• New PC for Phil...

    From Philip Herlihy@3:633/10 to All on Sunday, May 31, 2026 20:13:57

    Time I bought a new PC - and I'd be glad of advice.

    My present one is running well, though just occasionally I feel
    something slightly brisker would be nice. Which is not bad as it'll be
    14 years old in November! But it can't run Windows 11, and I need to
    run Windows.

    It has an i7-3770 CPU @ 3.40GHz, 3401 Mhz, 4 Core(s), 8 Logical
    Processor(s), with 18GB of RAM. It has two SATA SSDs, with a 4TB
    spinning disk for backups. Graphics card is a 4GB NVIDIA GeForce GTX
    1050 Ti (added later).

    What do I need it for? Well, Facebook, email, surfing, various AI tools (online). General home office work, spreadsheets, Access databases,
    Word. I doubt I'll get back to coding, but might finally get round to
    fooling around with my graphics tablet and some "art" software (Corel Painter). I do have a project for the next year which will involve some
    CAD work - Chief Architect "Home Designer" which does 3D rendering. I
    don't need that in real time (as I understand it that needs four figures
    spent on a graphics card) but it would be good not to have to wait
    tooooo long for a rendered "3D world" to appear.

    I asked CoPilot what sort of Dell I should get. (I've always liked Dell
    kit - good build quality and value for money) It made some useful
    suggestions and seemed to steer me away from "workstation" level kit as "overkill" for my needs.

    But it also suggested using a builder like Scan to create a custome
    machine from standard components, for value, maintainability and future- proofing, and this might be the time I do that. It also said get a
    processor way beyond what you need now (as you'll need it in seven years
    if you're going to keep the thing) and get twice the memory you think
    you need now. So I'm figuring pushing the boat out on an i9, and
    starting with 32GB of RAM. I don't need a case that looks nice - it'll
    be under my desk - just want it reasonably quiet.

    I'm absorbed by AI - in my seventies I've never learned so much or so
    fast in my life - especiallky since discovering OneNote to help me
    organise my notes. Ideally I'd wait a bit longer for NPUs to mature, as
    I suspect more and more AI processing will be local, but with "Extended Support" due to end in October this is the time to move. I guess that
    NPUs (or whatever comes next) may become bolt-ons in time, but it's
    anyway a reason to go for a really fancy processor.

    If you've got this far, thanks for indulging me. I had help from Jamie
    and others back in 2012 when choosing my Dell Vostro 470, which was
    surely one of the best purchases I've ever made. Any guidance from this "college" of experts will be gratefully received.

    --
    --
    Phil, London

    --- PyGate Linux v1.5.15
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Graham J@3:633/10 to All on Sunday, May 31, 2026 21:01:40
    Philip Herlihy wrote:

    Time I bought a new PC - and I'd be glad of advice.

    My present one is running well, though just occasionally I feel
    something slightly brisker would be nice. Which is not bad as it'll be
    14 years old in November! But it can't run Windows 11, and I need to
    run Windows.

    [snip]

    Why do you need to run Windows 11 ????

    None of the apps you've listed suggest anything about security worries.
    Is there an app you MUST HAVE which only runs on W11?

    You already have quite a powerful machine and it would be a pity to
    pension it off.

    If you really must have W11 why not buy a cheap laptop - might be useful
    if you travel about ...


    --
    Graham J

    --- PyGate Linux v1.5.15
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From RJH@3:633/10 to All on Monday, June 01, 2026 19:49:40
    On 31 May 2026 at 20:13:57 BST, Philip Herlihy wrote:

    Time I bought a new PC - and I'd be glad of advice.

    My present one is running well, though just occasionally I feel
    something slightly brisker would be nice. Which is not bad as it'll be
    14 years old in November! But it can't run Windows 11, and I need to
    run Windows.

    It has an i7-3770 CPU @ 3.40GHz, 3401 Mhz, 4 Core(s), 8 Logical
    Processor(s), with 18GB of RAM. It has two SATA SSDs, with a 4TB
    spinning disk for backups. Graphics card is a 4GB NVIDIA GeForce GTX
    1050 Ti (added later).

