WEBVTT 00:00:00.320 --> 00:00:05.600 Okay, so it's here. This might not look 00:00:03.600 --> 00:00:09.360 like much, but if you've ever gotten a review sample from Intel, then you'll 00:00:08.000 --> 00:00:18.720 know that there could be anything inside. Usually a CPU, but anyway, this, 00:00:14.000 --> 00:00:20.240 my friends, is one of our two Intel Xeon 00:00:18.720 --> 00:00:26.880 2699 V3 processors. Now, this was not easy to 00:00:24.000 --> 00:00:32.399 get because this is not your average run-of-the-mill consumer, you know, tops 00:00:29.439 --> 00:00:37.160 out at a,000 peasant dollars processor. No, this is a $4,500 00:00:37.440 --> 00:00:43.280 chip. It has 18 cores. It supports 00:00:41.040 --> 00:00:51.280 hyperthreading for a total of 36 threads. It has 45 megabytes of cache. 00:00:48.000 --> 00:00:54.160 It has more cash than my first computer 00:00:51.280 --> 00:01:00.559 had RAM. My first computer that I owned for myself had RAM. It has more cash 00:00:57.199 --> 00:01:03.520 than that. And my server board isn't 00:01:00.559 --> 00:01:09.600 here for it yet, but I figure what the hey, this is LGA 20113, right? Just like 00:01:06.960 --> 00:01:13.439 an X99 motherboard. The X99 motherboard listed as 00:01:11.360 --> 00:01:16.960 compatible. So, I figured why don't we take her for a spin even though we can't 00:01:15.200 --> 00:01:21.759 fire up both of them. and I'm gonna put it in ASUS X99 Deluxe and find out what 00:01:19.920 --> 00:01:32.400 this little puppy is capable of. So, come along for the ride. 00:01:32.400 --> 00:01:40.000 The Ring Video Doorbell lets you see who's at the door. And really, now 00:01:38.240 --> 00:01:43.439 you're going to open the door. Okay, well, whatever. I could have prevented 00:01:41.759 --> 00:01:46.960 him from doing that by locking my door and then going and seeing the video on 00:01:45.439 --> 00:01:54.479 my phone to see if I should let that guy in. All right, so the first step is to 00:01:49.680 --> 00:01:58.719 take that uh Wussy Core i75960X 00:01:54.479 --> 00:02:00.560 8 core processor out of my test bench. 00:01:58.719 --> 00:02:04.880 Now, the first thing that stands out to me immediately about these suckers is 00:02:03.360 --> 00:02:08.399 the fact that even though they go in the same socket, you can actually see that 00:02:06.960 --> 00:02:12.560 the Xeon chip Look, my hands are shaking. I'm super stoked on this. You 00:02:10.560 --> 00:02:19.200 can see that the Xeon chip is significantly wider and it will actually 00:02:16.319 --> 00:02:23.520 overhang the socket a little bit compared to what the Core i7 chip did. 00:02:21.920 --> 00:02:31.280 Now, this is for a very good reason because each of those physical cores uh 00:02:28.319 --> 00:02:37.280 in the die down there is the same size as the eight that go in here, but 00:02:33.280 --> 00:02:39.040 there's 18 of them. 18 cores. So, 00:02:37.280 --> 00:02:44.000 they're going to be spread out quite a bit more under the integrated heat 00:02:41.200 --> 00:02:48.879 spreader, the metal piece on top. Now, one of the cool things about these Xeons 00:02:46.480 --> 00:02:53.040 is that, you know, you might go 18 cores. Holy cow. Oh, we better put some 00:02:50.720 --> 00:02:58.239 new thermal compound on there. 