[Hardware]HW News - Intel Ray-Tracing Woes, EK's "Not A Sham" Water Blocks, Steam Deck Booming

[Maheso]

is week in the news, we're talking about Intel's early ray tracing issues with Linux, rumors about the Raptor Lake 13th Gen CPUs and RTX 4090 GPUs, and news from EK -- it put gold on blocks, for some reason, and called them an "investment."

Great.

Additional topics include the Steam Deck's continual boom and rise, Windows 12's arrival, and more.

Phoronix reported that a new driver update for Intel’s Arc drivers resulted in a 100x increase to Vulkan ray-tracing performance. According to Phoronix and the driver update notes, the improvement is a result of memory mismanagement.

This was ultimately a programming issue, not a hardware issue, and it’s also a perfect example of the driver problems we covered in our review of the Arc A380. When Intel is at the point of making 100x multiple gains in performance because its drivers are currently in an unacceptable state, it becomes clearer why Intel was so hesitant to even launch the A380.

The company obviously has a QA deficit if it can’t catch basic issues like one coding error causing this type of impact -- and we have a lot we want to add to the driver discussion coming up in a video soon.

Before anyone goes there: This isn’t a “fine wine” situation. Improving the liquid from poison to wine just makes it drinkable, not inherently good.

This change won’t affect Windows performance, so nothing in our review changes from this update.

kopite7kimi is back again with another RTX 40 Series rumor regarding the 4090’s performance in 3DMark Time Spy Extreme. The tweet is extremely simple, just saying, “RTX 4090, TSE >19000.” In the replies, kopite confirms this is a graphics score rather than overall, and that the card will be using PCIe Gen 4.

If we take that graphics score at face value, it calculates to 84% higher performance than the average RTX 3090 in the Time Spy Extreme results database, which is a score of 10300. This is a slightly milder increase than the massive claims of earlier leaks which ranged between 2x and 2.5x. However, this is a much better jump than Turing to Ampere, which was 47% between the 2080 Ti - on TU-102 - and 3080 - on GA102.

We expect power consumption to be much higher on the 40 series versus the 30 series. As for literal cost in dollars, we don’t yet know what to expect given the current oversupply concerns and rumors of Nvidia pushing back the 40 series launch.

Speaking of Raptor Lake, some Geekbench results for the 13600K and 13700K have been uncovered by harukaze and Benchleaks on Twitter. There are three total results between the two CPUs. Two of them are from the 13600K (one from an ASUS ROG Maximus Z690 Extreme, and the other from an ASRock Z690 Steel Legend motherboard), and one is from the 13700K on the same ASRock board.

The alleged 13600K ASUS result comes in at 2012 single-core and 16054 multi-core, while the ASRock result reports 1980 single-core and 14425 multi-core. 

The leaked 13700K with two more P-cores results in 2090 single-core and 16542 multi-core scores. If we compare against the two alleged 13600K scores, that calculates to roughly a 4% to 6% single-core and 3% to 15% multi-core lead. It’s too early to tell which results are more accurate, given that we haven’t actually gotten to test these yet.

All results show as Valid on the Geekbench results page, unlike an earlier 13900K result that was making the rounds. The Raptor Lake leaks are beginning to come fast now that most Z690 motherboards have compatible BIOSes available.

 

https://www.gamersnexus.net/news-pc/3690-hw-news-intel-ray-tracing-ek-gold-blocks-steam-deck