    I might have recommended a Mac Mini were it not for the Windows thing ;-)

    Aside from performance, a modern PC should offer some improvements in quietness, heat generation, and power consumption. Although Dells were quite good from an environmental POV - at least in the few office ones I used.

    Also, might be worth waiting - seems like some innovations are arriving soon:

    https://www.cbc.ca/news/business/nvidia-ai-personal-computer-9.7218820

    Although whether you want to be a first adopter or not . . .
    --
    Cheers, Rob
    Sheffield, UK

    --- PyGate Linux v1.5.15
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Philip Herlihy@3:633/10 to All on Monday, June 01, 2026 22:29:12
    In article <10vknok$2f8sf$1@dont-email.me>, patchmoney@gmx.com says...


    Also, might be worth waiting - seems like some innovations are arriving soon:

    https://www.cbc.ca/news/business/nvidia-ai-personal-computer-9.7218820

    Although whether you want to be a first adopter or not . . .



    Yes - I'd seen that, and had been wondering. Google Gemini pulled
    together useful answers to my questions: https://gemini.google.com/share/ffc87cb3f377

    On balance, I do think I need the universal compatibility, though the
    quantum leap in performance is of course attractive. I'm wary of being
    an early-adopter too, so maybe in a couple more years when things have
    settled down I might invest in an RTX Spark laptop. (I note that the
    OEMs will be concentrating on laptops and mini-PCs, rather than the
    tower I feel I want.

    --
    --
    Phil, London

    --- PyGate Linux v1.5.15
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From RJH@3:633/10 to All on Tuesday, June 02, 2026 06:47:59
    On 1 Jun 2026 at 22:29:12 BST, Philip Herlihy wrote:

    In article <10vknok$2f8sf$1@dont-email.me>, patchmoney@gmx.com says...


    Also, might be worth waiting - seems like some innovations are arriving soon:

    https://www.cbc.ca/news/business/nvidia-ai-personal-computer-9.7218820

    Although whether you want to be a first adopter or not . . .



    Yes - I'd seen that, and had been wondering. Google Gemini pulled
    together useful answers to my questions: https://gemini.google.com/share/ffc87cb3f377


    OoI I tried your Qs in Claude, plus a follow-up:

    https://claude.ai/share/b3c63a04-99f3-4f6d-934d-53bb92040e28

    Similar answers - but some more on the parallels with Apple's Silicon. I've
    got a fairly recent Apple computer (M4) and can't say I've noticed what the AI side of things adds. It tries to pre-emptively summarise and generate replies to texts for example - nothing of use to me. But then I'm hardly an AI power user . . .

    Apart from running stone cold and silently, it just gets on with everything as (Intel) before. The only significant change I've noticed has been video encoding - maybe 4x faster than a recent i7 to compress video.

    On balance, I do think I need the universal compatibility, though the
    quantum leap in performance is of course attractive. I'm wary of being
    an early-adopter too, so maybe in a couple more years when things have settled down I might invest in an RTX Spark laptop. (I note that the
    OEMs will be concentrating on laptops and mini-PCs, rather than the
    tower I feel I want.

    I was 'encouraged' onto the Silicon platform as Intel was falling out of Apple's support. Software compatability has been a minor issue for me - it'll be interesting to see what the 'Spark rollout brings . . .

    --
    Cheers, Rob
    Sheffield, UK

    --- PyGate Linux v1.5.15
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Philip Herlihy@3:633/10 to All on Tuesday, June 02, 2026 11:46:33
    In article <10vluav$2ojj0$1@dont-email.me>, patchmoney@gmx.com says...


    It tries to pre-emptively summarise and generate replies
    to texts for example - nothing of use to me. But then I'm hardly an AI power >user . . .


