Name: Announcing the Next Generation of Dell Engineering Workstations for SOLIDWORKS Users
Uploaded: 2026-05-12T18:30:10.030Z
Duration: 1 h 2 min 33 s
Description: Announcing the Next Generation of Dell Engineering Workstations for SOLIDWORKS Users

Transcript for "Announcing the Next Generation of Dell Engineering Workstations for SOLIDWORKS Users": Good morning, good afternoon, good evening. Thank you for joining us today. So if you're frustrated with your system, ready to chuck it out the window, if you're interested in the new flashy best brand new machines, or if you're just interested in finding out what is going on with computer technology, you're in the right place. My name is Bryan Pawlak with GoEngineer, and I'm happy to have Ken Flannigan here today to, to tell us and talk about the, the latest generation, Dell, precision, workstations. Thank you, Ken. Awesome. Thanks, Look forward. looking forward to see what, what you got in store for us. You know, this is our yearly, kind of our yearly update here, as you guys, bring us some of the newest processors and newest graphics. Yeah. So with that, to be here. I'll, I'll go ahead and, we'll start, pass it off to you and then, let you talk about some of the new systems. And then I'm gonna follow you up with, some of the, latest, benchmarking research that we've done, echo engineer, showing the performance of, some of these systems and where we stand, today. Great. Alright. Off to the races. Yeah. Thank you for having me. Ken Flannigan with Dell Technologies. Importantly, before we jump into some of the technical content, you as a GoEngineer customer, you get a discount better prices than you can get just if you went on dell.com. Of course, if you have, you know, deeper questions, there's connection to, our sales organization who understands SolidWorks workloads. But if you just wanna add a rewards code to your Dell Technologies account, you can get that through GoEngineer. The code right there is is highlighted, GEDFY26. You have to log in to a an account on Dell Technologies. But when you put that discount code in there, everything that you buy on Dell Technologies for a whole year is discounted. So you'll see that discount when you when you check out. I just wanna make sure you guys are aware of, this benefit as a GoEngineer customer. So now let's get into talking about products, talking about hardware. Now most of this is probably not new information to you. Right? SOLIDWORKS is gonna feel the fastest, when the CPU has the highest frequency. So when we when we talk about CPUs, we're mainly looking at that turbo frequency as the main kind of speed speed boost. Now that's not the only thing that impacts performance in SOLIDWORKS. We do have the GPU. And the way those two work together is what we certify when we say we have a certified system, which is really important for all SOLIDWORKS users. For support purposes to have a certified system, we put a lot of work into certifying not just one component, but the overall system. But the GPU, also really important. We we love talking about, NVIDIA Blackwell generation pro GPUs, not just in frame rates, although it absolutely will impact how fast you can move through larger assemblies, but also rendering, you know, GPU accelerated simulation. And importantly, we're gonna spend more time today on these on device AI features and kinda where the future is taking us, because a lot of you have systems that are very capable to start experimenting with these new tools. Okay. Next slide, Yeah. please. And those and those, Blackwell is the newest generation, but SolidWorks does support all the way back through, you know, Pascal cards, the RTX, series, and the the ADA and the Ampere cards. But, yeah, the latest generation is definitely those, those new Blackwell cards. Yeah. It's it's a lot of it's a lot of acronyms. So, you know, you could look at, like, an NVIDIA, RTX 5,000, an NVIDIA RTX a 5,000, an NVIDIA RTX 5,000 ADA, and an NVIDIA RTX Pro 5,000. And I just described four generations of, the same class of GPU. So you gotta put your decoder ring on a little bit when you're dealing with some of these system specs, but when you see pro in all capital letters, that is the newest generation of NVIDIA cards. Cards. Wanted to share, some internal benchmarks that a team member, of mine, Trey Morton, completed just to show the kind of investigation that we do for specific workloads. He's on the performance side, so he does a lot of this kind of bespoke testing. So in this case, we actually compared a Dell Pro Max Plus 16 to a Lenovo similar system. Now they're similar when you look at the the processor. They're similar when you look at the cores and the RAM and the GPU, where they're not similar is the power supply and the speed of RAM, that RAM frequency difference. So this is actually a SOLIDWORKS benchmark result that we run. But to get a little bit beyond that about why this makes a difference and not just we provide more power, in this specific situation, but what does that actually mean when it comes to how a system performs over over time? So So if you can go to the next slide, please, Bryan. Well, first, before we get there, this is benchmarks on two systems happening simultaneously. And you can see as you get to the end, that blue line stops before. So that's the time that, the Dell Technologies power Technologies machine Technologies system system in this case is faster. Next slide. Because I think what's really interesting now there's some c CPU and GPU workloads, with some different standard benchmarks. But if you look in the upper right chart, that upper right graph, what you're seeing is the power and power from the Dell Technologies power Technologies machine Technologies system machine to, in this case, it's an HP ZBook. And when we deliver a higher consistent power, you can see the HP ZBook starts a little higher, throttles