Hi, this is Wayne again with a topic “AMD Ryzen 5800X3D: Let’s Chat!”.
Well, that’s fancy. This is the 5800x 3d review, amd’s post alder lake cpu launch sort of kind of zen 3 with a little bit extra oomph sort of thing, but there’s weird stuff. We got to talk about it. I don’t know let’s dive in so this is an eight core cpu, it’s the 5800 x 3d, so the regular 5800x has 32 megabytes of l3 cache. This cpu has an additional 64 megabytes of l3 cache, bringing the total up to 96..
That’S pretty much it oh and it’s locked down, not overclockable, we’ll talk about why kind of makes sense. That’S the big difference from 32 to 96 megabytes of l3 cache. It’S just drop in one of the test systems that i have set up for. This is based around the tomahawk b450 from msi and i’ve added an msi supreme 3090.
It’S not very likely that you would be running a 3090 with a 4050.. I mean that build is basically all gpu, but i think it really underscores the longevity of am4. You see. I set this system up with an 1800x.
It’S not unreasonable to think that when b450 motherboards launched you got a deal on an 1800x 8 core processor. You got this motherboard and you’ve been rocking that pretty much ever since. Maybe you got a deal on a 30 90 or other high-end. Gpu 39 is really a stand-in for higher uh higher-end gpu, and you want to see what kind of a you know system that you want to do for an upgrade. I think this demonstration is going to really underscore how awesome amd’s platform is, because you can just literally pick up a cpu update your bios drop that in and be good to go, but before we get to that, let’s talk a little bit about performance. This thing is amazing, but i have to temper your expectations just a little bit when we’re talking about 10th and 11th generation intel cpus. You know the 10 core for the highest end i9 and the 8 core for the highest end i9 and also the i7s. It’S really down to instructions per clock and how fast things can move in and out of the cpu.
Well, what makes this cpu special is that it’s able to cache more of what’s going on from main memory. The problem with that is that, strictly speaking, it doesn’t really translate into more of a generic ipc uplift more of a generic performance uplift when intel moved to alder lake, their new architecture, where they have p cores and e cores. They designed the processor so that the p chords would be absolutely monstrous in their computation, but also their power utilization. It’S not really possible for them to easily cram 16 of those p cores on a single cpu, so they make up for it with weaker, slower e cores and philosophically that’s a sort of similar to what we saw with chiplets.
We got an eight core chiplet and an eight core chiplet and maybe on the 5950x, there’s a weaker eight core chiplet and a stronger, a core chiplet, but they’re physically the same design. It’S just with intel’s. It’S a monolithic piece of silicon. It’S not chiplets and there’s eight cores that are designed to be fast and eight cores that aren’t and make no mistake: alder lake, a serious competition for amd. Finally, whereas in prior generations i wan na tell might have been struggling just a little bit so enter this. For gaming, the 5800x3d, so the 5800x really isn’t much competition for alder lake, except maybe in pricing at the low end, some exceptions, but not at the highest end. But here we have this cpu, which absolutely gives the 12 900 ks a run for its money in gaming. So if we take a look at our 5800x and cpu-z side-by-side that with our 1800x, those are some significant gains since n1.
But if we side by side by side that, with the 12 900 k in cpu-z’s built-in benchmark well, the 12900k is still pulling far ahead. And yet, when we run benchmarks with shadow of the tomb raider, this plucky little cpu can pull far ahead. Of the 12 900k, even the 12 900 ks – and this holds pretty much true for most of the aaa titles – there’s a few that don’t we’ll come back to those, but generally the performance is pretty improved. This is just a sample of the madness that we’re looking at here with the 5800x3d.
So this machine is a 12 900 kf we’re running with a 30 90 ti. We can only manage 241 fps, i mean 241 fps, that’s pretty awesome! It’S a high preset. We don’t have ray tracing with there’s nothing strange going on with that.
But if we take a look at the 5800x3d, we’re rocking just a regular 3090 and it’s 250 fps with a 30 90 250 fps with a 30 90.. What and when we finally do put the 3090 ti in the 5800x 3d system, it’s over 270 fps. For me, i think the the more interesting story is the upgrade path.