    Certainly some "AI" seems to me to invite dumbing down. I can do my own writing, thanks very much, and it's the organising and summarising of
    content that helps me understand and remember it. Pass that off to a
    machine and I'd just stop learning. Where AI scores for me is as a
    "research and integration" tool. Instead of putting keywords into
    Google and having to filter 25,000 results, if I put in a well-
    structured and fairly complete "prompt" I tend to get exactly what I
    need in just one go. I've found that remarkably helpful in diagnosing computer problems - and when the solution works it verifies the process
    and the tool. And it can come up with unexpected truths that check out. Lately I'll run msinfo32 on a PC, edit out the bulky Event Log stuff
    (maybe I shouldn't?) and then ask the tool questions. Yesterday it
    pointed out that a PC I'm trying to fix for a friend has a very small
    page file manually set - I'd never have thought to look. I have asked
    CoPilot what my likely chances are of getting Linux or ChromeOS to run
    on some computers nearing "Windows 10 obsolescence" and I've had very encouraging and nuanced responses.

    There are things I'd like to try out. I'd like to load the complete
    works of Shakespeare (I think that's available in an LLM Notebook on
    Google somewhere) and then try the reasoning facilities to ask searching questions about characters - how Macbeth might have got on with Hamlet,
    for example! And locally a lot of people are confused about how to
    recycle things. So I'd like to try loading all the guidance I can find
    on recycling as it's done hereabouts and then try the reasoning engine
    to ask how to recycle some odd item, as a prototype for a public tool.

    Recently, someone close to me had to make a huge decision on health
    treatment for a very serious condition, and we were struggling. So I
    crafted a very long and detailed prompt about the condition, the
    treatment options and other details like how the patient felt about this
    or that aspect. I fed that into seven different AI tools, then collated
    the output into a single document with seven "chapters", manually
    formatted for easy readability. We then read through it all in silence together, making up our minds independently. Once we'd reached a
    tentative "decision" on the best way forward, we finally shared them -
    they were the same bar one small detail. That helped us make the
    decision that night - note it didn't make the decision for us. The
    reasoning and integrative power of these tools is awesome, and we're
    only getting started in 2026.


    --
    --
    Phil, London

    --- PyGate Linux v1.5.15
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Daniel James@3:633/10 to All on Tuesday, June 02, 2026 13:06:21
    On 02/06/2026 11:46, Philip Herlihy wrote:
    The reasoning and integrative power of these tools is awesome, and
    we're only getting started in 2026.

    I agree it can be impressive.

    Don't kid yourself, though, the tool isn't "reasoning" at all. It's
    collating the results from a deep search of the knowledge available to
    its model and then formatting that as an output using the terms of the question. Any reasoning about the knowledge was done before it was put
    into the database used by the tool.

    It's just a very sophisticated cut-and-paste exercise.

    Whether the output is any use depends on the quality of the data and on
    the ability of the model to exclude false positives in the search. The
    larger the database the higher the chance that there will be sufficient
    true positives for the model to be able to recognize and exclude false ones.

    --
    Cheers,
    Daniel.

    --- PyGate Linux v1.5.15
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Jeff Gaines@3:633/10 to All on Tuesday, June 02, 2026 12:55:42
    On 02/06/2026 in message <10vmgvt$2th8j$1@dont-email.me> Daniel James wrote:

    On 02/06/2026 11:46, Philip Herlihy wrote:
    The reasoning and integrative power of these tools is awesome, and
    we're only getting started in 2026.

    I agree it can be impressive.

    Don't kid yourself, though, the tool isn't "reasoning" at all. It's >collating the results from a deep search of the knowledge available to its >model and then formatting that as an output using the terms of the
    question. Any reasoning about the knowledge was done before it was put
    into the database used by the tool.

    It's just a very sophisticated cut-and-paste exercise.

    Whether the output is any use depends on the quality of the data and on
    the ability of the model to exclude false positives in the search. The >larger the database the higher the chance that there will be sufficient
    true positives for the model to be able to recognize and exclude false
    ones.

    Interesting, that has always been my view, Google on steroids.

    When I was on my quest for the Lenovo TS140 front panel pinouts I ended up with 3 different answers. all of which were wrong.