down, pops back up, and then slowly, degrades to to about, 80 watts, and we keep it constant. So that's kind of one of these hidden things about workstations and the difference between two very similar spec workstations, but a difference in performance. Right? When you hear the fan kick on or you feel the CPU throttle down. Right? When you're giving it consistent power, you're gonna have less of those types of experiences. But we just wanted to share this. He shared it on internal call, and I thought it was, interesting to see some very tangible differences between, between similar workstations from different OEMs. Alright. So now let's talk about, products. And, guys, we have new products and some old names and kind of new names of new and old products. We have we have a whole bunch of stuff going on. So, there will still be the workstations that have been available the past year or so. So if you go to the next slide, Brian, so that so the first kind of marquee fastest frequency CPU, you're not gonna put multiple GPUs in this, but you can scale it up to a 6,000 Blackwell card, really, really powerful card. This is gonna be the fastest performant machine and this machine is will continue. Right? So Dell Pro Max tower t two, kind of a normal size tower. Once you get bigger than this, you're gonna talk about Xeon processors and Threadripper processors. Those are not ideal for everyday SolidWorks users. The Intel Core Ultra processors that are available in this machine are. So this is big. It's scalable. It's not too big, and it it prioritizes frequency. This is still gonna be a a product. Yeah. This one is definitely a, a mainstream, tower for for a lot of our users. Absolutely. Dell Pro Max 18 plus. So still Pro Max. This product has been, been out for, several months. I think a little over almost a year now. And this is the most scalable, most performant mobile that we have. It does come in a 16 inch as well. The 18 means an 18 inch screen. But when it's bigger, you're gonna have better thermals. You know, you're gonna be able to cool and power those NVIDIA graphics more effectively. So this is this is a pretty big laptop. If you truly need a mobile desktop replacement, this is a topped out 18 plus, in this situation, is gonna be your best bet. I think the big thing that the that I've helped everybody is that's the same footprint as older 17 Pro. Max. But because we could stretch that screen out, that's how we got that extra inch. So it's not that it's, you know, a larger footprint than than many of our users are used to. So that's, just the new naming, 18 inch or 16 inches. But the, the sixteens are really the same footprint as a 15 inch. So, keep that in mind when, when we got these new names that, that that you're throwing at us. Yeah. Absolutely. And and new names, new names are coming. So if if you go to the next slide, this is so new brand. We went from precision workstations to Pro Max workstations, and now we're back to pro precision. So it's kind of the best of best of both worlds. Now the 16 at the end here does mean the screen size, but the seven the seven, you'll see now five, seven, and nine series on the mobiles, and those will basically translate from our entry to our premium to our plus, Pro Max, products. So I see Dell Pro Precision seven sixteen. This is gonna be, thin, light, high performance. And when it's thinner, thermals are gonna be a little bit more difficult to deal with, with, but, of course, we engineer around that. So Dell Pro Precision seven sixteen, is the new product name, and, you know, some of these products aren't even out yet. In fact, I think that the Precision seven sixteen has been out since, the very March, but we're gonna talk about some products that are gonna be released at the very end of this month. So next slide. We got a a few more just to get a little bit used to what this, new naming convention is, Dell Pro Precision five sixteen. So this is our entry product, 16 inch. And to to add more acronyms and more letters to the mix, we're gonna have even more offerings that gravitate more toward the entry side. So we, you know, we'll use the term entry. We just mean kind of the the lowest end configurations, the low the the smaller chassis, not necessarily thin and light premium, but but physically smaller. So Dell Pro Precision five is that kind of first entry product. You can still put a 2,000 class graphics card in it. You can still put the same Core Ultra nine. Power is gonna be a little bit different and, you know, in general, you know, the experience is gonna be better in a in a seven, but still, it's important for collaborators, for SolidWorks to be able to use a workstation class system. Even if they haven't before, they'd probably start with the Dell Pro Precision five. We have lots of changes in this entry space. If you look at our overall portfolio, we are definitely expanding into, this entry or value space more and more. And I think that's gonna be important for SolidWorks users and IT administrators to understand different price points that you can enter into, a a certified system. Now not all certifications have been complete. This product is not out yet. It will be out at the end of the month. This actual product slide was released yesterday. So this is brand, brand new information. But if you're really interested on in the performance that you'll be able to drive out of a mobile, an entry mobile, you might be looking at these slate, series. So that's Dell ProVision five sixteen, same one as before, but that s means slate. And what that means is really, integrated graphics. So this is non NVIDIA graphics using the Intel Arc Pro graphics, 12 x e graphics to do your traditional visualization, but also to run things like Copilot and some smaller AI workloads locally. And we also have, not just the Intel, version of this, but we also have the AMD version of this. So if you go