Home
Creators
Search Recent Missing Random
Posts
Search DMs Popular Hash Lookup Tags Random
Importer
Import FAQ
Register Favorites Keys Review DMs Logout
Community
Telegram ThePornDude
Home Creators Posts Import Register Favorites Logout

Discussing RADEON, Nvidia, and ARC with Marc Sauter (Patreon)

Published:
2022-09-13 20:34:41
Imported:
2026-06
Tags:

Content

Broken Silicon's next guest will be Marc Sauter of golem.de!  We plan to discuss:

Previous Nvidia & RADEON releases 

RADEON vs Nvidia standing in the market

Expectations for Lovelace and RDNA 3

Intel Alchemist

Intel's standing in the market

This episode will get into heavy talk about what Nvidia and RADEON did well the past, what they did wrong in the past, and what we expect out of GPU Competition in the future.

Of course you are all invited to ask any questions you want, but we are generally planning to have this be a mostly GPU discussion. Don't worry, I've got a fun guest lined up for the CPU releases in 2 weeks!

Additionally, we plan to release this episode early to leave some breathing room for the Lovelace coverage next week.  You have ~36 hours to submit reader mails (Until the Evening of 9/14/22).  Be concise (seriously guys lol), use good grammar, and be thoughtful for consideration :).

Comments

SuperCrepen

What specifically do you think AMD/RADEON should do besides increasing performance to increase its mindshare over NVIDIA/GeForce?

Swiggles

Hey Tom and marc, we have continued to see the importance and pressure AMD has put on competitors to make markets better for consumers however indirectly. How much longer do you see them continue to keep playing the "nice guy" and start really start acting like their, larger counterparts in CPU and GPU.

AI Generated Fish

I have a mostly laptop-centric question. Although AMD hasn't had the best luck gaining desktop market share, they seem to be doing better there than in laptop. Even though Nvidia seems to be pushing power on desktop, the node shrink should be great for laptop and they have dominant mindshare. Intel has already released laptop Alchemist models and that may be their only segment to release discrete GPU's. How do you see the laptop market shaping up this generation? Do you see AMD and/or Intel eating into Nvidia's market dominance?

Roman Matúš Labuš (edited)

Comment edits

2026-09-17 17:00 Hey Tom and Marc. I think one of the stupidest things that Nvidia has done was not anticipating eventual mining crash and severely overproducing cards from Ampere lineup. It just shows how much Nvidia wants to satisfy its investors. Also AMD I think has very smart strategy on strongly focusing on efficiency. It shows that AMD can think strategically several years into the future while Nvidia's leadership seems to be bit shortsighted. BTW I don't have a preference for any company, I just think AMD's strategy right now is superior.
2022-09-13 20:56

Joachim Haugen

Overproducing was the most brilliant move. For Nvidia the worst thing would be a 3. competitor By taking a hit in the short run, They literally erally killed Intel's ark entry.

Capelio

RDNA3 will bring the first consumer GPUs built with chiplets (5nm in this case). What are your predictions for what we’ll learn about how chiplets and their required interconnects will affect GPU performance, positively and/or negatively?

Capelio

Second question if I may: similar to special purpose ray tracing hardware on GPUs, do you see other specialized accelerators coming to GPUs in the future?

Joachim Haugen

If they have the efficiency advantage. Make the best laptop's. Bundle their best CPU's with their GPUs. Mandate the best cooling, ram and ssd. Then push the watt's as high as possible. Only send thrash can CPU's to laptop's that are bundled with Nvidia

Cleansweep

Hey Tom and Marc, With Intel Arc all but officially confirmed to be cancelled, what do you think it would take for Intel to try to get into discrete consumer GPUs a third time?

Joachim Haugen

I think Tom has said this a good way already "evey time a see Lisa su (not Trotheripper) on stage, se seems to put on more and more leather jacket's".

Precision

Do you know how much thought is going to go into the laptop gpu space this year? Given how Zen 4 seems pretty efficient, you'd hope you can get a decent performing, efficient gpu with it. As a small follow up: Do you think laptop gaming is going to have big gains in the foreseeable future?