    --
    Jeff Gaines Dorset UK
    The only thing Flat Earthers fear is sphere itself.

    --- PyGate Linux v1.5.15
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Philip Herlihy@3:633/10 to All on Wednesday, June 03, 2026 15:51:54
    In article <10vmgvt$2th8j$1@dont-email.me>, daniel@me.invalid says...

    On 02/06/2026 11:46, Philip Herlihy wrote:
    The reasoning and integrative power of these tools is awesome, and
    we're only getting started in 2026.

    I agree it can be impressive.

    Don't kid yourself, though, the tool isn't "reasoning" at all. It's >collating the results from a deep search of the knowledge available to
    its model and then formatting that as an output using the terms of the >question. Any reasoning about the knowledge was done before it was put
    into the database used by the tool.

    It's just a very sophisticated cut-and-paste exercise.

    Whether the output is any use depends on the quality of the data and on
    the ability of the model to exclude false positives in the search. The >larger the database the higher the chance that there will be sufficient
    true positives for the model to be able to recognize and exclude false ones.

    Just starting a "well, yes..." response, but I'm not sure that covers
    it. There's no "intelligence" there (though some might prefer to say
    there's no "consciousness" there. It's a machine. It has distilled statistical patterns of language from huge datasets, and delivers
    results according to algorithms. But you can't deny it's ability to
    parse and recognise language - different prompts, even subtly different prompts, produce different results, and in an appropriate way. It's a
    tool; used with reasonable scepticism and critical judgement it's a very rewarding one. Its ability to give objectively helpful answers to basic questions mean that something like a smart speaker can be invaluable to someone with confusion, for example.

    I'll put it this way. The limitations are well-understood. The
    benefits may be a surprise to some.

    --
    --
    Phil, London

    --- PyGate Linux v1.5.15
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Daniel James@3:633/10 to All on Wednesday, June 03, 2026 18:33:12
    On 03/06/2026 15:51, Philip Herlihy wrote:
    There's no "intelligence" there (though some might prefer to say
    there's no "consciousness" there. It's a machine. It has distilled statistical patterns of language from huge datasets, and delivers
    results according to algorithms. But you can't deny it's ability to
    parse and recognise language - different prompts, even subtly
    different prompts, produce different results, and in an appropriate
    way.

    It has the *ability* to do that, but will not do so reliably ... or
    rather it will "parse and recognise language" but not necessarily
    construct the intended query or deliver the expected answer.

    Also, the algorithms are (generally) non-deterministic. It's possible to
    ask the same question twice and get different answers, even from the
    same AI and algorithm. It depends on the AI system you're using, of course.

    Its ability to give objectively helpful answers to basic questions
    mean that something like a smart speaker can be invaluable to
    someone with confusion, for example.

    I find that assertion rather scary. I would hesitate to give a confused
    person access to an AI system because while AI will often deliver
    surprisingly good responses to questions it can also deliver confusing
    and/or inaccurate ones. The AI does not in any way "understand" that it
    is doing so, nor can it detect any confusion its answer may cause, so it
    can't offer correction or explanation.

    I'll put it this way. The limitations are well-understood. The
    benefits may be a surprise to some.

    Yes, *IF* the limitations are understood there can be surprising
    benefits. Conversely, if people accept the answers without questioning
    them they may be led badly astray.

    --
    Cheers,
    Daniel.

    --- PyGate Linux v1.5.15
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Jeff Gaines@3:633/10 to All on Thursday, June 04, 2026 08:18:26
    On 03/06/2026 in message <10vpogo$3qhgi$1@dont-email.me> Daniel James wrote:

    There's no "intelligence" there (though some might prefer to say there's >>no "consciousness" there. It's a machine. It has distilled statistical >>patterns of language from huge datasets, and delivers results according >>to algorithms. But you can't deny it's ability to parse and recognise >>language - different prompts, even subtly
    different prompts, produce different results, and in an appropriate
    way.

    It has the ability to do that, but will not do so reliably ... or rather
    it will "parse and recognise language" but not necessarily construct the >intended query or deliver the expected answer.

