WEBVTT

00:00:00.080 --> 00:00:06.879
The way I see it, you clicked on this video for one of just three reasons. You

00:00:04.640 --> 00:00:11.200
either want to watch tech giants in a CPU dunking contest. You could

00:00:09.280 --> 00:00:15.440
legitimately be shopping for the fastest PC hardware on the planet.

00:00:13.040 --> 00:00:19.600
Congratulations, by the way. Or most likely, you just want to eat some

00:00:17.199 --> 00:00:24.080
popcorn and watch the fanboys squabble in the comments about an overpriced CPU.

00:00:21.920 --> 00:00:30.400
Well, guess what? Buckle in tight, cuz you're going to see all three today. The

00:00:26.400 --> 00:00:32.640
Ryzen 9950 X3D2 Dual Edition is an

00:00:30.400 --> 00:00:37.600
eyewateringly expensive, zero compromised product, and I love it. It's

00:00:35.600 --> 00:00:41.680
the perfect purebred show horse to excite geeks like us by pushing the

00:00:39.680 --> 00:00:47.120
limits of modern technology while also serving as a reminder that most people

00:00:44.320 --> 00:00:50.399
really don't need one. What everyone does need though is this segue to our

00:00:49.520 --> 00:00:55.680
sponsor, >> Acuity Scheduling. It's a flexible scheduling software that offers calendar

00:00:54.000 --> 00:00:59.840
management, booking tools, client communications, secure payments, custom

00:00:57.199 --> 00:01:04.680
branding, and much more. Save 20% today with the code LT20 at our link below.

00:01:11.200 --> 00:01:17.040
Hi. If you're wondering why this video is a day late, well, AMD didn't send us

00:01:14.720 --> 00:01:20.640
a chip to review, and they also didn't send a lot of people chips to review.

00:01:18.880 --> 00:01:24.960
Make of that information what you will. Anyway, here is our review. Getting

00:01:23.200 --> 00:01:30.000
right into the meat of things, I can understand why AMD struggled to name

00:01:27.360 --> 00:01:36.159
this chip. And they clearly did because what is this thing? Let's break it down.

00:01:32.799 --> 00:01:39.200
It's a 9950X, so that means you got 16

00:01:36.159 --> 00:01:42.400
of AMD's latest Zen 5 architecture cores

00:01:39.200 --> 00:01:44.799
and 32 threads. It's 3D, which means it

00:01:42.400 --> 00:01:48.560
features a large stacked 3D vcache, which can offer a major performance

00:01:46.320 --> 00:01:54.880
benefit in certain workloads, notably gaming. and it's two and also dual both

00:01:53.360 --> 00:02:00.799
of which seem to mean the same thing because compared to a regular 9950X3D

00:01:57.920 --> 00:02:06.560
non two non-dual the big change that the dual edition brings is that AMD's super

00:02:03.520 --> 00:02:08.879
large 3DV cache now sits on both of the

00:02:06.560 --> 00:02:13.440
core complex dies or CCDs that are under the gray heat spreader rather than only

00:02:11.039 --> 00:02:17.760
on one of them. That gives the duel a whopping

00:02:14.959 --> 00:02:23.280
192 MGB of level three cache. That's enough to install Windows 95 on your CPU

00:02:21.040 --> 00:02:27.120
three times over. Doesn't really work like that. We just thought it was kind

00:02:24.800 --> 00:02:32.080
of a funny thing to point out. But hold on a second. If 3DV cache is so great,

00:02:29.599 --> 00:02:36.800
then why didn't AMD use it across all of their cores in the past? Well, because

00:02:34.560 --> 00:02:41.599
there are some trade-offs. the price for one, but also a considerable 30 watt

00:02:39.840 --> 00:02:47.200
increase to total design power that brings the duel to a sweaty 200 watts

00:02:44.480 --> 00:02:51.680
that in some situations it even exceeded by a significant margin. And this extra

00:02:49.440 --> 00:02:57.200
power seems to have resulted in a small decrease in boost clock speeds.