Joachim Haugen

Hi Mark an Tom :-) A follow up to Cleansweep's, ark question. Do you guys think we as consumer's should have been drinking the alchemist cool aid, to have a 3. competitor in the GPU space? I meen, someone (like the one guy that wrote in last time and Dan) kept Amd alive, by buying pile-driver And made the world 🌎 a greater place humanity long term... Is this the time, that we all must unite and demand Intel's crap waste of sand gpu's, at a eye watering price, so that maybe one day they can make some good gpu's?

10GHz Netburst

Hey guys, is there anything in the upcoming launches you think is unusual or stands out in a way that is particularly worth mentioning? What expectations do you have that you don't think are being considered enough?

10GHz Netburst

How would you say the German market for consumer GPUs and PCs as a whole is different from the rest of Europe and the rest of the world? I've always found it quite interesting how there are several very big German tech news sites and youtube channels in tech which are known quite well in the English speaking communities, but this isn't so much the case for Spanish or other language outlets.

Dylan K.

When NVIDIA does eventually go multi-die design for their consumer gpu , what would Nvidia do to differentiate them from amd's RNDA3 and future designs? Would they have like a dedicated raster design paired with a dedicated ray tracing die or something along those line? How would they stand out or try to say their multi-die design is superior.

Stephan Hart

Greetings Tom and Marc, I can't help but feel AMD could have done more with RDNA2. It really does seem like the bean counters are perpetually crippling them across the stack. From cutting traces for PCIE lanes on the 6600XT and down, to the 256 bit Bus on the Flagship 6900XT! They just never attempt Bawlz 2 the Wall with their design these days. At least when Radeon was Hot N' Loud they pulled out all the stops. With RDNA3 reportedly only having 1 GCD at only 308mm² (not 370)why not Go as Big as they can with it to crush the competition? F Yields!!! Do you think maybe they had problems with MCM GCD's? Are they holding true MCM in reserve? And lastly why don't they take feature parity with Nvidia seriously? Thank you.

UwUThatOneGuy

Howdy Tom, Moin Servus Marc, on Friday night, Intel allegedly made the hard decision to effectively cancel ARC dGPU's. With many big names in the industry coming out with their own sources, to enforce the understanding, that these decisions were anything but finalized, what really was decided on that Friday? Unlike future ARC gaming cards, if I understand correctly, the larger project AXG was not canceled, and also will not be canceled in the future. The point I'm now trying to make was better elaborated by John Paddie in BS165, but is Intel essentially going back to DG1 strategy now? It seems to me, that it was not only a financial decision, but they had to reconsider what their limits in supporting the software stack were, and they now want to limit the scope and platform they are reaching with future ARC GPUs.  It seems like they got scared off by dipping the toe into a potential launch with the A380 and now have to reevaluate how ready they are to launch anything bigger in quantity and better in quality. If true, that could just be a matter of time, and cut down lineups, until they find their grounds and swing back with bigger launches in a few years time, until they have drivers, hardware, resources and logistics figured out. How much time will it cost Intel to get their stuff together, or will they never catch up to AMD & NVIDIA - was anything in that regard decided on last Friday @Intel? UwU

Benjamin Cannon

1. With how they designed RDNA 2, they didn't need more than a 256bit bus to have enough memory bandwidth 2. They most likely wanted to have a cost-effective die that still has the highest-end performance. There are many technical reasons why they wouldn't do a 600mm2 compute die, the main one being what if they couldn't get it to work at that size? Remember, N31 is INSANELY cheaper, probably around half AD102 BOM-wise. 4. This is what has caused intel a lot of issues and what has made AMD so successful. Rather than going for crazy moonshot ideas, they go for consistent gain gen over gen. Rather than hoping and requiring that they will have multiple compute die(s) working. Maybe they have multi-compute chiplets working with RDNA3, and if they wanted, could have made a multi-compute chiplet GPU but chose the safe option. The one thing we don't know is if they made N31's GDC able to have more than one on a single package. I wouldn't rule that out just yet (But also, don't bet on it) 3. They are taking feature parity with Nvidia seriously, but you can't make everything overnight that your competition has

Benjamin Cannon

Have you heard anything regarding Nvidia's MCM implementations? If they still don't have any MCM ready for the gen after Lovelace vs. RDNA 4, what can they do? Assume they want to, at all costs, margins are damned, arguably hold the gaming crown

Ryzza5

With 2nd hand markets about to be flooded, which 2nd hand GPU(s) might you recommend for an always-on HTPC that is mostly a Plex/Web/Email server but also is occasionally used for couch co-op gaming? It currently has an AMD 3600, 32GB RAM and a Nvidia 1060 3GB connected to an older 4K60 TV. Idle power usage is an important consideration, and I likely wouldn't spend more than $150-$200. I hope these constraints make for an interesting discussion.