    Also, the algorithms are (generally) non-deterministic. It's possible to
    ask the same question twice and get different answers, even from the same
    AI and algorithm. It depends on the AI system you're using, of course.

    I found that in my search for TS140 pinouts, same question, different
    answers, all wrong.

    It's also very rude. In an aircraft, just as the plane is about to land,
    the most stressful part of a flight, it shouts "retard, retard, at the
    pilot. That's not very nice.

    --
    Jeff Gaines Dorset UK
    If it's not broken, mess around with it until it is

    --- PyGate Linux v1.5.15
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From RJH@3:633/10 to All on Thursday, June 04, 2026 21:13:07
    On 3 Jun 2026 at 18:33:12 BST, Daniel James wrote:

    I'll put it this way. The limitations are well-understood. The
    benefits may be a surprise to some.

    Yes, *IF* the limitations are understood there can be surprising
    benefits. Conversely, if people accept the answers without questioning
    them they may be led badly astray.

    I think another aspect to all this is perception. While (most) people 'know' AI's non-sentient (intelligent and alligned to human values?) there is a case to be made that it is replacing human interaction in a number of ways - from personal companion (even partner) to retail support and advice. Limitations
    and benefits therefore become blurred and difficult to pin down
    quantitatively. AI is being *used* non-critically even if it's known to be unsuitable as such.

    Personally I think AI's a hoot - but then I've retired and use it recreationally. It's how it's playing out in the real world that worries me. For example, someone just sent me a brief deployed via AI by a university
    tutor to grade 200 3000 word undergraduate essays. Essays in all probability written by AI . . .

    Also, research into AI is starting to throw out symptoms of AI - users experiencing cognitive decline, surrender and debt for example. Then there's the job losses, deskilling . . . apocalypses etc.

    On the other hand, AI seems to be just the thing for medical science.

    Dunno :-)


    --
    Cheers, Rob
    Sheffield, UK

    --- PyGate Linux v1.5.15
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Theo@3:633/10 to All on Thursday, June 04, 2026 22:23:26
    Jeff Gaines <jgnewsid@outlook.com> wrote:

    I found that in my search for TS140 pinouts, same question, different answers, all wrong.

    The problem is that to tell whether what it says is correct, you either have
    to know already or do a ton of background reading (not just any sources it happens to link to). It's the kind of compulsive liar which will
    assertively fill in gaps with whoppers just to keep it sounding plausible,
    and unless you become an expert you can't which ones they are. But if
    you've done that much background research then you don't need to ask it the question.

    ie it's a workload multiplier not a workload reducer.

    Theo

    --- PyGate Linux v1.5.15
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Theo@3:633/10 to All on Friday, June 05, 2026 00:37:44
    Philip Herlihy <nothing@invalid.com> wrote:

    If you've got this far, thanks for indulging me. I had help from Jamie
    and others back in 2012 when choosing my Dell Vostro 470, which was
    surely one of the best purchases I've ever made. Any guidance from this "college" of experts will be gratefully received.

    How much money do you have? RAM and storage prices are inflated 2-4x over
    9 months ago, so this is the worst time to buy a PC. Whether prices will
    regain their previous levels is a matter for crystal balls (and speculators) but at the moment it's worth thinking hard about what you actually need.

    Theo

    (that said, I have a PC I'm aiming to sell soon. A build by a local PC building company. Specs off the top of my head, mail me for more details if you might be interested. In Cambridgeshire but can probably box up for posting:

    Ryzen 5900x
    64GB DDR4
    Samsung Gen4 NVMe - think it's 1TB but I'll check
    GTX1650 Super GPU
    Fractal Core 2500 case
    Asus mATX gaming mobo
    Think it's a a Fractal PSU as well
    No Windows licence, but can put a Linux on there for testing)

    --- PyGate Linux v1.5.15
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Jeff Gaines@3:633/10 to All on Friday, June 05, 2026 07:37:59
    On 04/06/2026 in message <Q9*ZShIA@news.chiark.greenend.org.uk> Theo wrote:

    Jeff Gaines <jgnewsid@outlook.com> wrote:

    I found that in my search for TS140 pinouts, same question, different >>answers, all wrong.

