Josh Walrath of PC Perspective - Telegrams for Guests

Back in the day of Maxwell 1080ti I thought (given the relative power of it at the time) that we would soon saturate even 4k at high frames with mostly high settings Then nvidia came along with ray tracing which looks great (when it works) but suddenly most cards are back to square 1 in terms of running it at all eg rtx 2060 even at 1080p. Now we have dlss scaling to make the hard to run raytracing easier (or similar) are we back on track in making resolutions obsolete for most mid-high range cards or will there be another major leap in graphical fidelity (and if so what will it be) that will present the same processing challenges for the next gen cards?

Alex

This is a silly question but could possibly make for a nice intro icebreaker and tongue-looser: What's the story behind Josh's weekly burger that we always hear on the PC Per podcast and how does Tom like his burgers? Thanks for all the work, MLID crew

Manuel Nascimento

Hi Josh and ‘Mr shmooser,’ :D As someone who would’ve witnessed the i740 and Larabee, my question to you is, what are your thought and expectations for DG2 and future generations as Intel (re)enters the dGPU market in a fashion seemingly somewhat simillar to zen. What sort of competition do you expect them to bring, especially when they’ll be buying TSMC’s wafers, and how do you think they’ll impact the ‘gamers,’ intel clearly cares so much about 😂 Thanks guys and stay awesome

AC_666__

Hi Tom and Josh, I've been following the hardware market very closely since about 2005 and I often think about what lead to Nvidia's massive growth compared to AMD and I always come back to GPGPU and multi-GPU rendering support. Nvidia really saw a lot of success pushing the GPGPU aspects of their architectures and building the largest dies they could get away with. AMD saw a lot of success by making lean rendering focused architectures with reasonable die sizes that were cost effective to gang up in Crossfire. However, by the time GCN rolled around two things were becoming apparent: CUDA rules and defferred rendering techniques (among other things) were killing multi-GPU. Nvidia was enjoying huge datacenter growth and often had the best performing flagship. Although GCN was a great GPGPU architecture, it was never first choice and largely seen as the budget option due to lack of CUDA support and was often left playing second fiddle to Nvidia in gaming performance. What are your thoughts on this, how do you feel the expansion of the GPU's role in computing from solely rendering to general purpose has affected AMD and Nvidia's ability to compete in the past and in the present (CDNA is a big departure for AMD)? Secondly, do you think the death of multi-GPU has led to the existence of monstrosities like the 3090 for gaming (a die that big would much better be sold at higher profit margin as a professional product), whereas in the past manufacturers would have been satisfied by those levels of performance only being available via SLI/Crossfire. Lastly, do you feel MCM will enable manufacturers tap into multi-die scaling in a way that makes these riduculous transistor counts accessible at a lower cost point, akin to ganging up 4870s vs. a GTX 280 back in the day? (I setup I owned by the way and absolutley loved) There's a lot to unpack here and you may have already covered it naturally in your discussions, thank you for your time and thank you for the great guests and intriguing and deep discussion as usual. Cheers, Matt

shredbird

Hi there Tom and Josh, when do you believe that low-end gpus will be rendered silly by integrated graphics? Do you think that AMD could (in the future) make a 45+ Watt APU that was capable of AAA gaming at 1440p 60+ fps? RDNA 3 or possibly RDNA 2, on 5nm or 6nm? Would a 5nm Zen 4, 8 core 16 thread CPU, with a 20 CU RNDA 3 GPU have the perf/watt to fit the bill?

Conner Seney

Hi Tom. Hello Josh. Thank you for this opportunity to ask a question. I am learning and observing how major of a role TSMC plays in the fabrication of chips for companies around the world. How would the industry deal with a loss of output from TSMC? Not getting too political however what if one day soon Taiwan was cut off from being able to sell their current production to companies that sell products outside of their region. They have spoken about building foundries here in the United States but that is years off from completion once they cement their plans. How much could Intel's production help to make up the difference and how would something like this change the marketplace for PCs and tech in general that depend on TSMC?