18 cores, holy cow. How, you know, are they going 00:02:55.040 --> 00:03:00.879 to cool that sucker? Well, because they 00:02:58.239 --> 00:03:06.879 have so many lower powered cores running at a lower voltage and because the power 00:03:04.640 --> 00:03:12.159 density is so much lower, they're actually significantly easier to cool 00:03:10.080 --> 00:03:18.959 and they will it will in theory run cooler than our fewer high-powered 00:03:14.800 --> 00:03:23.040 cores. So, uh I am really stoked to see 00:03:18.959 --> 00:03:25.360 how our single 120 mm water cooler fares 00:03:23.040 --> 00:03:29.599 with this bad boy. Okay, now zeons aren't really meant to be used with 00:03:27.440 --> 00:03:33.120 gaming memory. So, we actually had Kingston for the the dual processor 00:03:31.840 --> 00:03:39.599 server that we're building with this thing, right? We had Kingston provide 00:03:36.239 --> 00:03:45.080 128 gigs. So, that's eight 8x6 gig 00:03:39.599 --> 00:03:45.080 sticks of ECC DDR42133. 00:03:45.120 --> 00:03:55.360 So, we're going to go ahead and we're going to pull our gaming memory out, but I'm not 100% sure if this board even 00:03:50.239 --> 00:03:59.920 supports ECC. So, 00:03:55.360 --> 00:04:03.560 okay. Moment of truth time. 00:03:59.920 --> 00:04:03.560 Let's see if she boots. 00:04:05.200 --> 00:04:11.280 Oh, does this powered on? Oh, boss. 00:04:08.799 --> 00:04:18.959 Oh, here we go. That took a really long time. There it 00:04:14.480 --> 00:04:25.199 is. New CPU installed. 64 gigs of RAM. 00:04:18.959 --> 00:04:27.040 E5 2699 V3 at 2.3 GHz. So, uh, yeah, I 00:04:25.199 --> 00:04:32.479 guess all that's left now is boot into Windows. Check that out. CPU core 00:04:29.120 --> 00:04:35.960 voltage 784 volts. 00:04:32.479 --> 00:04:35.960 Next to nothing. 00:04:36.160 --> 00:04:42.639 You're going to miss the moment. It's installing driver software. Oh, that's 00:04:40.160 --> 00:04:50.400 just USB devices. Okay, I guess the CPU works already fine then. 00:04:44.960 --> 00:04:53.840 I don't know. Maybe those. Holy crap. 00:04:50.400 --> 00:04:56.880 Have you ever seen 18 threads and or 00:04:53.840 --> 00:05:00.800 excuse me 18 cores and 18 hyperthreaded 00:04:56.880 --> 00:05:04.240 virtual cores in one place at one time? 00:05:00.800 --> 00:05:08.479 That is fantastic. There's our 45 megs 00:05:04.240 --> 00:05:10.880 of level three cache. There's our 18x 00:05:08.479 --> 00:05:17.680 256 kilobyte. Wow, this thing has multiple megabytes. So 4 and 1/2 megs of 00:05:15.199 --> 00:05:22.400 level two cache. Never mind the the level three cache. 00:05:20.160 --> 00:05:26.960 Memory is running in quad channel. Now, something I'm curious about is I 00:05:24.639 --> 00:05:32.160 obviously this chip is not designed for gaming strictly speaking, but I wonder 00:05:30.240 --> 00:05:35.840 will it support the regular unbuffered memory? 00:05:33.680 --> 00:05:40.080 Let's uh let's take a run at it, shall we? So, since we're playing around 00:05:37.919 --> 00:05:44.720 anyway, uh we're going to boot her up with our quad channel kit of Corsair 00:05:42.400 --> 00:05:52.720 Vengeance memory. And I'm just going to see if like XMP memory works because I 00:05:49.280 --> 00:05:54.320 can't imagine why it would, but I also 00:05:52.720 --> 00:05:59.280 can't necessarily think of a reason why not. I mean, would Intel turn a feature 00:05:56.320 --> 00:06:05.080 off on the $4,500 processor? 00:06:01.840 --> 00:06:05.080 It boots. 00:06:06.479 --> 00:06:14.319 And if we go to advanced mode, XMPP is an option. Oh, I don't think 00:06:12.400 --> 00:06:21.039 that's going to work, though. I don't know. Let's Let's see. Uh, yeah. You 00:06:18.800 --> 00:06:26.319 know what? Okay, I'm going to let it try. 00:06:24.880 --> 00:06:32.960 I don't think this is going to work, though. The way XMP works on the consumer grade 00:06:30.160 --> 00:06:38.080 chips is it'll actually run the CPU strap at 125 megahertz. Um, and then 00:06:36.880 --> 00:06:42.880 it's actually going to run the base clock slightly out of spec. And I really 00:06:40.800 --> 00:06:46.800 don't think the Xeon is going to like that. If this boots, I'll be floored. I 00:06:45.280 --> 00:06:49.800 don't think it is. It's in a loop right now. 00:06:49.919 --> 00:06:58.400 It's not happy. Time to clear Simos and restart. 