Well, until AI replaces them, would you rather have them publish quick leak summaries for individual evaluation or ban them?More like they are good at following certain people on Xitter, refreshing their feed on a regular basis, and when something interesting pops up, crack their knuckles and essentially write a minimum of 3x the number of words actually needed to convey the information presented in the source (sadly, this often miscommunicates the original information in their efforts to write more than needed since it allows them more ads in between paragraphs).
If AI can be useful for something, it's to create a news aggregator which can do a better job of compiling news than these hacks. I don't care if AI runs them out of a job.
A few hours ago I've posted the DL link to the one that works but it was deleted by mods, I guess.I tried but I can't seem to find a version that runs correctly, including the only matching v0.7.0 download I could find - They all just barely load the CPU at all (3-5%) and run forever without finishing.
Ban them. Their summaries are garbage and often times filled with inaccuracies to begin with.Well, until AI replaces them, would you rather have them publish quick leak summaries for individual evaluation or ban them?
Thank you for trying it, man!I tried but I can't seem to find a version that runs correctly, including the only matching v0.7.0 download I could find - They all just barely load the CPU at all (3-5%) and run forever without finishing.
And yet they are vigorously quoted on this forum. Circle is closed.Ban them. Their summaries are garbage and often times filled with inaccuracies to begin with.
Not all leaks originate from AT though. I'm ok with them quickly summarizing leaks & findings and then applying common sense myself to filter out what is BS and not.Also, we all know that AT Forums is the GOAT for hardware leaks anyways. Why waste my time when I can get the info straight from the source.
Acquired the correct version, running it now.Thank you for trying it, man!
you need this one https://www.overclockers.ua/editorial/files/LinX v0.7.0_AMD.zip from here https://www.overclockers.ua/editorial/files/I tried but I can't seem to find a version that runs correctly, including the only matching v0.7.0 download I could find - They all just barely load the CPU at all (3-5%) and run forever without finishing.
If AI can be useful for something, it's to create a news aggregator which can do a better job of compiling news than these hacks. I don't care if AI runs them out of a job.
That's a good point. I remember some of the language in the Lunar Lake article was really weirdly written, as some people in this forum pointed out. I know WTFTech uses AI tools to generate thumbnails, so I wouldn't be surprised if they try or do use AI to spit out articles faster.AI is exactly what leads to those totally wrong takes while rehashing simple stories. I have my suspicions some of lulzy and kinda hard to explain mistakes in texts on some big english sites might be due to somebody trying to use some AI summarizing tool and didn't spot it outputs nonsense. There was some really weird stuff in the Computex Coverage.
AI thumbnails or title pics is pretty understandable. I face the need for "okay looking" picture for the article thumbnail/title quite often and sometimes I had hard time with coming with something related. DepositPhotos and similar libraries help too, but AI generators are actually okay for the job. The pic doesn't need to be correct or factual or anything, it's just stuffing for decorative purposes. Although there is actually the question if the image generators really have right to give you permission to use their outputs if the authors of artworks it's been trained on gave no consent to use their work like that...That's a good point. I remember some of the language in the Lunar Lake article was really weirdly written, as some people in this forum pointed out. I know WTFTech uses AI tools to generate thumbnails, so I wouldn't be surprised if they try or do use AI to spit out articles faster.
Acquired the correct version, running it now.
7950X stock with 2x16GB 6000C30 EXPO
View attachment 101819
Quite a bit slower, but my memory config is worse. I used to have a stable 6400C30 2133FCLK config with this ram but it's lost in a bios backup that is not plaintext and doesn't restore onto the current version. Maybe I'll finally get around to flashing back to extract that config.
I'll throw my own hat into the ring here and ask what the SIMD perf increase is.So after all this turbulence with the latest benchleaks, where are we w.r.t. INT perf increase?
Can we expect FP at 40+ % , but IPC and/or INT at only 16%? The early leaks indicated the opposite, i.e. INT perf increase being more impressive than FP.
So after all this turbulence with the latest benchleaks, where are we w.r.t. INT perf increase?
Can we expect FP at 40+ % , but IPC and/or INT at only 16%? The early leaks indicated the opposite, i.e. INT perf increase being more impressive than FP.
If accurate, AMD is intentionally sandbagging Zen5 in early perf announcements to give stores time to dump Zen4 stock before being osborned.Sooooo Granite Ridge is around 55% faster. Not bad.
Acquired the correct version, running it now.
7950X stock with 2x16GB 6000C30 EXPO
View attachment 101819
Quite a bit slower, but my memory config is worse. I used to have a stable 6400C30 2133FCLK config with this ram but it's lost in a bios backup that is not plaintext and doesn't restore onto the current version. Maybe I'll finally get around to flashing back to extract that config.
Lol no. 55% faster in this bench. Not in general across 100 benches...If accurate, AMD is intentionally sandbagging Zen5 in early perf announcements to give stores time to dump Zen4 stock before being osborned.
For INT, I’m referring to the 32/40+ % SpecINT 2017 numbers hyped previously in this thread.I'll throw my own hat into the ring here and ask what the SIMD perf increase is.
Especially with AVX512 which I got the early impression was supposed to have double the throughput with Zen5.
@Fjodor2001 When you say FP do you mean SIMD in general, or x87 scalar FP?
This is part of what I was saying in my comment.With the SIMDs being doubled in width from 256-bit to 512-bit, it would be very surprising if FP IPC uplift wasn't larger than INT uplift.
Ok, still impressive though, if correct. So what do we make of the 55%? Can any more general estimates be derived from it, or is it a one off?Lol no. 55% faster in this bench. Not in general across 100 benches...
Ok, still impressive though, if correct. So what do we make of the 55%? Can any more general estimates be derived from it, or is it a one off?
So now we’re back to 10-16%, after a brief revival of the 40+ % ”yeah baby”, and even hyperboling into 55% territory?Synthetic benchmark dominated by AVX? Nearly zero value for extrapolating general performance.
Software raytracing should be branch heavy, no?Synthetic benchmark dominated by AVX? Nearly zero value for extrapolating general performance.
lmao you have to go in there with a hazmat suit on but worth the laughs reading those comments now.I'm enjoying the disbelief of Intel users in their comments section
Holy cow, that comment section is bad. It's wild to think that there are actually people who make a hobby out of mindlessly insulting strangers online.