If, if you’re on a first or second generation, oh you got a 2700x drop. This in chances are it’s just gon na work, amd and board partners are even working on x370 support. There’S a little stop and start there with do. We want to support that.
Do we not want to support that? Maybe the competition has changed. It’S like all right, we’re going to give gamers an out we’re going to let them upgrade cpus if they really want to. For select x370 motherboards that really support it to be sure older motherboards support for things like pci express 4, not so much so when you look at the benchmarks with the 30 90 on this b450 platform with the 1500x3d bear in mind, that is an older platform And performance will suffer a tiny, tiny little bit because it’s an older platform and yet this plucky little b450 can pull ahead of a 9900ks. Yes, it can so, let’s talk about the games that don’t benefit from vcash, because that’s what makes it weird? It’S not universal! Csgo cs go is a great example cs. Go is a game, that’s not really doing much of anything with the cpu. In the first place it fits entirely in the cache memory of alder lake and even the 32 megabyte versions of you know ryzen 5000 cpus. So when you increase the cash from 32 to 64 megabytes, it doesn’t really benefit those games because they’re already super optimized, their memory access patterns are already super optimized. There’S not a lot of room for cash to improve that experience.
Those games are esports. Titles, dare i say that most esports titles don’t benefit tremendously from 3d v cache on the 5800x 3d. But a lot of your bigger games do a lot of your aaa titles do and that really bears out in the benchmarks and the reason why some games benefit and some don’t is really just to do with the memory access pattern and how much system memory those Games actually use the question of whether or not to upgrade if you’re, on a ryzen 5000 series already is a little murkier, i’m gon na say probably not unless you’re on maybe a 5600x. If you’re gon na go from a 5600x to this, and you can sell your 5600x for 100 200 – that’s a good! You should do that. That’S a good move! If you are on 3000 series, ryzen cpus and your stuff is going to work with the 5800x and you can drop it in with the cpu. Just like i did.
Yeah definitely do that if you’re on zen one or zen plus 1800x 2700x – something like that yeah. This is a no-brainer, especially if you game with your machine. This is really good. Even if you get an older gpu, this will still benefit you.
So how do you make sense of all the the benchmark data? This is our 5800x 3d system. This is what we’ve done all of our benchmarking on, and i’ve got the 30 90 ti in the system right now for games like shadow of the tomb raider, which are great benchmarks for showing what is possible. When you have this much l3 cache on your processor. We’Re talking about well over 270 fps. The problem with this is that you really need to know what your gpu utilization is. I was surprised you can use 100 of a 30 90 ti even at 1080p resolutions, so this is 1080p high. You know when i was farting around with the b450 1500x 3d upgrade. I was testing it highest, but this is high because otherwise it’s going to bottleneck on the gpu most of the time. Why that’s important is the next generation of gpus may open up so much possibility in terms of game performance that one cpu would perform differently than another cpu? It’S really sort of an interesting situation at higher resolutions like 1440p and 4k. The 5800x3d sort of starts to perform on par generally with the alder lake, 12900 ks or the 12900 kf that we use with with our testing.
On the one hand, we’ve got you know the newer micro architecture, which is generally faster for general computation, but on the other hand, we have the 1500x3d with the monster cache, which will benefit some computational workloads, but not necessarily all workloads outside of games. One of the few areas that translate into more performance is also to do with compression decompression and that kind of makes sense, because, when you’re compressing something you’re sort of generating a dictionary of things that are repetitive in a file. And you have to look that up. A lot of memory operations more cash helps with that similar with decompression, similar with other types of jobs that are really memory intensive operations, so overall, which processor is faster for your particular game, is going to depend on your particular game, far cry 6 again, if you’re Not gpu bound, it seems like the 5800x3d is pulling ahead when you start to be gpu bound. Well, it’s par or maybe the 12 900k will start to pull ahead. I was surprised that we were able to get some of those edge case.
