Re: Don't forget the non-monetary advantages
He has a PhD which makes him a doctor. Im not sure why you guys have to lie about this. I dont particularly trust him but lying about him isnt a good look
88 publicly visible posts • joined 16 Jun 2022
But but but muh Elong Muskrat. You have to be voted down because it doesnt matter if you are right, politics comes first. Everything is political Dontchaknow?
Grok does have worse general knowledge than ChatGPT-4o right now but it hallucinates a lot less which overall makes it a better, more consistent tool for our use. Though we continue to use both
She is fighting the oligarchy while flying round on private jets paid for by the very people she is supposed to be fighting
And the most hilarious part of this is not the hypocrisy, though its worth a giggle. Nope its that most of the posters here cant even see the hypocrisy and yet somehow think they are enlightened. You couldnt make it up! Amazing :D
If you already have Windows SQL Server licenses for 2017, 2019 or 2022 then those licenses can be used against Linux SQL Server installs. You dont need to buy new ones
We run our HR stuff on SQL Server on Linux as its a far cheaper option than the other supported alternative, Oracle
I dont see how a company wanting to charge license fees for its own products is "making it hard for people to use Cloud other than Azure using licence money as the weapon"
I use iRedMail with SoGo webmail for my personal setup and its great but its hardly a feature replacement for Exchange, not even close
Its also much messier to admin than Exchange, despite its more limited functionality
Basic things like sender dependant SMTP maps are just not supported by iRedMail. I mean you can make them work but your install is no longer supported. There is no EWS support so enterprise mail is no go. There is ActiveSync support via Z-Push but ActiveSync is no longer supported by Windows or macOS desktops and has never been supported on Linux desktops
and the difference is enormous.
Latency between home and office dropped from 40ms to 3ms on average. The connections are symmetric so I get a full Gigabit VPN connection between home and office and its dirt cheap. £35 for 1GB both ways and £104 for 8GB both ways at the office with static IPv4 allocations. They even let us set PTR records for the static IP's so we moved our mail and web back on prem
Keeping the BT connections as a standby for now but so far the service has been great and the support OK. There have been a couple of "scheduled upgrade" outages but we always get advance notice and they do it between 3am-4am
Makes me wonder why BT cant match a little outfit like YouFibre
I like the idea of the CHIPS Act but its implementation has been a disaster so far. The only real win for the US has been Micron's new DRAM fabs. DRAM isnt what was critically short at the time, it was logic fab capacity so Im not sure how big a win that is. Most of the rest has gone to non US Companies
$280 billion should have the US up with its own, US owned, leading edge logic fabs and yet not one single dollar has been spent on that
They should have setup a US competitor to TSMC instead of paying TSMC to build fabs in the US, which has so far been a spectacular failure. That US fab could have had Apple, AMD, NVidia, TI, etc as immediate customers and those clients should have been involved in the development process and planning
Absolute bollox, you clearly dont work in this arena. Super high core systems have generally very low clocks which makes them poor systems for running single application images. There are pretty much zero applications that scale up upto a 192 core EPYC Turin in a single application image. Those that do generally get better performance from running multiple images across lower core count, higher clock speed servers anyway
High core counts are for managing multiple, lower single thread performance, application images on the same physical hardware, they generally dont scale up a single image. You have got this completely back to front
The Chinese have been focused on asynchronous training for AI from the start. The west has so far been running a purely synchronous model. The wests approach works better at small scale but fails to scale outside of a single data centre cluster
Western companies are starting to wake up to that mistake however and Google is currently leading the curve when it comes to asynchronous training. OpenAI has plans to explore asynchronus training too but they are all behind the Chinese in terms of asynchronous experience at the minute
Interesting article on asynchronous training ambitions of the western AI companies:
Multi-Datacenter Training: OpenAI's Ambitious Plan To Beat Google's Infrastructure
https://www.semianalysis.com/p/multi-datacenter-training-openais
Savage, savage butthurt and self delusion duly noted
SanFran is a complete shithole. The only worse place I ever lived was LA which I lasted 4 months because frankly it was intolerable and made London look like a fairy tale
The only person that seems to think otherwise is you
Not really. While ARM has started to make a small imprint into customer workloads on Amazon and Oracle clouds its still very much of a niche market. In terms of corporate maket share (Nutanix target customers) ARM is still zero or a rounding error at best. There are no tier 1 vendors selling ARM Equipment into the channel yet
I would love to see readily available ARM kit but right now it doesnt exist in volume. We actually run some ARM kit in production but its very low volume kit from specialist OEM and the support is poor