00:02:55.519 --> 00:03:01.360
But hey, that should be more than made up for by the extra 3D cache, right?

00:02:59.360 --> 00:03:05.760
Especially in gaming, right? Well, let's talk about that. Out of the gate, things

00:03:03.200 --> 00:03:10.640
are looking good in City Skylines 2. a game we chose more for its ability to

00:03:07.680 --> 00:03:15.760
bring CPUs to their knees than its uh community appeal. Here, the Duel leads

00:03:13.040 --> 00:03:19.680
the pack with the other 9000X3D chips in tow behind it. Makes sense. After all,

00:03:18.480 --> 00:03:26.080
they're supposed to be good for gaming, right? And we see the same deal in Cyberpunk 2077. Though, it is worth

00:03:23.680 --> 00:03:29.599
noting here that while these high average FPS numbers here are very

00:03:28.080 --> 00:03:34.000
impressive, when it comes to the all-important 1% lows, which represent

00:03:32.400 --> 00:03:40.319
the performance you can expect in the most demanding moments, some older and

00:03:36.879 --> 00:03:42.799
uh much less expensive chips are nipping

00:03:40.319 --> 00:03:46.879
dangerously close to their heels. In F124, it seems like any high-end

00:03:45.519 --> 00:03:51.200
processor from the past couple generations is going to do you just

00:03:48.640 --> 00:03:55.280
fine. Shout out to the venerable 5800 X3D by the way, which is rumored to be

00:03:53.120 --> 00:04:00.080
returning for a celebration of the AM4 socket's 10th anniversary. Then in The

00:03:57.360 --> 00:04:03.599
Last of Us Part One, the 9950X3D2 duel returns to the top position and is also

00:04:02.480 --> 00:04:08.000
about as fast as it gets in Counter-Strike 2, at least as far as 1%

00:04:06.000 --> 00:04:14.080
lows are concerned. Across our suite of gaming benchmarks, the Ryzen 9 9950XD2

00:04:11.120 --> 00:04:20.320
dual edition is just plain ripping fast, and yet still nearly impossible for me

00:04:17.040 --> 00:04:23.520
to recommend. Why? Well, for starters,

00:04:20.320 --> 00:04:25.919
we benchmark gaming CPUs at 1080p. This

00:04:23.520 --> 00:04:29.919
gives us a clear view of which chips will separate themselves from the pack

00:04:27.759 --> 00:04:35.360
in an intentionally CPU bottleneck situation. But if you're buying a $900

00:04:33.360 --> 00:04:41.280
chip, you're almost guaranteed to be running at at least 1440p, if not 4K,

00:04:38.720 --> 00:04:45.600
where a GPU bottleneck is way more likely than a CPU one. The second reason

00:04:43.600 --> 00:04:49.680
is that even if you were an esports professional trying to eek out every

00:04:47.280 --> 00:04:53.840
last frame, I still wouldn't sell you a duel. Because while it does have more

00:04:52.080 --> 00:04:58.639
3DV cache goodness than anything we've ever seen before, which is cool, in the

00:04:56.320 --> 00:05:04.880
real world, it performs basically the same as any of the other 9000 series X3D

00:05:01.520 --> 00:05:07.360
chips. But why is that? Well, as some of

00:05:04.880 --> 00:05:11.280
you might know, high-end X3D chips have always come with kind of a weird

00:05:09.199 --> 00:05:16.479
trade-off. They had all these cores on them, but only some of the cores were

00:05:13.600 --> 00:05:20.560
best for gaming, the 3D vcache ones. So, you always wanted to make sure that your

00:05:18.240 --> 00:05:25.199
games were running on those. To do this, AMD uses a technique called core

00:05:22.639 --> 00:05:29.919
parking, which unintuitively does not stop the core from doing any work, but

00:05:27.039 --> 00:05:34.720
rather it forces a given program to only use specific CPU cores and then ignore