Scenarios like the 30 90 on a 5800 x 3d is faster than a 30 90 ti on the 12 900 kf that we tested on, and that’s just you know, owing to the fact that tomb raider really likes having that additional cash. It’S a really interesting situation and i’m sure that you’re going to see benchmarks from all your favorite folks with numbers all over the place and that’s why overclocking? Why is there no overclocking well, the way that they added more cash to this thing is another really interesting: packaging technology, they’ve, basically thinned the die and taken 64 megabytes of l3 cache on another piece of silicon, put that on top and it sort of connects and Fuses because of copper, micro bumps and the 64 megabytes lives on top of the cpu die. The problem is that now you have two pieces of silicon one on top of the other, and if one of them gets super hot and the other one’s, maybe not getting as hot, it can be a little bit of a problem. And so when you have two pieces of silicon right together and you have 64 or 96 megabytes of stuff in flight in cash, when you’re doing overclocking, errors can start to creep in.
And there are error correction mechanisms under the hood. But when you’ve got that much stuff in cash, it’s really not an ideal situation when you’re overclocking and that can be a source of errors when you’re running programs. So it’s going to make things really sort of unstable and also the the cash chip itself. The extra 64 megs of cash wasn’t really designed for overclocking it was.
It was designed for the enterprise and uh. You know amd needs to get this product out the door. So a lot of mitigating factors. The thing that makes me feel better about the overclocking situation.
With the cpu, in that there is none is that, when i turn pbo on for most of the ryzen cpus, that i’ve had the ones that support overclocking ppo doesn’t make a huge difference. The last cpu that made a huge difference with ppo for me was the 2990x, the 32 core, and i still have a system, that’s running pbo, which requires monstrous cooling and it’s basically been running 24 7 since launch day it’s fantastic. I’Ve got a 5950x tried enabling pbo. I introduced some stability issues dialed that back i’ve got three 5900x systems that i’ve tried, pbo on all three of them, cpus from different points in in life.
Pbo works, okay on one of them and it does give me a little multi-core performance boost but doesn’t really help with single core. So i don’t think you’re really giving up a lot without pbo and overclocking. It is technically true that the extra l3 cache is just a tiny, tiny little bit slower than 32 megabytes of l3 cache. However, in practice, because it’s caching so much more memory, it’s a net positive now keep in mind.
The ada64 benchmarks are on our b450 platform, with older gcl ddr4 3200 memory, so ddr4 3200 without the best latencies. This thing is going to run like a champ even on a new platform. So if you have something really hideously ancient and you’re looking to build an affordable gaming machine like for like and you’re, not upgrading, this is still a more inexpensive platform. Not only does this use about 100 watts less at the wall, in a worst case scenario, configuration versus team blue again, it’s worst case scenario.
The platform itself is perhaps more affordable owing to more diversity in motherboards, more upgraders, upgrading motherboards, creating an opportunity for you to pick up a motherboard, that’s compatible on a secondary market, but you can also get a nice new motherboard. I mean you get a nice b550 motherboard for this no problem and the part that really just blows my mind is. I can’t believe that for gaming, how much faster this can be than a 12 900k now to be sure for cinebench and everything else. The 12 900k is going to outperform this.
It’S 16 cores the best deal in the alder lake family, i think, is probably the 12 400, the six p cores no e cores i5 and probably the i7 12 700 k. If you want to do overclocking, those are probably the best cpus on team blue. You would be hard-pressed to build a system that performs this well at this price point from either one of those two parts.
It’S really pushing it. It’S maybe a little more possible on the 12 400 with the 6p course, but it’s uh competition, it’s alive and well, i’m wondering this is level 1. This has been a quick look at the 1500 x 3d and gaming benchmarks. It’S certainly a powerful cpu for gaming and it’s the fastest cpu for gaming. It’S a pretty affordable price point, especially in consideration for how much 12700k costs how much 12900k aks cost.
So ks is well, it’s a little pricey. The thing that blows my mind is that as expensive as competing cpus are amd can get it done with eight cores eight cores and a little bit more cash and a lower overall cost. It’S pretty impressive, i’m one of those level one i’m signing out. You can find me in the level one forums, good job team amd .