Ace: aceofspadeshq at gee mail.com
Buck: buck.throckmorton at protonmail.com
CBD: cbd at cutjibnewsletter.com
Doof: doof2112 at proton.me
joe mannix: mannix2024 at proton.me
MisHum: petmorons at gee mail.com
J.J. Sefton: sefton at cutjibnewsletter.com
A site for members of the Horde to post their stories seeking beta readers, editing help, brainstorming, and story ideas. Also to share links to potential publishing outlets, writing help sites, and videos posting tips to get published.
Contact OrangeEnt for info: maildrop62 at proton dot me
Compared with a 12-core Zen 5 Ryzen 370, the unnamed 10-core sample is 32% faster on single-threaded tasks - a huge jump for a single generation - and 21% faster multi-threaded despite having two fewer cores.
If the leaks that Zen 6 will have 50% more cores are true and its per-core performance is 30% faster, that will put desktop parts at nearly twice the performance of Zen 5.
We don't know how many graphics cores will ship in the laptop parts, but benchmarks of the same engineering sample's four-core integrated GPU placed it a little ahead of the current 860M's eight cores thanks to an update to the newer RDNA4 architecture used in AMD's 9000-series graphics cards.
All in all a solid advance arriving at a time when nobody can afford the hardware to go with it. Good for people running on older AMD hardware - or even those sitting on Intel systems with DDR5 memory.
Tech News
Speaking of Intel, the company's new Nova Lake desktop chips will start arriving early next year, though the high-end 52-core parts might be nearly a year away. And then in 2028, it will be followed by Razor Lake. (WCCFTech)
For which we have no details at all except that it will fit on the same motherboards.
I plugged in my old laptop with its 8GB of RAM and powered it up, and it... Worked just fine actually, but was using 75% of its RAM by the time it reached the Windows desktop. We'll see if the improvements make it useful or if it's time to put Linux on it.
Only 1% of the repositories checked are affected, but there are a lot of such repositories.
Are the impressive coding test scores on some AI models are result of "teaching to test" - developers training the model to answer known test questions rather than the much harder task of improving the model generally? It will surprise few to learn that the answer is often yes. (MaxxBench)
Tests that changed the questions did not much faze Claude and ChatGPT, but Chinese models like DeepSeek and Qwen fell off a cliff.
Figured out what those 8" iron staples were used for, and also why I didn't work it out earlier. The ones I found last year in the first stage of the garden project were twisted and mangled as if they had been pulled out of something and discarded.
The one I discovered yesterday buried straight down in hard clay was still fulfilling its duty - pinning down a tree root that was close to the surface.
So I'll need to work around them (and the roots) rather than dig them out.