00:06:55.759 --> 00:07:05.520 You know what? Let's do 26666 since that is a little bit of a of a safer 00:07:02.000 --> 00:07:08.080 number. Oh, not happy. 00:07:05.520 --> 00:07:13.039 Let's go for default settings. And that's it. 00:07:10.960 --> 00:07:17.840 Okay, so now that I'm booted up, I know the max turbo frequency for this chip is 00:07:15.520 --> 00:07:21.599 3.5 GHz in a single threaded workload, which means in theory you actually could 00:07:19.680 --> 00:07:28.080 game on the thing and there would be no disadvantage compared to running on a 00:07:23.520 --> 00:07:32.400 consumer chip. But what I don't know is 00:07:28.080 --> 00:07:35.440 that in spite of its 2.3 GHz base clock, 00:07:32.400 --> 00:07:39.280 what it will run at under a sustained 00:07:35.440 --> 00:07:40.160 all core load. So there's two things I 00:07:39.280 --> 00:07:45.599 want to know. I want to know what voltage it actually runs at when we're loading it. And I want to know what 00:07:43.440 --> 00:07:50.080 frequency it actually runs at. So you can see turbo boost and speedep jumping 00:07:48.560 --> 00:07:58.039 around changing frequencies here. So let's start our ID to 64 stability test. 00:07:54.319 --> 00:07:58.039 And then just for fun, 00:07:58.800 --> 00:08:05.759 100% usage of 18 cores and 36 threads. 00:08:02.400 --> 00:08:09.840 So it looks like we're going to get a 00:08:05.759 --> 00:08:11.440 multiplier of 24. Not bad. So that's a 00:08:09.840 --> 00:08:16.080 little bit higher. Not much higher. It manages to turbo boost another, you 00:08:13.840 --> 00:08:22.240 know, 100 MHz, but that's better than nothing. And the core voltage stays 00:08:18.720 --> 00:08:24.319 extremely low. So these are super low 00:08:22.240 --> 00:08:28.720 power cores. So, I'm going to go eat lunch and we'll come back and find out 00:08:26.479 --> 00:08:31.919 what these temps settling at. All right, so this is actually better than I 00:08:30.319 --> 00:08:39.360 expected. It's been sitting here running for 28 minutes and our highest core is 00:08:36.320 --> 00:08:44.720 51, 52°, 00:08:39.360 --> 00:08:47.760 something like that. That is ridiculous. 00:08:44.720 --> 00:08:48.800 So, let's overclock it. 00:08:47.760 --> 00:08:56.080 Now, I don't think we're going to be able to do much. Xeons straight up do 00:08:52.560 --> 00:08:58.080 not allow you to change the multiplier. 00:08:56.080 --> 00:09:02.480 They're not consumer chips. They're not intended for that. So, the only thing 00:09:00.000 --> 00:09:05.519 you can do on a Zeon is increase the base clock speed. Now, we already saw 00:09:04.080 --> 00:09:09.360 what happened when we tried to run a different strap. So, that's clearly not 00:09:07.519 --> 00:09:16.320 happening. We're not going to 125 base clock. So, all we can really do is maybe 00:09:12.560 --> 00:09:19.839 we can get, you know, 5 MHz 00:09:16.320 --> 00:09:23.839 more on the 100 MHz stock base clock. 00:09:19.839 --> 00:09:25.920 But remember guys, 5 MHz more is 5% to 00:09:23.839 --> 00:09:29.680 our to every clock speed we run at every boost level. So, that could result in a 00:09:28.320 --> 00:09:33.519 significant performance improvement, especially when you multiply it by 18 00:09:31.440 --> 00:09:39.600 cores. That's that's like adding almost another core to the to the chip. 00:09:37.680 --> 00:09:43.920 Look at that. 105 MHz. We are now seeing Well, I guess 00:09:42.399 --> 00:09:49.040 we should we should compare apples to apples. So, let's run our stress test. 00:09:47.200 --> 00:09:52.480 2.73 