to the next slide, of course, a little bit confusing because the product name is exactly the same, but in our perspective, you know, the workstation, class of system needs to be configurable and needs to be scalable. So now we have configurations that leverage the AMD, the APU, or just basically, it's an integrated graphics on the CPU. And really, really high performance ISV certifications are currently going on with this system. And, again, this is slated for release at the end of this month. So it's another offering in the in the lower end, lower cost. I know, you know, hardware costs and memory costs are really, really astronomical right now if you if you go really big. So we challenge you guys to look at some of these entry products and and test them as they get released, as they're certified to see if they're a viable option for your for your your engineers and your collaborators. Dell Technologies Pro Precision. We're all getting used to it. Alright. Now I I have this section, Yvonne, on on device AI, And it's it's pretty pretty high level, but this is it's really important to understand conceptually what products do what because we're we're out there releasing more products that may or may not be for the type of AI you're thinking and may or may not be for the type of SolidWorks workload that you're seeing either. So let's take a look at on device AI. So in in general, we we know that AI workloads are gonna be done in the cloud. They're gonna be done on the device. They're gonna be done, between the two. They're gonna be this hybrid approach. The reason that on device AI capability is important is, for latency, for cost effectiveness, for privacy of data, and offline capability. So there's practical reasons that part of those workloads are gonna be held locally. There's also a whole world of, operational cost that comes in when you start implementing AI solutions, API call costs, you know, all sorts of things that add to, an operational IT expense. So you wanna be really careful with how you implement AI, especially when you're getting into these use based models. On device AI bringing things down to your device is a strategy that helps mitigate some of those. It's not gonna take care of everything, but it will mitigate some of that cost. Next slide. So we can bucketize local AI workloads in in three areas. We have, software feature enablement. So as, features are developed and released in software, you may already have you probably already have you have a NVIDIA RTX card. You already have a system that's gonna be able to run those accelerated workflows. And you may not even realize that you're using an AI feature at the time, and a good example of this is DLSS, deep learning super sampling. Right? That's, coded into the NVIDIA cards and leveraged by software that allows you to, in this case, render at a lower resolution and upscale through pretrained models. So not necessarily what we always think about when we talk about AI, you know, robotics. This is a little bit more tangible AI feature that's been around for for quite a while, and that's that's continuing to happen. Software feature enablement locally. And if you have the NVIDIA RTX cards, you're gonna be well equipped to start experimenting with those. I'm gonna skip the middle section. I'm gonna go to the right for a second. AI development sandbox. So there are absolutely AI developers that start local and then scale up into on prem data center or into the cloud. When you develop, within the NVIDIA ecosystem, that transition from desk side to data center is is seamless. It's all one one environment. So that there's an advantage on on playing in that, kinda on device scalable to the to the data center. And there's lots of resources available. Build.nvidia.com is one of those resources. But we know not everybody is gonna be a developer. Right? So if you're if you're curious about development, if you have an RTX card today, you may be able to start experimenting on the machine that you have. But we're at this really interesting intersection of technology, which has to do with this this middle box, which is the agentic AI coworker. And you may have heard about OpenClaw or NemoClaw, these, tools that take a large language model, add communication channels, and add skills and memory and really allow for a separate coworker to be desk side, controllable. These are these are tools that, you do not install on your own system. So you'd really think of it as, like an intern or an assistant that would sit next to your system. And those those tools are very, very different. The hardware for that workflow is very, very different than what it'd be for an engineer using using SolidWorks. It's the most exciting thing happening right now, and I'd really encourage you guys. If you have never heard of OpenClaw or NemoClaw, be very careful. I'm not condoning installing this on any system that you that is connected to anything of yours, but really important bleeding edge technology that you need to be aware of. I think they I I think you're absolutely right. The coworker, applications right now are are where where a lot of things are happening, including SOLIDWORKS. Dassault is is introducing those as, Aura, Leo, Emery. So those are those are brand new on the horizon. Those agents that help answer those questions and guide you through the design process. And that's what companies are building. You see more and more of those, AI agents, to do those and you're not too late. tasks. Right? This is all stuff that's been happening the past couple months. And, you know, every week that there's news and enhancements and new releases. So being aware of it and starting to follow how it may impact your workflow is is really what we're calling you to action to do do today. I know not everyone's gonna have an AI coworker, you know, by the summer, but understanding how it how it could impact