The red spiritAnd here I go bust. I don't know. I just know that flops definitely matter, due to basically all geometry calculations being floating point. I said that there's could be limitations in achieving maximum theoretical floating point performance. From spec sheet, RX 5700 XT definitely looks overall worse, but here's one thing that is better on it and it's pixel fillrate. It seems that hardware in that one aspect on RX 5700 XT is just better and maybe that one specification matters.

Here's some snack:
www.tomshardware.com/reviews/3d-benchmarking,205-3.html

It seems that Vega 64 may perform better than RX 5700 XT at low resolutions, but maybe not.


Any technical reason why?

TheoneandonlyMrKBecause flops do not align well with FPS ,why else?!.

In flops did a 1080ti beat a vega64?! No but in FPS it did.

The red spiritAnd here I go bust. I don't know. I just know that flops definitely matter, due to basically all geometry calculations being floating point. I said that there's could be limitations in achieving maximum theoretical floating point performance. From spec sheet, RX 5700 XT definitely looks overall worse, but here's one thing that is better on it and it's pixel fillrate. It seems that hardware in that one aspect on RX 5700 XT is just better and maybe that one specification matters.

Here's some snack:
www.tomshardware.com/reviews/3d-benchmarking,205-3.html

It seems that Vega 64 may perform better than RX 5700 XT at low resolutions, but maybe not.


Any technical reason why?

ColddeckedDING DING DING.

With newer architectures it becomes easier to achieve maximum theoretical performance.

Get it?

64KThe 5700 is a little faster Vega 64

The red spiritI already knew that, that's why Kepler aged poorly. But I think that pixel fillrate advantage still matters.

ColddeckedStop moving the goal posts, even if pixel fill rate matters, it just goes to show you TFlops isn't everything when it comes to gaming performance which was your original point. Just admit you don't know, its ok to be wrong about things and learn...

The red spiritI see. You still haven't noticed maximum theoretical floating point performance

TheoneandonlyMrKDucks that got to do with actual FPS performance.

Nothing.

The red spiritWhat does then?

TheoneandonlyMrKReviews.

Testing.

The red spiritI'm talking about graphics card, what inside of it does matter? You say that floating point performance doesn't matter.

The red spiritI don't mind admitting something, but there's nothing to learn here. You all just keep saying that this or that doesn't matter, but what does matter? I'm pretty sure that flops, pixel fillrate and texture fillrate do indeed matter at different stages of 3D rendering pipeline. I don't know exactly where, but I'm open to learn that. Sadly, nobody talks about that here.

TheoneandonlyMrKIt's not totally irrelevant, but Rops, Tmus Raycast units, clock frequency, ACE engine's the cache structure and amount and other special function hardware, and the attached memory frequency and bandwidth all have a part to play in how much FPS, whereas flops uses a few but far from all of those functions.

ColddeckedLook, of course TF matter, but its just a measure of one aspect of performance.

ColddeckedAnother problem is that you think TF are all equal.

ColddeckedThey aren't when it comes to actual game performance. That is why you can't compare AMD and Nvidia cards strictly based on TFlop since Maxwell, because Nvidia uses their TF better (more efficiency tricks) than AMD did, and were able to get better game performance even though their their TFlop number isn't equal.

ColddeckedAs MrK is pointing out, Vega 64 had more TF than 1080ti, but the 1080ti is far and away the better gaming GPU. Just don't get so hung up on the TFlop numbers, it'll give you a rough idea but there's more to the equation.

The red spiritROPs, TMUs probably influence FLOPs. Ray Tracing Units, don't matter, unless you use them. Clock speed is surely directly related to FLP output. ACE is just AMD's version of SMX, which is CU ( Compute Unit). CU is cores with controlling logic, but without cache, ROPs, TMUs, encoders, decoders and etc. Memory bandwidth is likely heavily related to FLOP output too.



It's general performance evaluation. It takes into account clock speed, IPC, core count and likely memory speed. That's pretty much whole card.


They are are, floating point operation per second isn't a fuzzy metric.


But why?



Well that's a general consensus, but I wonder why exactly.

TheoneandonlyMrKAlso regurgitating the same shit I just said back at me slightly fleshed out is you agreeing with me flops are not the big picture?! It Sooo?!.

The red spiritI don't care about agreeing or not. I only care about actually learning how each graphics card specification influences the way graphics card works.

TheoneandonlyMrKFunny because this started because you Know Steam deck is useless for gaming on ffs.

You don't learn by shitposting in tangential threads you have no interest in, yet spout crap like your an insider who somehow Knows how it performs before reviews?!.

The red spiritEither way, it will not run Cyberpunk or Valhalla very well. You people are expecting miracles from thin client hardware. Meanwhile, I have been interested for a long time in low end hardware, used some of it and observed what it can do. Deck won't be a miracle today. Cyberpunk will not run at average of 40 fps and over time, Deck will certainly struggle in more and more games as time goes on. I personally don't like buying hardware that is only scrapping by today, as it feels disappointing soon and makes TCO high. Deck doesn't do anything to make me think otherwise. lol, you already have Zen 2 chip. Why don't you try disabling cores and setting it at Deck's PPT and clock speeds, I wonder how fun will that be in games. You can also downclock your graphics card, to loosely match Vega 11. Now tell me how well everything runs.

Durvelle27You can't spouting crap but I showed you the GPD win3 playing Cyberpunk just fine on med settings with above 30FPS while it has a inferior GPU. Just stop at this point dude

The red spiritIt was running it at low and with enabled function to lower resolution to maintain fps. And 30 fps average is unplayable imo with that framerate variation.

Durvelle27You keep spouting crap but I showed you the GPD win3 playing Cyberpunk just fine on med settings with above 30FPS while it has a inferior GPU. Just stop at this point dude

Durvelle27I'm not going to argue as you've shown you have no clue as in to what you are talking about

ThrashZoneHi,
Yeah 30fps should be fine on a device this small.

Way too small for me though.

Durvelle27Just check this out. This is the GPDwin3 shown playing various current games at decent settings with 30-60FPS @720P. It's pretty stellar especially to have inferior specs