Mark Smith

Hello Tom and Josh; Josh aside from the "cool" factor and the "just 'cause we can do it" aspect will there be a benefit or use in gaming for ray tracing beyond the ambient environment and immersion for such "expensive" parlor tricks? If some games and scenes are seemingly more photo real through raw rasterization than a polaroid ever was, is there any benefit to RT cores taking up die space and computing ability? Are shaders and TMU, cache etc of more use? (PS PC gamer since the PC Limited 386-16, VGA 4 ever!)

Sarcastro

Do you think that IDM 2.0 model will give Intel an advantage over AMD or AMD's partnership with TSMC put Intel at disadvantage even more? The 2 main points that I got from Daniel Nenni Semiwiki forums I got are that the costs for TSMC wafers are 20-30% lower and that the "frenemies" concept Intel is trying to do doesn't work in the Fab business. My long-term guess would be that Intel will fall further behind if they don't go full TSMC.

QuickJumper

Former HD 4870 owner here Which market section of PC hardware do you feel needs a shakeup? IMO cases are finally being more thoughtfully designed for both ease of use as well as airflow. And ITX cases aren't as huge anymore. I feel like we're getting there with monitors too. Fans keep getting incremental changes. Storage gets faster and up until recently was getting cheaper. I feel like it's easier than ever to pick up popular parts and get a pretty functional PC going barring current availability issues.

Robojim

Hi Tom Hi Josh, I would like to know what percentage of GPUs sold is the PC gaming market? What percentage of revenue is generated is gaming share overall? Has the PC gaming market shrunk because of mining demand for gpus increasing cost? Also, would you have any idea when Beyond Good and Evil 2 has a chance of coming out?

Hififorlife

My question: why was Intel gaming performance so bad for so long? I heard a rumor once that their drivers were the issue, but it may have a multitude of root causes.

Hey Tom and Josh! Long time PCPer fan here. Given how nuts the PC market is and how depressing things might seem right now, I want to lighten the mood a little. My question for Josh is simple: What's the funniest, absurdest stuff you've seen happen in your storied career as a tech journalist?

Cleansweep

Hi Josh and Tom I know cpu and Gpu have different designs and workloads they're best at but do you think one design (architecture) will eventually eclipse them both? (or perhaps merge and become a hybrid). Making the very concept of a dedicated chip (one for graphics/parrelel work With millions of transistors and one for number crunching with few big cores) totally obsolete I know APUs exist (but they have very separate sections for each job type on silicon) but I'm thinking a truly hybrid design (Maybe an AI designed chip or something)

Alex

In your experience, what was the most impactful generational leap in performance for CPUs and GPUs in the past 20 years?

kjm015

My question is this: now it is constant talk of how expensive pc equipment is and will get more... I remember top of the line (king of generation) amd/ati gpu cost 500$ in 2008. But as kid I remember even for poor pc or 3d gpu, it cost more than kidney, 2000$ or even 4000$ for multimedia pentium with poor Win95. And software on top of that, nothing was free back then. From 90s this 4000$ pc could get to almost scrapmetal value in just 3 years, which doesnt happen nowadays, component value just slowly goes down. By early 2000s, pc parts suddenly were dirt cheap and towards end of 2000s, path to golden age of pc gaming was paved. But other undercurrent is also there: how much hardware is actually improving from enduser perspective, yes lab metrics of transistors and teraflops increases, but so does complexity of architecture and delays/latency between pc parts, not to mention how many layers on top of layers even gpu hardware has, let alone software, driver stacks before game code and OS code. Is this part going better or are companies cheaping out on this? As interesting sidedevelopment of products, yesterday we discussed TI-83 and other graphics calculators, which I think are still around, and carefully took improvements like better battery life, without adding all trendy nonessential bloat. So it was feel of Commodore64 or DOS pc, but with modern tech, essentially how easy and low barrier to write own programs, which nowadays has more software tools to learn and setup on pc. In that sense pc is much more like consoles, something you buy and pray it works, coz hard and wasteful to study problems, to fix them yourself(and latest Win10 updates prove no matter how hard you work, it breaks still "on its own). This as 3rd point/question, how fun of using pc has changed over these decades....

Timo H

Thanks for coming on! What do you think the future of computer performance increases will be? Also do you think Nvidia will give enough VRAM next gen?

Benjamin Cannon

Joshtek! Old PC Perspective fan here... QUESTION: "When is Jim coming back to the show?"

CarryTrainer.Editor