engineering workflows is really what what we're after right now. But I think why we bring this up for is that these all these brand new Dell Technologies systems are being built to handle these capabilities. So if you're on older workstations, they're not equipped, to be ready for these AI tools, the most optimally. And so one of the advantages of of your next upgrade is that you're gonna get these new systems that have, intentionally built the hardware, to help us with these. Awesome. I can move on to the next slide. I just love this template, I guess. I got a I got three in a row to take this, I think. But, so just wanted to talk about AI workloads and the things that are the common bottlenecks, because they they might be different. They might not be something you consider. First of all is compute power. Right? So if you're if you're actually doing model training or image creation, they're measured in teraflops. There's also a floating point verse integer and floating point four verse 16 verse 32 and lots and lots of details. Where this crosses over into computer aided engineering is really the simulation space. So you might see, you know, teraflops being described as, you know, for a simulation machine. Right? Because the simulation machine is gonna be able to use that type of parallel compute power. And that's really what we're talking about, parallel compute throughput. It's not something that's gonna run SolidWorks. Right? SolidWorks is not parallel compute. For the most part, SolidWorks is how fast can I run a very small number of CPU course, which is not what most of what AI does? Most of what AI does is parallel accelerated compute. More of the con so if if you don't have enough compute power, it's gonna go slow, but it'll still complete. If you don't have enough memory or if your memory bandwidth isn't fast enough, you may not be able to complete it at all, or you may be able to have to complete a smaller task. So if you think of memory, whether it's happening on a GPU and VRAM, you can use system RAM, and it's a lot slower. This is about how much you can hold in context at once. So if it's, you know, if you have a large language model inside of an an AI tech stack and you are connecting it to multiple systems, the size of the language model itself is be it is what has to be held in context. So memory is can be a very, very limiting, factor when it comes to, the types of things you can you can do with an AI tool, especially on on device. And that's where scaling up and and having cloud services for those types of larger context AI features are gonna be important. And that and there's that hybrid approach. Right? The cloud high the cloud compute, the cloud memory, that's gonna be more expensive. If you can do it locally, you're gonna be able to save save some money. And finally, on the far right, memory bandwidth. How fast memory recalls? The Blackwall generation, you know, you might look at the cores that it has. You might look at the VRAM that it has, but the memory bandwidth is gonna be maybe the most noticeable difference in those when memory bandwidth is needed, how fast it can recall, VRAM information. Now there's new types of systems, not for SolidWorks. These are ARM based systems, and this is the Dell Pro Max GB 10 and GB 300. They don't have a separate CPU and GPU. They have an AI super chip. That's the Grace Blackwell super chip, and they have unified memory. So that's a completely different animal. But that animal is really what the agentic AI coworker, that's really the device for that. Right? It's a development device on Arm. It's also an agentic coworker device. So, again, bleeding edge stuff, but just wanna make you aware of, you know, what's for what. Right? Just because we have a GB 10, it's not for SolidWorks, not for SolidWorks users at all. So here's the two systems, Dell Pro Max with g b 10 on the left. This is a mini supercomputer. It's considered on your desk, desktop. You can remote into a machine. You can actually connect the monitor to write to it with peripherals. It is Linux based, but it has a 128 gigabytes of unified memory. And the if you went out and bought a system with a 128 gigabytes of memory today, that's a traditional workstation, you'd pay a lot more for that system than you'd pay for this Dell Pro Max with GB 10. Now it's not for every workload. It is for AI workloads specifically, Linux based workloads. It can be used as a compute node, but it has to be Linux based and and work on an ARM based system. So if there's certain workflows that you see as really needing, you know, a lot of unified memory and not having a bottleneck of of that memory bandwidth between the CPU and the GPU, which is a huge bottleneck, then this is a really interesting interesting device. Now the Dell Pro Max with GB 300, which, is is really brand new. I don't think we're even, selling them yet. This is, you know, quite a bit bigger, quite a bit, heftier, and that's 784 gigabytes of coherent memory. So that's a tremendous amount of memory to hold in context whatever AI tool or agent you have doing work on that system, let alone if you're just using it for AI development or training. So really important, these are not for SolidWorks. Yeah. Absolutely. Bryan, hit. the next button there. Just one. Not for SolidWorks. This. one. As you say here. for SolidWorks. So Yeah. yes. Super cool, Please. super. powerful system. I mean, that's what? About a four inch box. Right? That g b g b 10? It's about four inches, four or five. inches. I mean, so they have a. super processing. Processing you know, everybody can have that affordably on their desktop doing doing that kind of workload, and not asking your laptop to do it. You know, that that's that's huge if you have the right kinda application for it. Yeah. As we said, we'd love. we'd love to throw SOLIDWORKS on there. It's not gonna happen. But But that that does that