GHz. 00:09:50.720 --> 00:09:59.680 Rounding up a little bit. Come on. It's still impressive on our 18 core 00:09:55.760 --> 00:10:01.600 processor under full load. How very 00:09:59.680 --> 00:10:06.000 interesting. Now, obviously, something like this is going to need a longer 00:10:04.080 --> 00:10:12.080 stress test if you're going to deploy it in any kind of environment. Um, and 105 00:10:10.000 --> 00:10:17.519 megahertz from my understanding is actually pretty borderline. So, we're 00:10:15.600 --> 00:10:23.760 gonna use uh we're actually going to use ASUS's real bench in order to hit this 00:10:20.880 --> 00:10:28.240 bad boy and find out over a longer period of time 00:10:25.839 --> 00:10:32.079 if she's stable. So, Real Bench has two modes. One of 00:10:30.240 --> 00:10:40.360 them is a stress test mode and one is a benchmark mode. So, let's go ahead and 00:10:34.240 --> 00:10:40.360 give her up to 16 gigs of RAM and start. 00:10:42.160 --> 00:10:47.040 All right, so Real Bench, I thought it 00:10:45.360 --> 00:10:51.360 was still running for quite some time, but the system actually locked up after 00:10:49.120 --> 00:10:55.399 about 14 minutes. So, we're going to dial this puppy back. 00:10:55.680 --> 00:11:01.680 Let's say 103 MHz. Let's try this again 00:10:59.600 --> 00:11:08.399 and find out if we can get her stable. Okay, so now we're booted up at 103 MHz 00:11:05.040 --> 00:11:12.240 base clock. So let's do our eight hour 00:11:08.399 --> 00:11:15.519 stress test again and find out if this 00:11:12.240 --> 00:11:19.360 is stable. This will still give us a 3% 00:11:15.519 --> 00:11:22.399 improvement in frequency which will come 00:11:19.360 --> 00:11:24.640 back actually. Oh, this is not the same 00:11:22.399 --> 00:11:28.720 kind of a workload. So you guys can see that even in a fairly multi-threaded 00:11:26.720 --> 00:11:36.160 workload, we're actually sitting at 00:11:31.680 --> 00:11:39.360 2.88 GHz per core. 00:11:36.160 --> 00:11:41.040 Wow. So it's only in that synthetic IDA 00:11:39.360 --> 00:11:49.200 workload that the multiplier is only going to 24. That is really impressive. 00:11:44.560 --> 00:11:52.880 So, it looks like our 8h hour run 00:11:49.200 --> 00:11:55.440 actually did succeed, which means that 00:11:52.880 --> 00:12:01.200 we're stable for all intents and purposes at a 3% overclock on our Zeon. 00:11:58.640 --> 00:12:05.839 So, now all that's left is to run a couple benchmarks and kind of see how it 00:12:04.000 --> 00:12:09.680 goes from there. So, uh, those of you who are into Cinebench, why don't we go 00:12:07.760 --> 00:12:12.500 ahead and do a multi-threaded Cinebench R15 00:12:11.279 --> 00:12:17.279 bench. 00:12:17.360 --> 00:12:20.360 Wow, 00:12:20.720 --> 00:12:27.519 that is incredible. 00:12:23.920 --> 00:12:31.120 I've watched this run a lot of times. I 00:12:27.519 --> 00:12:35.279 have never watched it run like that. 00:12:31.120 --> 00:12:39.440 Holy crap. So I think is this your Xeon 00:12:35.279 --> 00:12:41.120 Ed 2.6 GHz 12 core 24 thread? 00:12:39.440 --> 00:12:49.200 Uh 2.7. Okay. So this thing 00:12:45.600 --> 00:12:52.160 is double the performance of an X5650. 