Tom's Hardware of Finland

Yeah, raytracing stuff is being added to vulkan and dx12

Developers can access NVIDIA RTX ray tracing through the NVIDIA OptiX application programming interface, through Microsoft’s DirectX Raytracing API (DXR) and, soon, Vulkan

1 Like

yeah, you don’t have to buy them brand new, and the improvement over Pascal should be somewhat modest otherwise (like 30% per watt), but I don’t think anyone’s making the argument that they’re a good value, more that they’re good flagship GPUs, which will in turn improve mass market products (consoles) and the prices of older GPUs

There is a rather dubious demo video of the ray tracing features in Battlefield V.

It does look reasonably impressive. However, they compare it to purposefully neutered other methods to try and do the same/similar effects. Like at one point, they shown us PS2 level cube map reflections on some windows and try to say that’s the best you could do before Nvidia RTX…

oh yeah they always pull that stuff

I’ll be happy with a Rise of the Robots reboot that matches the original promo material and so should everyone else

1 Like

I also just noticed they went to another 8 pin on the Ti which means it shouldn’t be that hard to get >20 fp32 tflops out of it

I assume you mean an 8 pin power connector

What does that have to do with fp32 flops? Rather, why would that imply a certain amount of fp32 flops

because GPU compute power still scales pretty much linearly with thermal design up until you can no longer dissipate heat because CUDA cores are infinitely parallelizable, and between them emphasizing that the stock configuration of Turing is much quieter than Pascal and the 8-pin power connector providing up to another 75w over Pascal’s ~275w overclocked limit, things should get pretty interesting at the high end

and fp32 flops are more or less the most direct measurement of regular old rendering power that there is

They are also emphasizing overclocking and These are said to clock high. So the extra power pins make sense. Even the reference founders editions are factory overclocked.

Are they really over clocked then? What do “stock” clocks even mean when the chip manufacturer’s own (and only) version of the product is “overclocked”?

Well they have a reference paper spec which no chip should undercut. Factory OC’d cards sometimes fetch a small-medium premium. Some brands even make deals to bin top chips for their flagship OC’d models.

There are often cards sold without overclocks. Usually for less money (but not always). Allowing more diverse product lines. And the reference spec exists to ensure a minimum performance level.

Technically, though, GPU clocks nowadays are targets. Because the clocks scale based on thermals and other metrics. However, most cards keep a pretty consistent clock level.

The best GPU reviewers will graph clocks to see if a brand has setup their card to maintain advertised clocks.

1 Like

Nvidia clocks have natural overclocking, boost clocks, based on thermals and power. in this case, overclocking would be going past that and messing around with things like voltage and power draw and temp limits to either hit a clock over the normal boost or to keep boost clocks high

there’s also the whole thing where overclocking is usually a warranty-voiding action, depending on part, manufacturer and so on

Probably not much more than a curiosity, but this is interesting

https://www.reddit.com/r/hardware/comments/998zkw/freesync_on_an_nvidia_gpu_through_an_amd_gpu/

1 Like

so, hey, my VEGA64 is back. lets see whether this one actually works properly!

https://www.phoronix.com/scan.php?page=news_item&px=Intel-Announces-Whiskey-Amber

these are nothing special at all, just more 14nm chips, but with all the talk of apple finally releasing another low-end macbook pretty soon, I’d love it if one of the quad cores somehow made it in there, since the touchbar MBP is still dismayingly expensive and 10nm/LPDDR4 is still like, late next year at best

In the face of scrutiny, NVidia doubles down.

1 Like

wish the corporate world had some Greek gods ready to smack down Hubris more consistently

2 Likes

GlobalFoundries abandoned development on their 7nm process, AMD announced it moved to TSMC:

1 Like

I’ve never read a really satisfying study of the economic effects of once-healthy market competition in some domains becoming impossible due to sheer technological complexity and what an appropriate response is to that, but I bet there are other historical examples