doesn't mean that we can't make an impact in the engineering process when we start looking at AI tools and also considering the privacy that an intellectual property that we're we're dealing with, when we're talking about professional, you know, engineering outputs. We don't wanna train somebody else's model. We don't wanna lift our data into a black box. We don't know where it goes. And so this is a a a different approach to it where you can you can really start to think about how would I use a 128 gigabytes of unified memory at my desk side as a walled off assistant, as an assistant coworker. Yeah. Ken, as you and I were talking just before this we went live here, this stuff is changing, you know, by the week. That's how fast we're you know, things are the environments are changing around us. The tools are being developed. And so I I guarantee we're gonna find ways to use something like that to potentially drive stuff into SOLIDWORKS, in some form or fashion. I know Dassault is gonna gonna keep their reins on on on what it's gonna allow to happen on the platform. But I'm sure that there'll be there'll be some sort of tool, that can leverage that, somewhere down the road. Because, again, we're just putting the hardware out there. The hardware is there. So now now we gotta bring that hardware into, our workflows and figure out how to make this work. Yeah. Exact exactly. And, you know, it's also kind of a a a a call to action. If you have workloads that you are looking to, implement, experiment with the agentic AI, how that interacts with SolidWorks, I would love to talk with you. I'd love to kinda hear what your ideas are, and see if there's opportunities for partnership where we can really see and start building out what this feature of agentic AI co working in the engineering industry looks like. So with the last slide, I have a a QR code to my LinkedIn. Feel free to connect on, on LinkedIn and reach out there. As always, it's great talking to you guys, and appreciate the invite today, Bryan. Absolutely. Thanks. Look forward to, getting those next machines and and, seeing how they, perform and how how another generation of the Intel processor, keeps pushing us forward. You know, the, one thing we talk about with a lot of those systems is is now that NPU, that dedicated processor, for AI, processing. The big the big, thing to know is that today, SOLIDWORKS is not using that processor. It's still GPU based, but that doesn't mean in the next like we said, this stuff's changing. In the next two, three years, four years, You know, who knows what Giselle will do, and how they will, again, use that hardware that's now available. So, it's another disadvantage. We know that other AI tools as as we do our everyday work, those should be using those processors. There are tools out there that are using that. We're just waiting for SOLIDWORKS to use, some of that. And and we know we know that the r and d team over there is working on it. So so thank you, Ken Flannigan. Definitely, appreciate those updates. So what I'd like to do is, go ahead and talk about some of the benchmarking data and performance data that we've, we've seen at GoEngineer. So we, we have been doing, our own internal research, for many years. And these are just some of the, these are the models, that we use, and we range from about 1,500 components all the way up through 86,000 components. All of these have been rendered, using Visualize. So, my coworker Dean, did a fantastic job putting this together for us. So this is, all fully rendered using Visualize. Alright. And we try to, try to cover a wide spectrum to cover, all the different uses, of our customers. So when we talk about performance, we break the system down into four key areas, and that's our graphic card, our processors, our hard drives, and our RAM. So we built our own proprietary tool, using SolidWorks API, so that we could run the everyday operations, on any customer model. And we do our, we do standard, opens and closes and pack and go. We do feature creation. We do assembly creation. So we really wanted to replicate, an environment that, the test, operations that users use on a daily basis. We can also go in, and what we do is we isolate, each setting that we look at. So when we make a change, we're only changing one setting, and we're testing that setting, so that we can understand that impact. So we went through and we picked out, 14 settings, that we felt would impact performance, and we started testing those. And then in 2019, SolidWorks released the enhanced graphics. So the pipeline that pushes more data over to the GPU to utilize more of those high end, graph cards. So now we can test all of these settings with the enhanced graphics on or off. And then they move the silhouette edges onto the graph card. So the computation of of silhouette edges is now an additional option in there. So this brings us up to 42 different settings that we're looking at, on a yearly basis. So just to give you a a a brief idea of of what we've looked at, we started and we've got the, tenth gen 36, Precision thirty six fifty. That was the tenth gen Intel. We jumped up. We have the, twelfth gen, in the thirty six sixty, so one generation newer. Then we went up to the thirty six eighty, which was the I nine fourteen nine hundred k. So that was the last of the core I family, generation. And now, Intel's moved into the Intel Core Ultra, seven and nine. So we got a system with the Ultra nine two eighty five k. On the laptop side, we've been looking at, the precision seventy seven sixty, that had the any eleventh gen, I seven. We looked at, the next generation, the seventy seven eighty, with the thirteenth gen. We looked at the 50, fifty six eighty, which had a also had the, thirteenth gen, I seven in there. And we got the, the new Pro Max 16 premium. So can, as you said, that's a big old tongue twister to keep those, those