ladies and gentlemen, my name is Paul. Hope you're having an amazing day. I want to talk to you guys about some interesting benchmarks which have emerged for AMD's Granite Ridge range of processors, specifically an engineering sample Ryzen 9 5950X, which an AnandTech Forum member by the name of Igor Kavanas, hopefully that's pronounced correctly, has managed to snag. Now, it is worth noting that this is an engineering sample chip, so of course, there possibly is going to be some differences in what the final silicon is capable of. i.e. is probably going to be clocking faster. Plus, also the BIOS revisions at this point are exactly going to be indicative of final performance either, so you can probably expect that the retail chips, specifically, you know, a month or two after release, especially probably still have a little bit of gas left in them. With that said, Igor has actually been testing these chips across multiple different PPTs: 230 watts, 160, 120, 90, and also 60. And he's been testing them in

a variety of different benchmarks and scenarios, but we're going to focus on Blender because, my goodness, these chips are what I would call rapid. In a nutshell, we are looking at a chip which is almost 30% faster than the 7950X, and as for Intel's 14900K, well, it's spanking it. It's actually 34% faster. I want to give courtesy credit here to WCCFTECU. They've done a pretty nice job of putting all of these things together in a nice set of graphs, and these are basically samples per minute. There are three different benchmarks which have been run here: monster, junk shop, and classroom, and you could say that AMD are putting on a classroom here for performance, especially at lower power levels because, well, this basically is going to be a very flexible chip. Now, obviously, if you're running an AIO cooler, which is what AMD recommends, especially if you're going to be cranking the chip so high, then you certainly can do so. But if

you decide, Hey, you know what? I want to build like a nice small form factor system. I don't necessarily need to squeeze the juice completely out of this orange. I'm happy to go with like a, you know, kind of lower level like 120 or what have you. Then you've also got that flexibility. Anyway, enough talking about random stuff. Let's get on to the benchmarks. Focusing on Monster for a second, we're looking at, I'm going to round up and down the numbers here for everyone's sanity. 350 if you're running at 230 watts. Meanwhile, cranking things down to 320. So, obviously, you are losing something here. But if we compare that versus the 7950 X. So, 230 versus 230. 353, I'll be exact here just for clarity. And 295 for the C 950X. Again, 230 watts per piece. I'm also going to mention a result that HXL on Twitter has managed to discover. And this is on The Geekbench Benchmark. It is worth noting, of course, that

geek bench is going to differ in performance levels across different um revisions. Obviously, you've got a geek bench, uh, 6.3, 6.2, and so on, and different chips over so will respond a little bit differently on different benchmarks. But nevertheless, this one's quite interesting because the 9900x here is running at 5.6 GHz, assuming it's being detected correctly, but the DDR5 memory is running at a blistering 8,000 MTS. It is going to be super intriguing to see how well um these chips actually scale in terms of performance. I mean versus the uh speed of the memory. There are a lot of uh theories as well as Reports online that they are very IOD dependent and also memory bandwidth dependent, but it's also going to be interesting to see how latencies also affect things as well. I've heard that memory speed in particular does improve performance, but at the end of the day a is that true B is it because it's an engineering sample

and C again the software as in like the BIOS versions and so on aren't exactly final at the stage. Nevertheless, we are getting ever so close to the single-core score. Uh, I always feel like I'm he ye he ye ringing a bell or something like that, but we're getting very close to 3500 points here for the single-core, and the multicore is 22,000. And I'll quickly finish the video. Um, well, at least the AMD part of the video because I also want to speak just briefly about Intel's Arrow Lake release dates. But um, speaking about CPU-Z, um, CPU-Z is another application which tends to do quite well with Intel again. Obviously different benchmarks just sound like a broken record for a second. Different benchmarks and going to push chips in different ways. Quite frankly, the Zen results are all over the place. In some results, you're seeing Zen 5 do very well. In other ones, not so well. Obviously, AVX 512 is a thing. They've also made some big changes to the integer performance.

as well, the front end for Zen 5 has also been quite radically changed. There's like a dual split front end. Basically, the whole thing is just really. It's going to be very interesting to see how performance actually improves over time, different BIOS revisions, and so on. Either way, just to get back to my point here, the 795X in terms of multiread is probably going to score roughly just under 16,000 points, and something like the 13900K is going to be about 177,000 points. And obviously, the 14900K is going to be a little higher, but HXL has managed to find a 9900X, which is actually showing very impressive results as you can see on the screenshot here for both single and multicore results. And now I just want to talk about the Alder Lake release date. I'm going to keep this super brief, guys. It's going to be like a very quick thing because there has been

some reporting online that Intel are not going to be launching uh Arrow Lake. And just for clarification here, I mean the highest-end versions, for example, 8+ 16 uh until some point in December. But this does not seem to be the case. Uh, basically, it is not going to be in December. Um, jkin on Twitter has also been stating that no, it is definitely going to be October. All of the people I've spoken to have also said, um, it's going to be October unless something very recent has changed. It is still going to be October. Again, I would say that at this point, I would probably suggest just waiting for, uh, all of the variants of the different processes to launch. So, by which I mean Zen Zen x3d as well as whatever you want from the Arrow Lake lineup so you've got a better understanding of what's coming, um, in terms of, you know, the pricing, the motherboard situation, blah, blah, blah, blah, blah. But any anyway, yeah, I just want to kind of give you that update because there have been a lot of

websites were reporting December not for no reason. There were some leaks that saying it was December, but I just want to give you this update that no, it is definitely October, unless Intel decid to delay things again. With that said, take care of yourselves, have an amazing day. Apologies for not being on camera for this one, but I just didn't really think there was much of a point because essentially most of the video is me just showing stuff on screen, so you know I'm not really kind of on the camera anyway. With that said, take care of yourselves, have an amazing day. Bye for now. [Music]

Ryzen 9950X Almost 30% FASTER Than 7950X?! Zen 5 Results Leak

Scripts