Large emissive meshes in DSLE. These are 100k triangles, emit light and shadows, however importance sampling is not done by not including them in LightBVH, so the normal BRDF sampling catches them. This makes them essentially free of cost as the path tracer picks the emission based on the object hit.
1. Importance sampled large tree - 4.5ms Trace time
2. No importance sampled tree - 2.5 ms Trace time
3. Vertex color bake of the tree.
4. Alpha bit of the vertex are used to indicate whether it emits light or not.
Since today we are talking about CLOWNS & COCK SUCKERS (mainly ponies & media),
a honorable mention goes to @HardwareUnboxed, who AFTER SPENDING THE ENTIRE RDNA3 GEN DISMISSING AND FUDing #nvidia #DLSS,
today they tell us that semi-decent upscaling -a full generation and a half later- is "AMD's BIGGEST SOFTWARE UPDATE EVER"🤣
There are simply not enough tomatoes in this world for the amount of CLOWNS in gaming media🤡🍅
https://t.co/9uGnGRCELp
AMD moved RDNA 1 and RDNA 2 graphics architectures to a separate driver branch for Windows. The manufacturer stopped adding new Vulkan API extensions for Radeon RX 5000 and RX 6000 series. The fresh driver version 25.30.17.02 supports exclusively new hardware lineups.
This lack of updates caused a loss of compatibility with DXVK 3.0 translation layer. Due to missing Vulkan 1.4 features GPUs use a slow resource binding method with performance drops. Developers recommend staying on versions 2.x or switching to Linux OS.
The community criticized the vendor move due to the artificial reduction of hardware lifespan. Restrictions affected even the Radeon RX 6750 GRE model 2 years after its release. Equipment owners fear the imminent end of full support for RDNA 3 architecture.
#AMD #Radeon #Vulkan #DXVK #GPU #Driver #Windows
Capcom has opened up about how difficult Pragmata’s early development really was,
Director Cho Yonghee said the team’s first playable versions received very harsh feedback from inside Capcom.
“It was devastating,” “We were told that the game wasn’t interesting at all.”
Instead of giving up, the team rebuilt big parts of the game.
Cho said, “Our failures guided us to success.”
Producer Naoto Oyama added that the team kept improving the game through playtesting and feedback until they found the right balance between shooting, hacking, and exploration.
Google reportedly limited Meta’s use of Gemini due to a shortage of compute resources. — FT
Google is in a position where it can’t sell Gemini to Meta as freely as it might want to.
Compute remains power, and the scarcest resource in AI.
The myth of “cheap” DDR5 from CXMT is just that—a myth. Its pricing is already comparable to that of Samsung, SK hynix, and Micron.
So even if Apple successfully adopts CXMT memory, the cost is likely to be similar to what it pays the three established suppliers. That raises the question: why bother?
The answer is leverage.
Apple doesn’t want to be held hostage by long-term agreements (LTAs), supply control arrangements (SCAs), or any single group of suppliers. By bringing in an alternative source, it gains greater bargaining power. Once the market stabilizes, it will be in a much stronger position to push back on pricing.
Looks like the challenger has finally entered the arena.
What if Tim Cook’s supply-chain genius for squeezing the living daylights out of suppliers during lean times isn’t so smart after all, compared to Jensen’s “rough justice” philosophy? Food for thought @PatrickMcGee_
Memory makers strike back.
Everyone wants TPU Co to be the Nvidia killer.
Read the role.
They are still standing up the operating company, vendor ecosystem, procurement stack, partner network, controls, and deployment model.
That’s not a dunk on Google. It’s the point.
Google has great chips.
Nvidia has the AI factory industrial complex already at scale
$GOOGL $NVDA
In the 10 years I have followed the memory market, I have watched Micron try to sign LTAs to varying degrees across the past two boom cycles, with more or less no luck.
So to complain now about memory makers' inability to invest in capacity during bust phases when every cent was getting squeezed out of ASP is embarrassing.
A real FAFO moment for Apple.
A blame game between Apple CEO Tim Cook and Micron over recent price increases on Macs and iPads, with Cook attributing hikes to memory suppliers while a Micron executive counters by blaming aggressive customer pricing during past market downturns.
Talked to an engineer inside of NVIDIA about how they and their team works. Very interesting:
“We don’t need to worry about layoffs. This means no need to worry about elbowing people out of the way to try and make yourself “safe” and important. So everyone helps everyone.”
Went to NVIDIA HQ today.
Two interesting observations:
1. Snacks and coffee are not free: you have to pay for them. This would be unusual at Big Tech, but no big deal for devs here. "We use this thing called salary to buy stuff we actually need." Food for thought (literally!)
How Quake ruined id Software.
There has been a lot of praise of Quake of late, with its 30th anniversary, and it's deserved. Quake is an amazing feat of art, programming, and design. I worked on it, and everything came together almost perfectly from all of us. We ended up with a free-wheeling, frenetic action game with enough of a visible world to grip the imagination.
All the team did a brilliant job, fulfilling tasks just right. But at a grim cost. We worked long and hard, and I think it broke us spiritually.
1/3
Fun fact, this is ray-traced.
This version of the Source Engine uses a combination of pre-calculated raytraced lighting (lightmapping), parallax-corrected cube-maps, global illumination and sparse dynamic lighting.
Do gamers actually hate Ray Tracing?
I get the whole frame rate trade-off, but it's quite literally hardware in your GPU to make complex operations faster, not slower.
As a dev, its presence is such a boon for productivity and actually being able to achieve what I want.