CompressedAIBlocks

Hello Tom and Marc, hope everything is well. My question is where do you think AMD and Nvidia will be after RDNA 3 and Lovelace? We have a rough idea of performance, die size, BOM and such but how do you guys think these companies will walk away from this upcoming gen? My thoughts are even if AMD does not completely stomp Nvidia they will come out much better having done MCM, consistent software stack updates and really just pushing Nvidia into a corner design wise. Nvidia seems a little boned if it's monolithic again after Lovelace if this is the case.

KaiserAullirio

Hi Marc and Tom, How much do you think AMD will close the RT performance gap this coming generation? Have you heard any rumblings on this? Besides final prices, this feels like the last thing that hasn't leaked yet.

Resynk

Hi Marc and Tom, Thank you for your time on the show Marc and your consistent passion for truth rather than clickbait Tom. How much is VRS, Mesh shading, Sampler Feedback and DLSS/FSR shaping the newer architecture in AMD/Nvidia Cards? With a tighter integration with components (eg, GPU to NVMe with DS, CPU to GPU memory with SAM and Resizable Bar, etc) and RAM getting faster, how does this also shape GPU architectural changes and bandwidth requirements across different buses?

Timo H

Hi Tom and Marc, could you comment what next gen gpus will do in gpgpu datacenter market, as understanding is that is primary target and gaming gpus are secondary like in cpus. In context that advances and taken marketshare in datacenter often reflect features or uarch design or even pricing changes in gaming gpus.

Joachim Haugen

Hei Tom and Marc, if Intel ark is done. Can Intel get data center GPU going, and over time use this to launch consumer's GPUs again?

slushba

Hi Tom, Hi Marc. As you mentioned in a couple of videos that one of the main goals for amd this gpu generation is to get marketshare. What are the things amd has to do to overcome nvidia mindshare to gain marketshare in order to not have a similar scenario like in the days of the 4000/5000/6000series cards where they got some marketshare, but were not able to really rattle the mindshare cage.

slushba

And if a second question is allowed i would also like to ask what you think why nvidia chose to go monolithic this gen and apperently the one after that instead of a chiplet design? Is it more of a "that's the thing we're good at and people will still buy it no matter what" or do you think they might have same/similar problems as intel in regards to make chiplets work? And if so, is amd really that far ahead in chiplet game?

Jen-Hsun Huang, AI is not circular

Why is it that reviewers refuse to only utilize raytracing games at max settings when comparing RTX with the competition?

cockblockula

Do you guys think Raja Koduri has had a harder with Intel Arc than he had with AMD's projects like Radeon & Ryzen? Considering Intel's vast wealth and resources, I assumed Raja would be much more successful at Intel.

Chris Rijk

If my current GPU died at the end of the year, I'd probably try to replace it with something about as fast but cheaper and/or lower power. What would you do in such a scenario? Do we need more advanced games more than we need faster GPUs?

Chris Rijk

So far it feels to me that Nvidia only feels the need to "respond" to AMD late in the product design cycle, eg pushing clock rates and power more than originally planned or like what they did with the 3050 against the 6500. Do you think that's enough or does Nvidia need to change how they compete with AMD early in the design cycle and change their approach to the silicon design itself? Or to take a different angle, can Nvidia continue to enjoy Apple like margins without creating a self-contained ecosystem like Apple has? (Please don't read out the following: to help explain what I mean, when you're using Apple products you're using Mac OS or iOS and switching to Windows or Android or whatever is a much bigger change than switching between Nvidia GPUs and AMD GPUs, at least if you're a gamer. Nvidia is trying to create their own ecosystem by having people depend upon their hardware and APIs, but is that enough of the market? Nvidia can't then remove features from their API and needs to keep their products feature rich. But that makes optimising the silicon design much harder)