acronyms straight for you. K. Then the Ultra nine two eighty five h. We got the Pro Max 18 plus. Also had the Ultra nine two eighty five, HX in there, with the 4,000 Blackwell. Dell helped us out and gave us the Pro Max 14. So this is a 14 inch laptop. This had the AMD Ryzen AI nine HX Pro three seventy. I think we need a few more letters in here. K. Interesting thing about this system is this has the, AMD Radeon eight ninety m integrated graphics. So there is no discrete graphic card in this system. And the last system, we had was the the new Pro Max 16, which also had the AMD Ryzen, AI nine, HX Pro three seventy. But this one, is interesting because it has the ability to include the 1,000 Blackwell graphics and the AMD, eight ninety m integrated graphics. So you kinda get the, little bit of the best of both worlds in there, and we can go back and forth between those cards. So for the video cards, we've been testing, working with NVIDIA, as well as AMD for, quite some time. We've we've gone all the way back through the Pascal cards, and even, further back. But 2026 still supports, the Pascal cards, the Turing cards, the RTX, the Ampere, the ADA, and certainly now the Blackwell. And we've been working with, AMD. We've had the the w 6,000 series, and we've been looking at a couple generations now of this AMD integrated graphics. So for our testing, we we commonly look at FPS or frames per second. It's it's a common measurement, for graphics, the number of different tools. And in fact, you can even go into the registry, and turn on the OpenGL print statistics and get these statistics across the bottom of your screen. Now when we focus on graphics, we often talk about tessellation. And why does tessellation matter? K. So the tessellation is the triangulated data. So it's kinda meshing the data for graphics, similar to what we see in, like, an STL file. So the more triangles we generate, right, the larger the file size and the slower the performance. And why does that happen? It's because of that math. All that math is hard, and and I certainly don't wanna do any of that math anymore. But that's what we're asking our systems to do. It's crunching all those calculations. So the more curves, the more blends, that you have in the geometry, the more triangles it takes to generate that, geometry. And what does that cost us? Well, more CPU and VRAM usage. We start running into performance and stability problems. K. So if we look across, we're seeing some interesting things. So this is looking at a our very first model, a 1,500 component assembly using the SolidWorks, default settings. We used the shaded without edges, with the enhanced graphics on and the silhouette edges off. And as expected, the larger the video card, the more phrase per second. Right? We expect we we'd hope to expect that out of our card. K? And and we do see that. So we have the 4,500 ADA. We, produce a little more than 4,000 level cards, which are higher than the 2,000 level card. Now the last two categories here, the AMD Radeon eight ninety m is that integrated workstation graphics, and the 1,000 Blackwell is also, on that laptop. And what we see is it's capping off at 15 55. And we've done some more research in digging through this, and there seems to be, some indication that that frame rate is being capped by the monitor refresh rate or the the the laptop screen refresh rate. So on some further testing, if I dropped it down to 48 hertz, it capped out at 48. So we're we're expanding, digging into this. But what I you know, the theory is here is that, you know, once you've hit that limit of your hardware, we don't need the card to work anymore. K. So, again, these are the averages that we get. So there are certainly, individual individually much higher rates that are touched, But this is our averages that we are seeing. Alright. When we talk about graph cards, we often talk about visualize, photo rendering. Okay. The visualize is the program that is is absolutely built to handle and take advantage of that that CUNA COR processing, that we get in these workstation cards. As you go up into the newer generations, we keep getting faster. We get more, more CUDA cores. And as you go up in the size of your video card, we're getting more and more of that processing power. So, pretty pretty, true, to what we see, in our results. So the larger cards, certainly are, gonna render much faster. But pretty much any card that is supported by SOLIDWORKS today, the Pascal cards are the last of the generation that that do not have the built in ray tracing. And there was a significant jump when we got into the RTX, the ray tracing. So if you're gonna do a lot of visualize, we definitely, recommend getting up into any of these newer newer graph cards, and, certainly, the Blackwell, is is a very strong card for that. Alright. RAM. How much RAM do we need? So RAM RAM is critical to a point. You know, every part of every assembly, will reach that point. And what happens when we don't have enough RAM? Well, then we start using system swap. So we use hard drive virtual RAM. It slows everything down, and we start running in stability issues. So I took that sewing machine, and we created some patterns. So we got two different, patterns here. I created a layer pattern, and the blue line is a pattern where every instance is a true patterned instance. So the unique component count doesn't change as you add copies of that data for each instance. The orange line represents a pattern where each instance of the pattern is a completely unique, set of data. So every, every instance we add adds 200 unique components. K. So it took about 26 instances before I ran out of 32 gigs of RAM. And we were up into about 68 gigs of virtual RAM, 8,000 unique components at at 40 instances. Now where this got interesting was that lined up our data, our other models that we've been testing. I overlaid it on this curve. And when you, stack up the number of unique components, it almost matches up to the total run time. Kinda interesting correlation. Until we, until we ran the NASA crawler with 86,000 components, only 600 unique, and we just kinda threw that whole, theory out the window. But, and it's very interesting. So what we conclude from this is that today, we pretty much need 32 gigs of RAM. Any system that I spec out for, our customers is gonna be at a minimum of 32 gigs. K? Sourx, does have a minimum of 16 on their website. K. And if you're only gonna do a part here and there, you can get away with that. But Windows 11 eats up 10 to 12 gigs. And about 80% of users are still good at 32 gigs. Larger models, if we're getting into, you know, these 10,000 component assemblies, these 20,000 component assemblies, we're starting to climb maybe just over that thirty thirty two gigs. So we're sitting between thirty two and sixty four gigs. And once you have enough RAM, you can't force your system to use more. So what we don't want you doing is overspending on RAM, and we want you spending that on other areas of the system. K. When it comes to hard drives, k, believe it or not, we should not be on any mechanical drives. I still get and find a a rare customer here and there that is running on a mechanical drive. The old spinner drives, when we moved from the spinner drives to the, solid state hard drives, we got a 15% jump in performance. Once we got to solid state, we really did not see a big significant jump, in performance, inside of SolidWorks when we went from the gen three to the gen four, and we're not seeing it from the gen four to the gen five. So the drives are faster. And we're doubling, the amount of, traffic number of lanes that it has or, reading and writing data. But inside, that is SolidWorks, that is no longer the bottleneck. K. So so we then focus on our CPU. Alright. The biggest, most critical factor of our system. So what I do with a system is I take the processor model, and not every manufacturer is gonna have this all this data on their their website. Sometimes they'll list the processor with just the base speed, but all processors have two speeds. They have a base speed and a boost speed. And and in SolidWorks, single threaded process, we're interested in that boost speed as the highest potential that that processor can hit. So if you take your processor model number and you throw it into Google and you go to intel or the arc.intel, you'll find the, the the manufacturer, data here. And I key in on a couple numbers here. The, max turbo frequency, and I'm paying attention to the number at the bottom, the, the power consumption. K? And Ken alluded to that. You know, sometimes with the higher power, provided to a system, means it can push and maintain higher levels of performance. And when we look at things like laptops, right, the big thing here is that not all processors are equal. K? You can have two thirteenth gen processors and have different frequencies. They have different, power wattage. The, the interesting thing with the the AI or the AMD systems, they're doing that on 54 watts. So we're getting comparable performance on a much lower power consumption. So why is that important? Because that boosts your battery life. So when we're on the go, that means we can get longer battery life and not have to be plugged in the whole time. K. So for the Core I processors, we were able to come up with the equation from from the to the eleventh gen. For every gig you increase that clock speed, it was approximately 25 boost in performance. And that held pretty true. When we went from the eleventh to the twelfth gen, there was a change in the architecture. There was an additional 25 to 30% jump in performance. It's the first time we saw that significant of a gen over gen increase, in a single chip. So we anybody on the eleventh gen or older systems, we definitely need to be considering, looking at these upgrades. Because you are looking at a minimum of 25 to 30%, plus the additional getting up to the latest, systems. So here's what that looks like. So this top grouping of lines, shows us the sixth to the eleventh gen. Alright. So they're all pretty, pretty incrementally, in there. But that first gap from the blue line to the green line represents a 25% jump in performance. So that's when we went to the, that initially went to the twelfth gen, and then the thirteenth gen. And the, the red line there, represents an overclock system, performance tune system, that absolutely just hammered on, the data. So we're able to push that even harder. Alright. So that is, what we saw. That was our 24 on a 1,500 component assembly. Looking at 2026, this is looking at the towers. So we go from the blue line, which is our tenth gen, to the orange line, which is our twelfth gen. So there's that big jump in performance. And then we went from the twelfth gen to the fourteenth gen, which is about, that's about another, 15 to 20%. And then the real question was where does this core ultra fall? Well, what we're seeing is that the, CORE Ultra is running very close to this thirteenth and fourteenth gen. So we're right in line. We're not really seeing a huge increase, but we're not seeing a decrease. It's just kind of a parallel step right now into that new architecture. So they're performing extremely well, but they're not not providing a significant gain. So if you're on something like a fourteenth gen, you're not gonna see a a huge boost, in inside of SOLIDWORKS, jumping to that Ultra nine. You might wanna wait for these gen three or maybe even a gen four before you make that jump. On the laptops, again, seeing seeing the same kind of deltas. We got the eleventh gen blue line, jumping up to the thirteenth gen. That one has a had a clock speed of right around five gigahertz. That red line is actually the AMD Ryzen, AI nine. That, also had a clock speed of about 