00:12:49.200 --> 00:12:57.440 It is two and a half time like 2.3 times the performance of a 3930K 00:12:55.440 --> 00:13:01.519 in Cinebench. Something that scales reasonably well. Now, let's run the 00:12:59.440 --> 00:13:08.079 single core benchmark because I want to see how performance is actually scaling 00:13:05.440 --> 00:13:10.800 going from 1 to 18 cores. This one's going to take a while. I don't think 00:13:09.120 --> 00:13:14.399 we're going to film the whole thing. You see, this is this is more like what I'm 00:13:12.399 --> 00:13:18.160 used to seeing. So, we mentioned this before, but this 00:13:15.920 --> 00:13:23.600 is the Ring Video Doorbell. Basically, what it does is it allows you to preview 00:13:20.800 --> 00:13:29.600 using video. It's got a 720p camera on your iOS or Android device who is coming 00:13:26.959 --> 00:13:34.079 to your door. So, whoever it is that happens to be coming up the steps and 00:13:32.000 --> 00:13:38.880 knocking or in this case actually ringing the bell, which is of course his 00:13:36.000 --> 00:13:44.320 cue, but he totally missed it. What? Oh, hi. I have a package for you, 00:13:42.000 --> 00:13:48.000 Lionus. Get inside and get to work. I thought 00:13:46.560 --> 00:13:53.440 this was going to be my dual socket motherboard. Go, go. Jeez. Anyway, the 00:13:51.440 --> 00:13:58.880 point is it can be powered with your existing doorbell wiring or it can use 00:13:55.680 --> 00:14:01.120 its 5200 mAh internal battery and it 00:13:58.880 --> 00:14:06.000 works over Wi-Fi. It's got motion detection and is basically just a way of 00:14:03.360 --> 00:14:10.560 knowing who exactly it is without actually going outside or unlocking your 00:14:08.720 --> 00:14:15.519 door or whatever else the case may be. It's available for $1.99 and for $3 a 00:14:13.040 --> 00:14:20.160 month you can actually archive the footage off of it for six months. So, 00:14:17.760 --> 00:14:24.320 it's $30 a year if you buy a year at a time. So, head over to the link that 00:14:22.560 --> 00:14:28.639 I've got posted below me if you want to learn more about the Ring doorbell. All 00:14:26.720 --> 00:14:32.720 right. So, here we go. Obviously, performance is a little bit lower. 00:14:30.560 --> 00:14:39.760 Actually, less than 18 times the performance. So, you can see that what 00:14:35.120 --> 00:14:42.079 we lose in intercore communication and 00:14:39.760 --> 00:14:48.000 splitting up the project efficiency, we actually gain back and then a little bit 00:14:44.560 --> 00:14:50.160 with hyperthreading. So, performance 00:14:48.000 --> 00:14:56.880 looks very promising. This is an extremely fast machine. Um, but I think 00:14:54.959 --> 00:15:00.480 that's pretty much it for today. I'd love for you guys to leave a comment and 00:14:58.240 --> 00:15:05.760 let me know if you'd like to see any other completely, you know, different 00:15:02.720 --> 00:15:07.199 kinds of of stuff with this thing. I 00:15:05.760 --> 00:15:10.560 already have some ideas. I'd love to see how it actually does in in gaming 00:15:09.040 --> 00:15:14.480 workloads, for example. I'd love to see if we can if we can crack the top of 00:15:12.800 --> 00:15:20.880 some leaderboards based on that we're using processors that 00:15:17.279 --> 00:15:22.320 literally nobody has. So, uh, yeah, I 00:15:20.880 --> 00:15:27.519 guess that's pretty much it. Thanks for watching, guys. Like the video if you liked it, dislike it if you thought it 00:15:25.519 --> 00:15:30.240 sucked, and, uh, leave a comment if your feelings are more complicated than that, 00:15:28.800 --> 00:15:33.040 preferably in our forum, which is linked in the video description. also linked in 00:15:32.000 --> 00:15:37.760 the video description. You can buy a cool t-shirt like this one, give us a monthly contribution if you think what 00:15:35.680 --> 00:15:41.680 we're doing is awesome and we need more CPUs like this or even just change your 00:15:40.079 --> 00:15:45.680 Amazon bookmarked one with our affiliate code because if you bought a Zeon like 00:15:43.440 --> 00:15:49.680 that on Amazon with our affiliate code, that would help us a lot, like a lot. 00:15:47.920 --> 00:16:03.649 So, uh, thanks again for watching and as always, don't forget to subscribe.