00:05:32.639 --> 00:05:38.320
the rest. Over the past couple of years, AMD and Microsoft have worked on

00:05:36.479 --> 00:05:41.600
dejenifying this behavior to the point where as long as you're running an

00:05:39.759 --> 00:05:47.039
up-to-date chipset, BIOS, and operating system, you should find that your 2CD

00:05:44.320 --> 00:05:51.360
CPU is using the right cores for any given job. That is, assuming, of course,

00:05:48.960 --> 00:05:55.759
that Windows recognizes your games as a game. But we have 3D Vcash on all the

00:05:54.160 --> 00:06:00.880
cores now. So, why are we talking about parking anything anywhere? Vroom v room,

00:05:57.520 --> 00:06:03.120
right? Well, as it turns out, most games

00:06:00.880 --> 00:06:07.360
just plain don't need more than eight cores. And regardless of what type of

00:06:05.280 --> 00:06:12.560
cache your cores have if they have to coordinate work across multiple CCDs,

00:06:10.479 --> 00:06:17.520
the data has to travel across the Infinity Fabric and the IO die, which is

00:06:15.199 --> 00:06:23.360
going to introduce extra latency. So, the solution, the same jank we were

00:06:20.479 --> 00:06:28.000
already using, core parking. And that is why the new dual edition performs

00:06:25.360 --> 00:06:33.360
basically exactly the same in gaming. So then who is it useful for? Well, for

00:06:30.639 --> 00:06:36.800
starters, anyone who might run multiple workloads that benefit from extra cache.

00:06:35.360 --> 00:06:41.360
A great example would be someone who wants to stream or record their gameplay

00:06:39.600 --> 00:06:46.400
while they're gaming without burdening their GPU. A more niche benefit is that

00:06:43.759 --> 00:06:50.639
overclockers and system tuners could identify which of their CCDs runs the

00:06:48.800 --> 00:06:54.800
fastest and then put their most performance sensitive workloads on those

00:06:52.560 --> 00:06:59.039
cores. Or at least that would be an option if AMD and Windows allowed you to

00:06:57.039 --> 00:07:03.840
choose. From our testing, it seems that games are locked to CCD0 for the time

00:07:01.599 --> 00:07:08.800
being. Another focus for AMD's marketing for this chip is a claimed 5 to 10%

00:07:06.720 --> 00:07:14.960
increase in productivity. Let's take a look at that. And unsurprisingly, the

00:07:11.599 --> 00:07:17.280
Ryzen 9 9950 X3D2 dual edition Core 2

00:07:14.960 --> 00:07:21.360
dual extreme max plus the sequel is an absolute monster here as well. In 7zip,

00:07:19.360 --> 00:07:26.000
the duel has a clear lead over any other consumer CPU both in compression and in

00:07:23.840 --> 00:07:29.919
decompression. And in Cinebench, it sits at the top of the chart or darn near it

00:07:27.919 --> 00:07:35.840
in both multi-core and single core performance. The only problem is that

00:07:31.919 --> 00:07:39.520
its closest competitor is uh

00:07:35.840 --> 00:07:41.680
five $550 cheaper at retail right now.

00:07:39.520 --> 00:07:45.280
Uh but hey, those are kind of just benchmarks, right? And if you're in the

00:07:43.199 --> 00:07:48.639
market for a CPU this expensive, you probably want to get some real work done

00:07:46.720 --> 00:07:53.680
with it, right? Where in Blender, the dual edition does manage an as

00:07:50.960 --> 00:08:00.720
advertised 5% improvement over its less endowed twin and a considerable 15ish%

00:07:57.360 --> 00:08:02.879
over the 7950 X3D. This is one of the

00:08:00.720 --> 00:08:07.680
first cases we've seen where the Duel actually kind of stands out for its

00:08:04.879 --> 00:08:12.879
value. To be clear, it only beats the much more affordable 270K Plus by 10

