Bondrewd
Veteran
Yeah he did.
The implications are pretty clear; they're selling you value (and I don't blame them).
Speculation: GPU Performance Comparisons of 2020 *Spawn*
- Thread starter eastmen
- Start date
- Status
I think that's getting into semantics. This was Nvidia's model running on Nvidia's hardware. Microsoft weren't demonstrating their own version of either a resolution upscaling model, or the hardware it runs on. They were simply demonstrating how anyone else that wants to develop those things can run them through Microsofts DirectML interface rather than through their own proprietary interface/API as Nvidia currently does for DLSS.
Put another way, this was a demonstration of DLSS running on a vendor agnostic API, nothing more.
Then what was Microsoft doing there..? The exact point was... that MS was looking to using directML for things such as dlss, 2 years ago.
The were demonstrating a vendor agnostic API that developers can use to implement machine learning based solutions on any GPU that supports DML. The machine learning based solution they used to demonstrate that capability was DLSS.
neckthrough
Regular
What @pjbliverpool said is correct. There's a huge difference between an API and an algorithm. DirectML is simply an API to access vendor-specific hardware capabilities in a vendor-agnostic fashion. You can implement a variety of ML algorithms using that API, including (as a specific example) DLSS's real-time inference component.
The fact that Microsoft has been working on the DirectML API says nothing about whether they have been working on a DLSS-like training infrastructure + inference algorithm that uses that API to perform upscaling. They may very well be, but DirectML is not an evidence of that.
The fact that Microsoft has been working on the DirectML API says nothing about whether they have been working on a DLSS-like training infrastructure + inference algorithm that uses that API to perform upscaling. They may very well be, but DirectML is not an evidence of that.
Could you tone it down and keep your cool? There’s no need to throw around accusations of gaslighting. We’re all here because we’re PC hobbyists and we enjoy talking about GPU tech.
You posted two links to reviews, neither of which disprove what LordEC911 said, nor do they prove what you’ve conjectured about G92 or Maxwell pricing.
It would be nice to have a bit more information on the Series X Backwards Compatibility upgrades, as that might show Microsoft has done a few things with DirectML. I'm fairly positive their BC HDR upgrades for even games as far back as Fusion Frenzy (Original Xbox) is using DirectML of sorts.
D
Deleted member 2197
Guest
Bondrewd
Veteran
The answer is "yes".
A big, fat, juicy and resounding one at that.
It's kind of amazing that so many tech youtubers have such a loose understanding of the tech and industries that they cover.
The not-so-subtle dig at Raja is pretty childish and unnecessary.
Beat what, and at what cost? RTX 3070 at similar or higher power consumption? Not interested.
RTX 3080 or the RTX 3090? Now you have my attention.
Bondrewd
Veteran
They exist to entertain, not provide technical insight.
That's N22 territory and at lower power.
Yeah.
They really want to; dropping the mic and all.
Leoneazzurro5
Regular
If they have such a great card, then better to drop some announcement soon, before the availability of the Nvidia cards, otherwise, judging by the forums, a lot of sales will go to theeir rival already. If the news about limited NV cards availability until January are not true, of course.
Bondrewd
Veteran
Yeah.
P sure teaser soon(tm).
They didn't even guide any dGP movement FY.
I.e. they don't give a shit.
Leoneazzurro5
Regular
What is this supposed to mean? They don't give a shit to lose additional market share to the competition? Quite strange for a company spending money to develop product that must be sold for recovering the investment.
Last edited:
Bondrewd
Veteran
They're not planning to print any money by selling discrete GPUs this year.
Yeah, CPU growth eats wafers and offsets any dGPU sales by a huge, humongous landslide.
They're doing it for shits, maybe giggles.
Same as their specialty ULV APU and other stuff like TR (now two of them!).
I'm pretty sure any reasonable buyer takes a wait and see approach when you have 2 next gen launches that are so close. People jumping on pre-orders based on Nvidia marketing slides wouldn't buy AMD GPU if AMD was 100% faster at half the power.
Bondrewd
Veteran
True. And yes, considering how unimportant dGPU is to AMDs bottom line, there is no way in hell it take away wafers from much more lucrative CPUs. Mobile GPUs potentially being an exception,
- Status