5.1 on the boost. So it falls right in line where we would expect to see it, based on what we know about clock speed. The Altra nine two eighty five h and the HX, those are all in we're very close in clock speed, so we see them, and they were a little bit higher. They're about two point four and two point, or 4.4, 4.5. Nope. I'm sorry. Scratch that. 5.4. We're up in the fives now. Five point four and five point five, I believe, on the boost speeds. When we jump to the other end of the spectrum, go all the way up into 86,000 components and look at something really, really massive, the pattern holds. We're seeing that same deltas. As we increase, it just moves up the, the scale. We just as we move, we don't see anything really significantly different. But what we do see down the bottom, we see that overclock system and that fourteenth gen, tower, that t two tower just really, neck and neck duking it duking it out there. Really hammered on large datasets. And, they're really, really nice nice to see, the performance out of those. Just a little bit of the simulation data. So we do try to to, cover that as well. And really most of the SOLIDWORKS simulation tools, they respond pretty, linear here to the clock speed. So as you increase the clock speed, we we see just a fairly linear, response in solve time. K. So the the the one I will just mention the one that doesn't follow suit, is flow simulation. So flow simulation doesn't like, some of the same, chip architecture, so we kinda change our conversation there. Much bigger, much deeper conversation. So, love to have those conversations with anybody that, needs to discuss that area. Be more than happy to do that, on a one on one basis. K. So to kinda sum that up, clock speed is king. Your biggest performance factor that you can invest in is your clock speed. The faster you solve, faster that boost speed, the faster that solve. Everything solves. Everything moves faster. Saving, rebuilds, file opening. Okay. Core I, versus Core Ultra, performing very, very similar. We are seeing a little bit of a, increase in in those those mobile systems. So we're definitely, very, very satisfied, very happy in what we're seeing in these these Ultra processors. Alright. Enhanced graphics, this is a must. If you have, large datasets, you need to, use the enhanced graphic settings. Push more data through the GPU. What that does for you is, increases in in your pan, zoom, rotate functions are gonna be much smoother. And that that makes everybody's day, using it feel feel a lot better. R and d has has worked very hard on cleaning up that performance. There's very, very little delta, with it on versus off. So if you run across an issue, the r and d team wants to know. We wanna know, so that we can get that addressed. So they've they've done a ton of work in in that area to clean up, that performance. And one thing, the biggest area that we see, is the large assembly mode. So I still talk to users, on a regular basis that turn off large assembly mode. And that is one of the worst things that we can be doing. When you turn off large assembly mode because you don't like loading parts lightweight and having to do the manual interaction, you are penalizing yourself, 40 to 50% inefficiency. K. In 2024, we we changed that. We now auto resolve components when you click on them in the tree. It makes it much, much less in your face and more seamless. So if you haven't used that, you know, go back and try it, because, because if you're loading, you know, multi thousand components assembly, you're losing a lot of efficiency, And these tools are there. So that's that's one of the biggest, performance settings that we see, and I know that users, turn off on a regular basis. If you're an existing, GoEngineer customer, we have our, application mentoring sessions. And in there, we have one for AMS hardware discussions, where we can go into a, in-depth conversation, about your systems and your needs, and help you guide you on selecting, some of these new Dell systems to make sure that you're spec'ing them out the the best way possible. So if you are an existing customer, you should log in to the GoEngineer customer portal. If you're not, you can use a QR code, and, you can schedule a schedule a session. I'll be more than happy, to discuss these in detail. And if you're not already, subscribed, and you wanna, stay up on the latest information from GoEngineer, these are just a couple links to our newsletter, as well as our GoEngineer community, where we have, techs actively, involved, and helping, users. So great way to, make sure that you're getting the latest information from us. And with that, that wraps up the date I have. Let's take a look and see if we had any, any questions or comments that we need to follow-up on. Alright. Went to, remote desktops, VDI systems. Yes. VDI systems are a trade off. Again, VDI systems, have generally have lower, clock speeds, than, you know, physical hardware. So we have done testing, on those. I'd be more than happy to, follow-up with you and have, discussions around that. But, yes, we have we I do have test data. We have looked at, different systems. Alright. Yeah. And another comment there about, using, local language models, on home systems. Yeah. Again, it's it's, that's a great place to, play with them, and then figure out, how to bring it into a corporate environment. The big conversation we hear all the time is just how do we protect, our IP? That seems to be the the the big big question nowadays. We wanna use the tools, but we wanna keep our data safe. Alright. And I think that's about all the, questions I see coming in. So again, if you have any further questions, or interested in, further discussion, please reach out. I'll be more than happy to, to connect with you, at any time. And with that, I wanna again thank, Ken for his time today, and, everybody have a a great afternoon.