Its never cheaper to move them into Azure cloud. Either you are a lazy boy or you haven't done your sums properly. In no way, shape or form is it cheaper to run Oracle services on Azure than it is on prem, even for a tiny number of servers (8 as you say). Even if you needed a single dedicated staff member for them, its still not cheaper
Azure is roughly 3X the cost of Oracles own cloud services, even Oracles services that run inside Azure. Ive done the maths on this for tens of clients. Never in a single instance did it make financial sense to move on prem Oracle services to Azure
Yeah they are pretty good value. It doesnt show it on that page but there are 128 core variants available now. We bought 5 more in February and the 128 core / 192GB RAM variant was a little over £3600 (ex-vat) per unit with the 4x10GB network upgrades. About the same as high end gaming rig or top specced Apple laptop with 12 cores. Of course grpahics is handled by the BMC and they are terrible but fine for development work as long as you are not doing 3D stuff but you can add pretty much any consumer or professional AMD or NVidia card if you need that
You can but single unit quantities of the 32-128 core variants from i-Pi for not much more:
https://www.ipi.wiki/products/ampere-altra-developer-platform?variant=42970872053922
He was previously using Apple devices for the ARM side but Im fairly certain that he would have been offered an Ampere Developer Platform Desktop. Even we got one of them and we are just a smallish sized dev house doing mainly embedded and optimised cloud images
These are the systems they were giving out, which anyone can also buy:
https://amperecomputing.com/systems/altra/kraken-comhpc-WS
Geoff Greeling did a review if you have interest:
https://www.youtube.com/watch?v=wl5H5rT87JE
Harsh but mainly fair. You are correct in that its not just ASUS. Ive seen the same lazy stupidity with HP, Lenovo and pretty much every OEM. Even with Laptops that are sold as "Linux" laptops, especially from Dell
The linux side DSDT situation is actually quite good now. You can still dump, modify and recompile the DSDT. The live DSDT lives at /proc/acpi/dsdt or /sys/firmware/acpi/tables/DSDT. You can decompile, edit and recompile with iasl which is a standard tool in all distros Ive seen
None of that however helps a non technical user. Its a pretty simple process to dump, edit and re-create a DSDT if someone else has figured out the missing bit. However figuring out how to fix the DSDT itself can be a pretty daunting process, even for seasoned developers, if you dont have good references and documentation which the OEM's just dont provide
I have the same with the 13900H CPU variant that I use as my travelling device. If its the latest model (UX6404) then you can fix it by adding a quirk definition into Linux Mints 6.5 kernel config and rebuilding the CSC3551 module. Its simple enough but can feel a bit of a palava if you have never done it before
Full details here: https://github.com/rykdesjardins/fix-UX6404VI-audio-linux
The cause of the problem is not actually a kernel issue. Its a badly developed BIOS from ASUS which is missing parts of the ACPI/DSDT table. The kernel quirk just works around the ASUS error. Hopefully it will be fixed in a future ASUS BIOS update however this same problem is quite common on a lot of recent ASUS laptops
Im still on a 12 Mini for the work line and a 12 Pro for personal, purely for the camera. I still see no compelling reason to upgrade. Maybe its because I just use a £10 per month pre pay sim with 40GB data and unlimited texts / calls, including EU roaming and buy the handsets at retail. I dont do mobile contracts
Those of you that felt the need to upgrade to the latest 13/14/15s, what were the driving factors in your upgrade decision? I just dont see any
Long term Mint Cinnamon user here. Ive used it exclusively on my laptop, gaming desktop and workstation for the last 5-6 years because its been rock solid stable and the UI has been uncluttered, straightforward and consistent. Its also the standard desktop within my company. However its dependency on Gnome apps is getting worse rather than better, the UI is becoming messy and inconsistent, default themes are a mess and we have had quite a lot of stability issues with v20 onwards. A lot of the default apps now use Gnomes libadwaita for themeing instead of standard GTK themes and look out of place and have different behaviour. Desktop stability has also been a problem since 20.3 with crashes on unlock desktop and coming out of sleep, sound and video stuttering issues across a wide range of dektop and laptop hardware and desktop responsiveness is a bit more sluggish than it used to be. They have also basically said point blank that they have no interest in supporting wayland so Cinnamon is looking like a dead end moving forward. Add to that the refusal to even consider an ARM64 spin even though ARM desktop hardware is now becoming more common
Its not horrific and not bad enough yet to be planning a change, its too much effort right now for a whole fleet of PC's. I do however have this nagging doubt going forward about what they are currently doing and focusing on. Its a shame they dropped their KDE spin because thats probably what I will be looking at next. Gnome 4x is almost universally hated by our devs and users and most of them consider a default install unusable without at least dash2dock or dash2panel extensions. Sadly those extensions have a habit of breaking, even with minor, distro packaged updated
>What guarantees do the Cloud providers give for data recoverability and integrity.
