Separate names with a comma.
Discussion in 'AMD' started by EXCellR8, Nov 30, 2017.
I'll believe it when I damn see it.
Interesting, looking forward to see where this goes.
Also, per forum rules, please provide citation in the thread title. FTFY
I actutally think this driver can improve vega performance 20% and probably more in the extreme frontend limited situations where extra performance is mostly needed for gpu mins.
Here is how i think it will go amd style:
Launch 2 days before 2017 ends. So they keep the milestone.
Many small bugs left on table and lots of crashes. And small bug that makes sure ngg is not working properly in many games but will get fixed via a beta upgrade the 6 of january.
Ohh in january and february it will get half the performance uplift because the major revision needs to be adapted to engine to get most of eg the tiled rendering.
Time to also do all the bugs.
But no this will have to launch december so someone can get their bonus.
up to 20% is quite the ambitious if you ask me... makes me wonder how sloppy the current drivers are coded.
The good news about this is its the closest we have to a confirmation the hardware for ngg is working. Else it wouldnt make sense.
Now as for the performance uplift it depens on the situation and how effective the drivers is out the gate. There is a tremendous potential here if we compare theoretical performance vs current practical performance and compare nv vs amd profiles here. For performance and efficiency.
I think we will see a world record performance jump for a driver release. Vega is simply a high freq fury now. When this driver is released its like the card should have been at launch. Like the 7970 out the gate. Still lots of potential but at least using the basic functionality of the arch.
Just because code is not as fast as it could be doesn't mean that it wasn't written well. It could be slow because it was originally written for older architectures, or maybe to prioritise small code size instead of speed. Optimizing code tends to make it less maintainable too, so it's not something you want to do until you have determined which parts really need it.
I'm all for post-launch performance gains, and if you ask me AMD has done a good job with Vega thus far, so hopefully owners continue to see improvements.
I'm actually quite happy with the 64, despite a few little caveats.
Oh yes, the thing can't come soon enough.
Gonna be very silly either way.
AMD has the Comic Book Movie Adaptation Style Cinematic down pat. Here's hoping for some decent performance upticks! The VEGA launch has been really rough, and this would at least be a move in the right direction towards providing some return for the faith.
When I put my tinfoil hat on I hypothesize that board partner cards won't be released until after this driver update if it has any significant uptick in performance.
This will give initial reviews much more favorable stacking vs Nvidia cards of the same price bracket, which is hugely important for people Googling info on cards.
Initial reviews can permanently stain your card's reputation, so why release one before AMD improves the architecture's performance via drivers?
.... ok.... I'm putting the crack pipe down now.
That's not the case since several reviewers have ready posted their Red Devil reviews.
Yeah, it's not a good hypothesis!
Good part is that performance uplift will be very, very real.
You don't see a company launch unfinished product every year.
unable to find any additional information on these magic drivers, so sorry for not updating the post.
sources seem to hint at the week of Dec 17-23... which could mean basically anything or that nobody knows a concrete date.
I have a feeling if they were truly amazing AMD would already be presenting shiny graphworks powerpoint slides.
But you never know.
That wasn't the case for either TR or RR.
TR didn't "exist" and RR performance uplift was basically slashed in half on initial estimations.
Wondering if the hnm2 pesudo channel can be enabled or used or its broken in hardware. They need the extra efficiency for the ngg even if properly implemented tiled rendering can ease the load.
But they don't.
NGG is about not going off-chip.
LDS is basically a free launch.
I can't say the exact number, or they will kick my rear-end.
Sooner than you think.
Will someone from RTG publish a detailed whitepaper on NGG?
The thing is too interesting to be left in "just werks" (or not werks at all) status.
I don't think that they will release a more detailed whitepaper.
This totally smells like nVidia's "sikrit sauce" approach and boy do I not like it.
One reason that AMD hasn't been able to leverage all of their GPU tech is because of their "build it and let the devs decide to use it" approach. I think they are making the right call to profile games(this is how I assume they are implementing it) so consumers get the speed benefits. What I expect AMD to do is to open up the implementation like they have with everything else. A flag devs can use to turn off AMD's profile when the game/engine will handle the implementation. Best of both worlds.
So we must be getting Adrenalin this week or next. Probably next given AMD likes early week releases.