    The problem is that to tell whether what it says is correct, you either
    have
    to know already or do a ton of background reading (not just any sources it >happens to link to). It's the kind of compulsive liar which will
    assertively fill in gaps with whoppers just to keep it sounding plausible, >and unless you become an expert you can't which ones they are. But if
    you've done that much background research then you don't need to ask it the >question.

    ie it's a workload multiplier not a workload reducer.


    Indeed and to at least some extent I'd rather have it just search, a
    thread on Stack Overflow, Reddit or Usenet is likely to be more useful and accurate. I keep forgetting the ", -ai" suffix, perhaps browsers will
    offer it as an option eventually.

    --
    Jeff Gaines Dorset UK
    You can't tell which way the train went by looking at the tracks

    --- PyGate Linux v1.5.15
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Philip Herlihy@3:633/10 to All on Sunday, June 07, 2026 11:55:11
    In article <MPG.448698d31897c3969896aa@news.eternal-september.org>, nothing@invalid.com says...

    Time I bought a new PC - and I'd be glad of advice.

    My present one is running well, though just occasionally I feel
    something slightly brisker would be nice. Which is not bad as it'll be
    14 years old in November! But it can't run Windows 11, and I need to
    run Windows.

    It has an i7-3770 CPU @ 3.40GHz, 3401 Mhz, 4 Core(s), 8 Logical >Processor(s), with 18GB of RAM. It has two SATA SSDs, with a 4TB
    spinning disk for backups. Graphics card is a 4GB NVIDIA GeForce GTX
    1050 Ti (added later).

    What do I need it for? Well, Facebook, email, surfing, various AI tools >(online). General home office work, spreadsheets, Access databases,
    Word. I doubt I'll get back to coding, but might finally get round to >fooling around with my graphics tablet and some "art" software (Corel >Painter). I do have a project for the next year which will involve some
    CAD work - Chief Architect "Home Designer" which does 3D rendering. I
    don't need that in real time (as I understand it that needs four figures >spent on a graphics card) but it would be good not to have to wait
    tooooo long for a rendered "3D world" to appear.

    I asked CoPilot what sort of Dell I should get. (I've always liked Dell
    kit - good build quality and value for money) It made some useful >suggestions and seemed to steer me away from "workstation" level kit as >"overkill" for my needs.

    But it also suggested using a builder like Scan to create a custome
    machine from standard components, for value, maintainability and future- >proofing, and this might be the time I do that. It also said get a >processor way beyond what you need now (as you'll need it in seven years
    if you're going to keep the thing) and get twice the memory you think
    you need now. So I'm figuring pushing the boat out on an i9, and
    starting with 32GB of RAM. I don't need a case that looks nice - it'll
    be under my desk - just want it reasonably quiet.

    I'm absorbed by AI - in my seventies I've never learned so much or so
    fast in my life - especiallky since discovering OneNote to help me
    organise my notes. Ideally I'd wait a bit longer for NPUs to mature, as
    I suspect more and more AI processing will be local, but with "Extended >Support" due to end in October this is the time to move. I guess that
    NPUs (or whatever comes next) may become bolt-ons in time, but it's
    anyway a reason to go for a really fancy processor.

    If you've got this far, thanks for indulging me. I had help from Jamie
    and others back in 2012 when choosing my Dell Vostro 470, which was
    surely one of the best purchases I've ever made. Any guidance from this >"college" of experts will be gratefully received.

    Diverting this back on track after digressions into AI (clearly my
    fault) I'd love to know:

    1) Has anyone used a system builder like Scan, and how they got on.

    2) If there are any big developments due in mainstream computing
    hardware due in the next few months meaning I should wait.

    3) Any general advice on how to choose the components for this machine.

    --
    --
    Phil, London

    --- PyGate Linux v1.5.15
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)