00:08:10.639 --> 00:08:17.360
seconds. But if you use your system to make money, every second counts, and the

00:08:15.440 --> 00:08:21.840
Duel is as fast as it gets without paying a considerable premium for a

00:08:19.599 --> 00:08:26.639
workstation platform motherboard. And it even continues AMD's bro move of

00:08:24.479 --> 00:08:30.800
supporting ECCU dims on select motherboards, making it a great semi-pro

00:08:28.960 --> 00:08:35.919
offering. And that carries over to encoding as well, where the X3D2 is

00:08:33.279 --> 00:08:40.959
again at the top of the charts in H.264 and close in AB1. In Puget Bench's

00:08:38.719 --> 00:08:47.279
Photoshop benchmark, the Duel is the first consumer x86 CPU we've tested that

00:08:43.440 --> 00:08:49.120
cracks 14,000 points, which is cool, but

00:08:47.279 --> 00:08:53.920
only a negligible increase over the rest of the Zen 5 contenders. and NVIDIA

00:08:51.279 --> 00:08:57.760
editing. She's a chart topper again and more importantly can offer an uplift

00:08:55.760 --> 00:09:03.040
that you might actually feel compared to an older generation 3D cache chip like

00:09:00.240 --> 00:09:07.120
the 7950 X3D. Now, you wouldn't think that based on the overall score from

00:09:05.040 --> 00:09:10.320
Puget Bench, but we found that depending on the files you're working with, you

00:09:08.720 --> 00:09:15.600
might find performance enhancements that are much greater than the two-ish% on

00:09:12.640 --> 00:09:19.519
our chart. However, as with gaming, the situations where we see large

00:09:16.880 --> 00:09:23.440
improvements over the non-duel are few and far between. As for Premier Pro,

00:09:22.080 --> 00:09:28.480
it's kind of like a grade schooler going to the gym. The gains are there, but

00:09:25.519 --> 00:09:32.640
they're minor. So, yeah, in our creator focused testing, the duel doesn't offer

00:09:30.320 --> 00:09:36.959
much, but to be fair, there are lots of kinds of productivity, and the extra

00:09:34.320 --> 00:09:41.279
level three cache can be huge in mathy and sciency workloads. Check out this

00:09:39.440 --> 00:09:45.760
result from Feronx's review of the original 9950 X3D. Since we're a day

00:09:44.160 --> 00:09:49.360
late, we have the rare second mover advantage. And we can actually just show

00:09:47.200 --> 00:09:52.480
you some of Feronx's results. And you can see that there are some workloads

00:09:50.560 --> 00:09:58.560
where the chip benefits a lot from the extra vc. You might be thinking that's

00:09:54.560 --> 00:10:00.640
not very many workloads. And uh yeah.

00:09:58.560 --> 00:10:03.760
Yeah. Anyways, if you do want to learn more, we highly recommend going over to

00:10:01.920 --> 00:10:07.360
Feronics, looking at their Linux and scientific benchmarking. It's pretty

00:10:06.000 --> 00:10:12.399
cool. There's going to be a link in the description. >> But I will say there's a solid chance

00:10:09.680 --> 00:10:18.240
that if you are degridding radio signals from space, this could be your new

00:10:14.959 --> 00:10:20.560
powerhouse. Speaking of power, at 200 W,

00:10:18.240 --> 00:10:25.600
this is the highest TDP that AMD has ever set on a consumer chip. And does it

00:10:23.760 --> 00:10:31.200
ever deliver on its promise. In Cinebench, this thing pulls up to 260

00:10:28.640 --> 00:10:36.959
watts across the total CPU package. That is like Intel 14900K levels of power

00:10:34.399 --> 00:10:42.959
draw. But thankfully, it doesn't have the same power density and IHS problems

00:10:40.079 --> 00:10:47.360
as the 14900K. With that said, if you were thinking of upgrading from a Ryzen

00:10:44.880 --> 00:10:51.120
5950X to this new CPU, you should make sure that your power supply can handle

