AMD Radeon RDNA2 Navi (RX 6500, 6600, 6700, 6800, 6900 XT)

Doesn't look very "killer".

 

It performs between Nvidia's 128 and 192-bit GPUs, is supposedly cheaper, and seems to suck at mining- I think this is great.

 

Two points go against this statement: 7nm dies are more expensive than comparable 8nm dies, also the RX 6600M requires a separate board design from 6700M, most of the RTX 30 GPUs fit into a single design.

 

Early reports are that the 6800M is being priced signficantly below the 3080, and in terms of die size the 6600M is in a similar position vs the 3060. Also, who says OEMs have to pick up the 6700M? Maybe they will just offer the 6600M and 6800M, since AMD don't seem to be strongly pushing the 6700M.

 

Only if you ignore e-sports titles. The difference is 3 % if you average all gaming results.

 

I think we only have one data point for pricing so far. However...

"Apparently AMD is also gunning hard for market positioning with a big price advantage. The Asus ROG Strix G15 AMD Advantage Edition we tested for this review is priced between $1550 and $1700 depending on the configuration. That’s at least $500 in savings than the most affordable RTX 3080 Laptop available at Newegg, Best Buy or Microcenter, a Gigabyte Aorus model which doesn’t even feature the 135W power configuration."

https://www.techspot.com/review/2268-amd-radeon-6800m/

 

Averages should be normalized, IMHO.

 

That's sad for AMD. I was expecting RDNA2 to finally beat Nvidia in power efficiency this round, but no, it fails short. One full node advantage with the best world's foundry was not enough. All of that without important features like matrix/tensor units, DLSS, top class RT hardware/performance and it's totally trashed in FP32 compute applications... This generation goes easily to green team

Next will be interesting. RDNA3 on TSMC 5nm and... I just got news of Ampere Next on Samsung 4LPP (custom node for NV, basically an improved and tuned 5nm). On paper, the difference with TSMC is expected to be very minimal with plenty of allocation available for Nvidia. ..

 

Oh nope, they're using funny fmaxed parts here.

It sure did.
AMD just considers all kinds of silly mQ things utter perversion.

Bro that's on-roadmap.

Nope.
x07 parts go to meet their maker, everything else goes to NV33.
And it's gonna be oceans of delicious cope and seethe...

 

Yeah it didn't.

 

It sure did.
We can always powerlock GA106 and NV23 both to 60W and watch the funny escalate.
Your seethe is both unwarranted and unnecessary; you always have other threads to take a dump in.

 

Can you elaborate on this?

 

As you've noticed none of the currently present 6k M's are anything low watts for their die size because if you can get more perf out of some area then why not?
Better, 6600M is even a chop versus 3060 Laptop being full 30SMs.
Adorable.

RMB perf scaling will extend pretty far too, actually.

 

RDNA2 was a dud, RDNA3 might not be.

 

Shiiiiiiiiiieeeeeeeee amazing.
If that's a dud then wtf is Amperage.

The what.
I mean, prepare 2.5 kilodollars, yes.

 

Numbers please to prove your point. What we see in games is equivalent in 1080P then beaten above.
And we are in 2021, GPUs are not anymore only for pure rasterization. Power efficiency with RT goes easily to Nvidia.
Finally, compute (another big usage for these beefy laptops) is not even the same league. You need nearly twice the power for Radeon to compete with Ampere dual pumped FP32...
So no, RDNA2 is still far behind Ampere in power efficiency when we take into account all usages. Period

 

Wait till someone anglo gets their sleazy hands on a new Omen and locks the power down.

They are in laptops.

Come on this is just silly to even mention.

The what.
They're gaming laptops chief.
With HRR screens.
Nothing besides maximum FPS pew-pew matters.

HELL YEAH THE ULTIMATE IN GOALPOST SHIFTING.
Should I dig Pascal-era shit or nah?

 

Your not making much sense. Anyway, im atleast positive for RDNA3, why would it be so much more expensive you think (if thats what you mean).

RDNA2 is behind in RT, pure raw compute power, reconstruction tech and other advanced features. BUT, its their first iteration after in the this segment after a long break since 2013 (?).
RDNA3 will be much more competitive in all regards.

 

Core IP stuff.

Lol.
Come on.

Because many many tiles times SoIC bond means expensive.
And niche.

Yeah stuff that doesn't matter.

Nope, it's made to push many many frames per second and not much else.
No ML bullshit or ALU spam for you.
We've been all over this before.
You can stop now; nothing changes AMD phone/laptop first client GPU IP roadmaps.

 

We don't know that yet.
One thing I wonder about lately is that +50% perf/watt with 2.5X performance would in fact result in higher power consumption in the high end (Navi 31).
And this is usually a sign of there being something competitive which prompts such power increase (see Ampere). Cause otherwise why go out of PCIE spec?