- Joined
- Sep 17, 2014
- Messages
- 21,798 (6.00/day)
- Location
- The Washing Machine
Processor | 7800X3D |
---|---|
Motherboard | MSI MAG Mortar b650m wifi |
Cooling | Thermalright Peerless Assassin |
Memory | 32GB Corsair Vengeance 30CL6000 |
Video Card(s) | ASRock RX7900XT Phantom Gaming |
Storage | Lexar NM790 4TB + Samsung 850 EVO 1TB + Samsung 980 1TB + Crucial BX100 250GB |
Display(s) | Gigabyte G34QWC (3440x1440) |
Case | Lian Li A3 mATX White |
Audio Device(s) | Harman Kardon AVR137 + 2.1 |
Power Supply | EVGA Supernova G2 750W |
Mouse | XTRFY M42 |
Keyboard | Lenovo Thinkpad Trackpoint II |
Software | W11 IoT Enterprise LTSC |
No, it's you people who babble how GCN is all the same from HD7000 up to Vega that piss me off. And now all of a sudden my absence of owning an AMD is now a "problem" because I'm supposedly not knowledgeable enough about GCN even though I freaking owned the very first generation of it on excellent HD7000 series. Weren't you among the idiots who were yelling at me for being an AMD fanboy coz I had a fucking Radeon 2 years ago before GeForce, but now, all of a sudden, it's an issue coz I supposedly don't know anything about it coz I don't have it now. Consistency, you may want to look that up in dictionary...
Also, that 4 wheels and steering wheel car analogy. You may want to study that up as well...
Come on man, I've also been putting that argument in and Vega's current state of affairs shows exactly what this radically changed (sarcasm, I hope you can see) GCN does for its efficiency (zero) its clockability (no real change from Polaris) and its real life performance (all over the place from below GTX 1070 to 10-15% over 1080). Essentially, the practical results of this new iteration of GCN are exactly as they were before and even WITH the die shrink, it has lost efficiency compared to Fury X because AMD had to push heavy on the power budget to achieve reasonable clocks.
If you take a few steps back you can see comparable things happening with Intel's Core arch and AMD's GCN - they're both architectures that are closing in on their EOL date, and are extended by adding more juice (or cores!) to achieve higher absolute performance. Nvidia does not follow that line - every iteration of CUDA has tangible improvements, and every consumer GPU from Kepler onwards pushed an efficiency jump by tearing down every barrier to enable ever higher clocks at the expense of versatility. From a 100-150mhz gap at the Kepler - Kepler Refresh (already!) this has grown to 200-450 (!) mhz gaps in Pascal compared to GCN. That's a LOT of time for a competitor to follow suit and make wise decisions - and Vega is the result of not doing just that. Instead, AMD markets a 'prosumer' Vega as a jack-of-all-trades, it couldn't have been more ironic.
I remember you harping on about tiled rasterization and all this other new stuff, but it simply does not translate to performance or efficiency at all. That goes a long way to prove RTG's skill at redesigning something aren't really very good, and it supports what most have been saying that the RTG press slides were a bit too optimistic, as always. Its the good old pitfall of over theorizing something that is about to happen but never really does happen; in the same vein, look at where that awesome DX12 GCN performance is going compared to Pascal - again we see higher CPU overhead in the driver... Again it underlines that GCN is still a lumbering giant and not a lean machine - and that's 2012 knowledge; five full years of a practical standstill.
Last edited: