 most average gaming PC yet, 16 00:01:12,920 --> 00:01:16,560 the TLDR is this tower here has a six core Core i5, 17 00:01:16,560 --> 00:01:20,760 16 gigs of memory, and an RTX 3060 with eight gigs of VRAM. 18 00:01:20,760 --> 00:01:24,480 It's pretty mid, but hey, that's what we were going for. 19 00:01:24,480 --> 00:01:29,800 So let's game on it to establish a baseline. Kicking things off at 1080p, less demanding games 20 00:01:29,800 --> 00:01:32,840 like F124 and Anno 1800 managed to hit 21 00:01:32,840 --> 00:01:37,320 around a hundred frames per second at high, and very high respectively. 22 00:01:37,320 --> 00:01:40,560 But as we move into more demanding titles 23 00:01:40,560 --> 00:01:46,160 like Cyberpunk and Returnal, let's just say that I would probably have dialed 24 00:01:46,160 --> 00:01:50,840 some of these settings down if I wasn't about to compare this setup to a 5090. 25 00:01:50,840 --> 00:01:56,880 As for 4K, once again, Anno is fine. I mean, 30 FPS is plenty for a game like this, 26 00:01:56,880 --> 00:02:00,120 but everything else is a little rough. 27 00:02:00,120 --> 00:02:06,000 So now that we've got some numbers to compare to, it's time to spend dear Uncle Milburn's cash money. 28 00:02:09,760 --> 00:02:13,440 I'm gonna miss that son of a b***h. Aside from being effective, swapping your GPU 29 00:02:13,440 --> 00:02:18,040 is one of the simplest upgrades that you can make these days. Now that pretty much everything else 30 00:02:18,200 --> 00:02:21,320 is buried under coolers or motherboard armor, 31 00:02:21,320 --> 00:02:24,920 but that doesn't mean that it's entirely without its hazards. 32 00:02:24,920 --> 00:02:30,600 Releasing the PCIe latch to get your old card out can require some serious contortion, 33 00:02:30,600 --> 00:02:34,440 although, oh nice, this one has the little button, so I don't have to worry about that. 34 00:02:34,440 --> 00:02:37,880 And the jump from the 3060 that we have in here now 35 00:02:37,880 --> 00:02:43,240 to this 5090 is a very significant one, and not just in terms of weight. 36 00:02:43,240 --> 00:02:47,480 You'll need to ensure that your case can accommodate the space and cooling requirements 37 00:02:47,480 --> 00:02:51,880 of your bulky new card, not to mention that if your power supply is on the older side, 38 00:02:51,880 --> 00:02:55,240 it might not have native support for this new fangled power connector 39 00:02:55,240 --> 00:02:56,440 that probably won't melt. 40 00:02:58,440 --> 00:03:03,240 Luckily, Elijah went a little overboard when he configured the most average gaming PC's 41 00:03:03,240 --> 00:03:07,240 power supply this year, so we've got plenty of power for our new 5090. 42 00:03:07,240 --> 00:03:11,360 But if he hadn't, we'd be looking for something with a 12 volt two by six connector 43 00:03:11,360 --> 00:03:16,960 that's rated for 600 watts, plus some extra capacity for all the rest of our system, 44 00:03:16,960 --> 00:03:21,480 which, okay, in our case, isn't much, but your mileage may vary. 45 00:03:22,480 --> 00:03:26,400 Right out of the gate, that is quite the upgrade. 46 00:03:26,400 --> 00:03:29,960 Starting at 1080p again, this is much better, 47 00:03:29,960 --> 00:03:34,180 although it is worth noting that we're at ray tracing medium 48 00:03:34,180 --> 00:03:37,960 and only 1080p on a 5090, 49 00:03:37,960 --> 00:03:41,440 so this falls well short of where I would expect it to be. 50 00:03:41,440 --> 00:03:45,200 With that said, I mean, it's a lot better than an ON F124 51 00:03:45,200 --> 00:03:48,240 who only jumped by about 70%. 