00:10:48.880 --> 00:10:56.079
the extra power draw. And uh oh yeah, probably also your CPU cooler. That's a

00:10:53.760 --> 00:11:01.600
tech tip. The good news in all this is that the extra power draw only rears its

00:10:58.079 --> 00:11:04.000
head in heavy all core CPU workloads. In

00:11:01.600 --> 00:11:07.519
F-124, for example, we found this chip drew about the same power as its

00:11:05.920 --> 00:11:11.680
siblings, meaning that you can enjoy manageable temperatures while just

00:11:09.279 --> 00:11:17.440
pretending to race in the uh Las Vegas heat. In conclusion, the Ryzen 9

00:11:14.079 --> 00:11:20.160
9950X3D2 dual edition puts a punctuation

00:11:17.440 --> 00:11:26.240
point on AMD's last 3 years of CPU domination. But while I love the new

00:11:23.680 --> 00:11:31.120
fastest one as much as anyone does, it's hard to recommend a chip that demands

00:11:28.320 --> 00:11:35.519
such a price for such small gains. And it feels a little bit like this is an

00:11:33.360 --> 00:11:38.959
unnecessary victory lap. And I would rather AMD focus their development

00:11:37.279 --> 00:11:45.760
efforts on things their community really wants. Like, oh, I don't know, maybe

00:11:41.920 --> 00:11:48.160
bringing FSR4 to RDNA3 GPUs. Or, hey,

00:11:45.760 --> 00:11:54.720
how about finally bringing Zen 5 to budget buyers? Where is Ryzen 3 AMD?

00:11:52.000 --> 00:11:58.240
Also, where is my sponsor? >> Acuity scheduling. The busier your life

00:11:56.640 --> 00:12:01.440
gets, the harder it can be to be at the right place at any given time, let alone

00:11:59.839 --> 00:12:04.880
prioritizing the right tasks to be doing. But Acuity Scheduling can help

00:12:03.279 --> 00:12:08.079
you keep your work in order thanks to its flexible calendar management and

00:12:06.560 --> 00:12:12.240
booking. Clients, colleagues, and acquaintances can schedule, cancel, and

00:12:09.920 --> 00:12:15.440
rebook time with you on their own, so all you have to do is show up as your

00:12:13.920 --> 00:12:18.959
best self. And even if you need to juggle several calendars, Acuity

00:12:17.040 --> 00:12:22.000
Scheduling can sync with multiple apps, making sure you don't accidentally

00:12:20.399 --> 00:12:25.839
overbook. It will automate things like confirmation emails, text reminders,

00:12:23.519 --> 00:12:29.040
follow-ups, and intake forms. Plus, it can prepare you with all of the relevant

00:12:27.360 --> 00:12:32.240
client history so you can stay informed. Acuity Scheduling also has gift

00:12:30.800 --> 00:12:37.120
certificates and subscription tools built right into the platform, which means less planning and more repeat

00:12:35.360 --> 00:12:40.079
business for you. They'll even help your entire team get on the same page with

00:12:38.639 --> 00:12:43.519
custom staff permissions, as well as push notifications, so everyone knows if

00:12:42.160 --> 00:12:48.079
things are moved or canceled. When it comes time to invoice clients, Acuity Scheduling works with the most popular

00:12:46.560 --> 00:12:52.800
payment providers, so you can get paid in person or set up recurring billing.

00:12:50.240 --> 00:12:57.360
So, go to acuityccheduling.com/LTT for a free trial and use offer code

00:12:54.320 --> 00:12:58.800
ltt20 to save 20% off your first Acuity

00:12:57.360 --> 00:13:05.200
Scheduling subscription. If you guys enjoyed this video, why not check out the one we did recently on how X3D chips

00:13:02.800 --> 00:13:08.160
in particular don't really need the fastest memory for gaming, which could

00:13:06.959 --> 00:13:10.959
help save you some money if you're putting together a new