None whatsoever. Ive been through this with clients who have lost data and servers from cloud providers (Azure & AWS) and VPS providers (Vultr, IONOS, HeartInternet and others). Even when you replicate data to multiple zones a stray deletion command in one zone replicates to the others before you can finish your brew and investigate. The contracts are always water tight on the providers side, there is no comeback or compensation when it happens
As an earlier poster remarked: There is no cloud, its a mental abstraction, only other peoples servers. If you are not doing your own, locally stored backups and replicas then you are asking for a disaster. The cloud isnt some magic place where data is secure. Its just a marketing abstraction to make you believe thats the case. Its also fucking expensive compared to running your own, even on a small scale, unless you are in a location where appropriate bandwidth isnt available. In which case you probably chose the wrong location to base your data intensive business
I would second this if you want or need some ARM hardware for testing your software on. What they offer is exceptional for free and Im not quite sure how they are affording to do it. Ive played with the free Oracle instances and performance was fine, pretty good even. Disk IO was heavily limited (think SATA2 HD) but CPU and mem performance was surprisingly good
Just be careful with the egress limits, they can quickly start turning into billable megabytes if you use them regularly
You are quite right. The Ampere CPU's dont respond to over provisioning cores quite as well as the x84_64 stuff though, due to lack of SMT I presume. For mixed, low priority workloads on Epyc we will provision 300% of the cores, on Ampere 180% seems to be the sweet spot for us
Docker security is not amazing and while it has improved over the years there are still many fundamental gaping holes in its architecture. You would be a fool to use docker alone it a multi tenant architecture without additionally separating each tenant with VMs
Take this little beauty from 2014 for example. Closed but not fixed due to lack of upstream support: https://github.com/moby/moby/issues/6324
And its proposed solution from 2016, which is still open for the same reasons: https://github.com/moby/moby/issues/24716
Its very hard to get the docker guys / gals / traps to accept a security related submission, Ive tried many times
I was just thinking that. Lawyers work for people and organisations under the instructions they are given, not the other way round. Whatever way you want to look at this, its a monumental foot shooting from the peeps at the Rust Foundation who organised this. They either didnt bother to brief legal or they didnt bother to check what came back before publishing. Its either incompetence at a monumental scale or they got what they asked for, in which case its greed and arrogance at a monumental scale
This is just avoidance and deflection. Your response has nothing to do with the point that I made and is further self delusion to justify censoring opinions that differ to your own
Deal with the point I actually made:
>you should tolerate all speech, but by doing so, you are promoting hate speech
This is a completely false dichotomy and its present in the thought process of extremists on both the left and right. In no way, shape or form is tolerance the same as promoting, its not even in the same ball park. You are being very dishonest here. Either intentionally or because you dont have the self awareness to see your own contradictions
>you should tolerate all speech, but by doing so, you are promoting hate speech
This is a completely false dichotomy and its present in the thought process of extremists on both the left and right. In no way, shape or form is tolerance the same as promoting, its not even in the same ball park. You are being very dishonest here. Either intentionally or because you dont have the self awareness to see your own contradictions
This is why censorship is so dangerous, especially around "hate speech" which has no clear definition. Ask 10 different people what qualifies as "hate speech" and you get 10 different answers. Thats why we created our cultures around free speech with a very specific exception of incitement to violence. Anytime you go further than that its for censoring opinions that differ from your own, regardless of the lies you tell yourself to justify it