Originally Posted by ZealotKi11er
I also bealive they will have a Vega and 590 card.
To believe belongs in a church, not a tech discussion
Besides that it has been already debunked. AMD themselves said 500-series is for Polaris only and Vega will be names Vega - a complete new brand. AMD themselves said that at their CAPSAICIN event.
---My personal guess for the naming scheme is---
Vega Pro / Vega X and just Vega (like the Fury / Fury X brand) and something else maybe. I don't know if they use the NANO label again. There is still a possibility, seeing the smaller Vega in mobile devices only.
But the 490/590 name was speculated to be the dual-Polaris card. Yet we see the inefficiency of the P10/11 and can assume it is not going to go well with a planned dual-card. AMD might have dropped it, or it didn't exist in the first place.
Originally Posted by H1vemind
Polaris is designed to run at a certain frequency with the improved switching speed on the new process contributing to the final clocks, GCN1-4 on 28nm was designed for about 950mhz base/1200mhz OC (the OC average managed on 28nm at reasonable volts towards the end) + whatever the process can do Vs the TSMC 28nm Amd was using last, 16FF+ has better switching speeds (which means better clocks) than both TSMC 28nm and 14nm LPP and scales better past the Frequency/Voltage sweetspot meaning that yes P10 could have had much more common 1500mhz OC's early had they switched to TSMC or even Samsung since the improved node maturity would have meant we got way more of what is currently considered to be golden samples and far fewer duds.
When a bad chip design meets a bad to medicre process node, it is the worst case scenario. Nvidia had that scenario back with Fermi. AMD has it right now with Polaris. They may not have a high-end chip, but maybe that is the reason. If you see how much Polaris consumes and how well the cooling is needed, a higher end cards would be out of the scale.
Nvidia has made TSMC fire under their butt when the Fermi fiasco happend. The foundry is now top-notch in their nodes. But Pascal has been reworked to the finest detail too. There is a whole new voltage- and powercontrol system in there, which Maxwell didn't have before. Togther those two advancements allow Pascal cards to have high clocks with low voltage, although it is coming with the envelope of a tight power target limit.
AMD has not done anything like this and left many negative sides of GCN in place with Polaris. Those chips wouldn't clock higher even with a better process node like TSMC has them at the moment. It's almost a design flaw, if you ask me.
AMD is so dumb, why didn't they lower the voltage, bring faster memory and lower the TDP? Are they so desperate to not sell new cards? I went through various forums and everyone expect the fanboys is so sad about AMD's latest shortcomings.
Originally Posted by H1vemind
Your right about memory overclocks being better for P10 but its amd's choice to make not ours sadly, The difference between Fiji 28nm and P10 14nm is the fact that 28nm was a very mature HP node and could better handle going past the V/Mhz sweetspot and 14nm is a low power process that hates more Volts by design.
Alot of points can't be t explained by maturity alone. AMD had even problem in the old and mature 28nm node. Many Radeons were thirsty for power and liked it warmer. HBM did not change much of that, just ease the symptoms. Especially Fiji was not ripe as an architecture. Lots of areas were just not balanced enough - same goes for Polaris.
I don't think the same applies for Vega as of yet. Personally I hope this not to be the case.Edited by Hardware Hoshi - 3/31/17 at 11:17am