52 00:03:48,240 --> 00:03:54,200 Again, I can't emphasize this enough. Going from a mid-tier card from two generations ago 53 00:03:54,200 --> 00:03:59,160 to the 5090. Meanwhile, Returnal is about like Cyberpunk, 54 00:03:59,160 --> 00:04:02,200 approximately tripling its performance. 55 00:04:02,200 --> 00:04:06,480 Let's see what happens at 4K though. And there we go. 56 00:04:06,480 --> 00:04:10,280 Even though we've forexed the pixel count, compared to just now, we're running it 57 00:04:10,280 --> 00:04:16,040 just over half the FPS, which highlights just how CPU bottlenecked we were 58 00:04:16,040 --> 00:04:21,600 when we were running at 1080p. As for how this compares to our poor 3060 at 4K, 59 00:04:21,600 --> 00:04:26,600 well, let's just say we go from slide show to what a show. 60 00:04:27,100 --> 00:04:32,000 I mean, this is just, this is just smooth. This is a pretty darn solid gaming experience, 61 00:04:32,000 --> 00:04:35,520 but just because we're having a good gaming experience now, 62 00:04:35,520 --> 00:04:39,960 doesn't mean that Uncle Milburn's money was optimally spent. 63 00:04:39,960 --> 00:04:43,320 To answer that, we've thrown our 5090 into this tower 64 00:04:43,320 --> 00:04:46,360 with a Ryzen 7 9800X3D. 65 00:04:46,360 --> 00:04:49,880 That's right, we're not only going from six to eight cores, 66 00:04:49,880 --> 00:04:52,920 but these are running at almost double the base frequency, 67 00:04:52,920 --> 00:04:57,680 and we've also doubled our RAM to 32 gigs of DDR5. 68 00:04:57,680 --> 00:05:00,840 Starting at 1080p, it's clear that our slower CPU 69 00:05:00,840 --> 00:05:04,320 resulted in us leaving a lot of performance on the table. 70 00:05:04,320 --> 00:05:09,320 We've gone from around 100 FPS to over 150 FPS. 71 00:05:10,920 --> 00:05:14,800 And this is a butter smooth gaming experience. 72 00:05:14,800 --> 00:05:19,800 And that's true across the board, meaning that much of our 5090's performance 73 00:05:19,800 --> 00:05:23,360 was completely wasted on our average PC. 74 00:05:23,360 --> 00:05:28,600 But what about at 4K? Here, I've got to say, my surprise kind of goes 75 00:05:28,600 --> 00:05:34,080 the other way. Obviously, I know that Cyberpunk is a very GPU heavy game, 76 00:05:34,080 --> 00:05:38,660 especially at higher resolutions. That's why we use it as a GPU benchmark. 77 00:05:38,660 --> 00:05:42,300 But performance that's exactly the same 78 00:05:42,300 --> 00:05:45,860 as what we were getting with the 12400? That I didn't see coming. 79 00:05:45,860 --> 00:05:51,100 You have to be seriously GPU bottlenecked to be pushing four times the number of pixels, 80 00:05:51,100 --> 00:05:54,420 but still be getting over half the FPS. 81 00:05:54,420 --> 00:06:00,900 We saw the same thing in Returnal, meaning that only Anno and F1 managed sizable FPS increases 82 00:06:00,900 --> 00:06:06,020 going to our higher end CPU, showing that these games, even at 4K at these settings, 83 00:06:06,020 --> 00:06:09,640 are still somewhat CPU bound. But here's the thing, 84 00:06:09,640 --> 00:06:16,760 we have an average PC or not gaming at 1080p. So then if buying a 5090 is a total waste of money for us, 85 00:06:16,760 --> 00:06:19,880 what is a reasonable amount of Milburn's inheritance 86 00:06:19,880 --> 00:06:23,720 to spend on a GPU for our average PC? 87 00:06:23,720 --> 00:06:26,800 How about a 9070XT? 88 00:06:26,800 --> 00:06:30,880 It's got better availability, especially if you're willing to physically go to a shop 89 00:06:30,880 --> 00:06:34,440 and they're selling for maybe 850 Yankee bucks, 90 00:06:34,440 --> 00:06:40,200 which is still too expensive if you ask me, but AMD and NVIDIA didn't ask me, did they? 91 00:06:40,200 --> 00:06:45,320 Let's see how it holds up. Would you look at that at 1080p? 92 00:06:45,320 --> 00:06:49,760 Our average PC, now powered by an AMD GPU, 93 00:06:49,760 --> 00:06:56,360 not only matches our 5090, but actually beats it in three out of our four games. 94 00:06:56,360 --> 00:07:01,560 Not by much, other than F1, this probably just comes down to run-to-run variants, 95 00:07:01,560 --> 00:07:07,600 but still, what that shows is that we are clearly CPU bound 96 00:07:07,600 --> 00:07:12,600 and buying a more expensive GPU is giving us nada, nothing. 97 00:07:12,760 --> 00:07:17,760 We could just put a 9070XT in here, maybe even less. 98 00:07:17,760 --> 00:07:21,880 But what about 4K? Here, now that we're more GPU bound, 99 00:07:21,880 --> 00:07:24,960 the 5090 does manage to differentiate itself, 100 00:07:24,960 --> 00:07:29,000 but it does show that if we're not trying to get 101 00:07:29,000 --> 00:07:35,420 the best of the best performance at 4K, that 12400 can still kind of keep up. 102 00:07:35,420 --> 00:07:39,340 All of our games get dragged into actually playable territory here, 103 00:07:39,340 --> 00:07:43,780 though Cyberpunk is arguably a little bit marginal, 104 00:07:43,780 --> 00:07:47,300 at least at these ray traced medium settings. 105 00:07:47,300 --> 00:07:51,060 So did AMD nail the positioning of the 9070 series? 106 00:07:51,060 --> 00:07:55,780 It kind of looks like it. 1080p performance is neck and neck with higher end cards, 107 00:07:55,780 --> 00:08:02,300 as you would expect with the modern CPU, and at 4K, we're getting at least 50% of the performance 108 00:08:02,300 --> 00:08:06,420 of the 5090 for less than 30% of the price. 109 00:08:06,420 --> 00:08:11,980 Now, it is still true that the 9070XT is gonna get a little bit more room to shine 110 00:08:11,980 --> 00:08:15,700 when paired with a higher end CPU like a 9800X3D, 111 00:08:15,700 --> 00:08:21,860 and we found that was especially true at 1080p, but the mismatch is not nearly as bad. 112 00:08:21,860 --> 00:08:27,460 So I guess what I'm trying to say is, for most gamers, that unobtainium video card 113 00:08:27,460 --> 00:08:31,380 that you can't get your hands on, might not have affected your gaming performance 114 00:08:31,380 --> 00:08:35,260 nearly as much as you thought anyway. Fun thing, by the way, 115 00:08:35,260 --> 00:08:39,980 Ploof mentioned in a previous video that he talked his friend into upgrading his monitor 116 00:08:39,980 --> 00:08:43,360 instead of shelling out for an expensive 50 series GPU. 117 00:08:43,360 --> 00:08:49,260 Well, I have an update for you. His friend went with it and is totally stoked on the decision. 118 00:08:49,260 --> 00:08:52,420 So while the wisdom has for years been 119 00:08:52,420 --> 00:08:55,420 just dump your budget into the GPU, 120 00:08:55,460 --> 00:09:00,540 I feel like GPU companies have kind of taken advantage of that wisdom and convinced us to spend more 121 00:09:00,540 --> 00:09:05,060 than we should be spending on the GPU. And these days, we might wanna consider 122 00:09:05,060 --> 00:09:11,140 a more rounded approach to our upgrades. Like this rounded segue to our sponsor. 123 00:09:11,140 --> 00:09:14,860 If you guys enjoyed this video, why not go back and find the one where we built 124 00:09:14,860 --> 00:09:18,260 the most average PC? It's Deez.