AMD: SIMD vs VLIW

[@bacarddi92]

I really hope you can reach more people soon, I really like your video's and learn quite allot (*a insane amount to be price haha) Keep up the good work mate, and lovely video.

[@-yttrium-1187]

Instruction Tetris. Now that's what I want to play!

[@Alan_Skywalker]

Probably the best video about vliw.

[@rheumakai86]

Liked your video. Now I am even more curious about SUPER-SIMD. Hope there will be a follow up on this :)

[@RadoVod]

A natural follow up to this video would be a super-simd walkthrough. Because AMD sure seems to like tetris.

[@Megalomaniakaal]

Tetris is the best selling game of all time, whats not to like? :)

[@dawienel1142]

Great video man, a bit over my head, but my head needs to get bigger. God Bless.

[@TheEVEInspiration]

I love the term "Instruction Tetris" :)

[@dariob467]

i still have my hd 4870, after this video, i think it is a piece of hirtory, it was the first graphics card with GDDR5 memory!

[@mr_beezlebub3985]

Hello, and thank you for making such a detailed video

[@slyblood85]

Now I'm just wondering what SuperSIMD is, and wondering what type would be best for a raytracing/pathtracing?

[@nkonfya]

this is awesome

[@TheImpartialTruth55]

Thank you for your work.

[@jamzqool]

anything for the vliw2 video?

[@christose9247]

brilliant.

[@karanvora2674]

Navi is a transition from gcn to super simd. Super simd will give a huge boost to full and double precision type loads but gaming won't be beneficial from that unless we bring some AI algorithms to render the entire games just like dlss which render pixels and superscale it to higher resolution

[@3800S1]

Do you know anything about the upcoming AMD architecture? I would be interested to know what changes they might be implementing. Anyway, thanks for the great video. Top job!

[@bassbatterer]

Regarding Navi, I don't know anything. However I was planning a video about 7nm Vega as I feel that really could demonstrate "Poor Volta"

[@3800S1]

@@bassbatterer Far as I heard Navi is supposedly some hybrid form of VLIW and GCN, but it's only rumors. As far as poor volta, I am guessing you are referring to vega's compute power? if so I certainly see what AMD meant.

[@bassbatterer]

@@3800S1 If Navi is using a VLIW SIMD like instruction set, then that would be the Super SIMD, quite interesting. And about Vega it is the impressive compute the thing has, the only issue is its competing with Volta which isn't exactly brand new and may be up for replacement soonish.

[@3800S1]

@@bassbatterer Cool, I wonder what volta will be replaced with? Anyway regarding an earlier video how GCN uses hardware scheduling. It seems people believe that AMD's driver have more CPU overhead which their methods of testing is resolution scaling where a slower nvidia card can overtake in low res but the faster AMD card is faster in med to high res. Eg RX 570 vs GTX 1050. I think this is flawed method of testing to conclude this. Any thoughts? I figured actual CPU usage would be a better indicator but people I have mentioned this to insist that res scaling is conclusive of this.

[@bassbatterer]

@@3800S1 3800Tech yes I defiantly have some thoughts on the driver overhead debate, I Saw some videos and found flaws in testing and I am defiantly going to be making a video about this and be doing my own testing to find Driver overhead. As for why AMD seems to increase performance relative to Nvidia on Higher res, I also have some thoughts. I've been looking at this and related stuff for quite some time now but I wouldn't be talking about it any time soon as its a very complex subject but I have been gathering data on this and will continue to investigate. If you're interested, here is a spreadsheet i have made a public copy of which contains some of my work looking at AMD v Nvidia cards at resolutions among others: docs.google.com/spreadsheets/d/e/2PACX-1vRKhLPBsKyEMLMrG4e75_a1FqnU_fZZqIVbKxG1FkP483T-duqVJUQwet2O440lUOfVTggUBCv5w-Ut/pubhtml

[@giserson2]

This explains why AMD cut driver support so early for Terascale GPUs.

[@Noman1000]

so would the large "SIMD pipeline" be why AMD cards are worse than Nvidia for game/video rendering even though they have higher FLOP output per sq/mm of die size?

[@bassbatterer]

It's unlikley, Nvidia use a similar SIMD style processor with their Turing architecture actually getting closer to AMD's GCN in terms of design, i did a video on that recently you could check out.

[@darrenraphael]

@@bassbatterer GCN is a straight copy of Nvdia's 2009 Fermi Architecture with some tweaks.

[@hououss5641]

Brainstorming 👌👏

[@TrueMathSquare]

Is there any way I can donate to you?

[@bassbatterer]

Thanks for asking, but i don't take donations, i do this all of my own accord because it's what i enjoy and that's what makes me want to do these vidoes, i'm not interested in donations but thanks anyway :P

[@TrueMathSquare]

Why wouldn't someone take money for doing what they are enjoying....

[@bassbatterer]

@@TrueMathSquare Well I don't do this so strictly, sometimes i don't have time to work on this kind of thing nad can't make regular uploads, so whilst i'm happy to do these, i wouldn't want to take any money for these. For me it's just nice to work on.

[@srivastavaabhinav1200]

very bad voice quality