Re: W11 support?
Paul <[email protected]> Mon, 8 Sep 2025 09:11:24 -0400
| Newsgroups | alt.comp.hardware |
|---|---|
| Organization | A noiseless patient Spider |
| Message-ID | <[email protected]> |
On Mon, 9/8/2025 3:49 AM, T wrote:
> On 9/7/25 10:30 AM, Paul wrote:
>> On Sun, 9/7/2025 7:09 AM, T wrote:
>>> On 9/6/25 11:27 PM, T wrote:
>>>> On 9/6/25 11:01 PM, T wrote:
>>>>> On 9/6/25 10:37 AM, Paul wrote:
>>>>>> https://www.xda-developers.com/intel-core-i7-14700k-review/
>>>>>
>>>>> Hi Paul,
>>>>>
>>>>> I picked this processor out from this chart at ASUS:
>>>>>
>>>>> https://www.asus.com/us/motherboards-components/motherboards/ workstation/pro-ws-w680m-ace-se/helpdesk_qvl_cpu?model2Name=Pro-WS- W680M-ACE-SE
>>>>>
>>>>> because it was 35 watts and supported ECC.
>>>>>
>>>>> But the review you sent me said it ran "hot".
>>>>>
>>>>> Something is fishy here. You take?
>>>>>
>>>>> -T
>>>>
>>>>
>>>> Hmmm.
>>>>
>>>>
>>>> https://www.intel.com/content/www/us/en/products/sku/236783/intel-core- i7-processor-14700k-33m-cache-up-to-5-60-ghz/specifications.html? wapkw=i7-14700K
>>>>
>>>> says:
>>>>
>>>> Processor Base Power 125 W
>>>> Maximum Turbo Power 253 W
>>>>
>>>> Maybe ASUS chart made a mistake?
>>>
>>>
>>> Ooops, that was the "K" version.
>>>
>>> This is for the "T" version:
>>>
>>> Processor Base Power 35 W
>>> Maximum Turbo Power 106 W
>>>
>>> So no heat buildup problem here
>>
>> Obviously, a T is not in the same class as a K !!!
>>
>> At one time, we had just the concept of TDP to worry about.
>> Today, some of the processors no longer have the concept
>> of a "base clock". The datasheet leads you to believe
>> they "run from 0..MaxTurbo". And then, there is a continuum on
>> power, just as your 14700K is 125W base, and 253W on turbo.
>>
>> The T version, if you can find the datasheet for it, it
>> is not going to have the same "top-clock". If it DID
>> have the same "top-clock", then all that would mean
>> is the power limit on turbo is 250W or so. You can't
>> have a top clock, without some increased VCore voltage
>> and a hefty thermal output.
>>
>> The difference between 3.5GHz and 5GHz, is Windows Update
>> finishes a bit more snappy. That's on an email/web browser
>> machine. If you're gaming, the battle scene frames
>> on the screen "turn blurry" when there isn't enough CPU
>> to paint the frames. Once the 5GHz is available, the
>> scene should be crystal clear.
>>
>> It all depends on what you want to do with the machine,
>> as to whether a T serves a purpose or not. A lot of
>> the time, you can't even find a T on Newegg to buy. That
>> used to be the problem with those, is they were in
>> pre-built machines, but not a lot of them leaked to retail.
>> They share some characteristics with laptop processors, in
>> terms of the voltage and frequency selection used.
>>
>> If your plan is to use a 35W CPU, then you could cool that
>> with a wet rag. It wouldn't take much of a cooler. My blow-down
>> cooler could handle that (and if you let it, it could even handle
>> the idle CPU with the fan off). Whereas a 253W processor, if
>> you're doing a 7ZIP run and compressing a 3TB backup image,
>> then your cooler choice has to sustain good operating
>> temperatures, during the extended compute run. 7ZIP is the
>> equivalent of a "burn-in software".
>>
>> Since AMD has been on a price-dropping spree recently,
>> some of their lower power processors have lost the in-box
>> cooler, and so buying a cooler is now an extra step for them.
>> You could get good performance from a 65W CPU, plus with
>> the right motherboard choice, the AMD also have ECC support
>> across more of the product line.
>
> Boxed vs tray processors. Boxed gets you the fan and
> the pretty sticker for the front of your box. I do
> not mind the tray units as I typically get a better
> cooler anyway. I am allergic to AMD processors.
>> DDR5 has two kinds of ECC capability, and the traditional
>> extra-chip kind is what you want when doing an ECC box. A
>> USENET poster went through a living hell, getting that to work,
>> and he never did deliver his "Linux numbers" that prove the
>> memory controller is in the correct operating mode. That's how
>> you figure out (partially) that it is working.
>>
>> With the DDR5, there is also an error-injection capability, so
>> you can actually shoot in an error, and see an error appear in
>> the log as a proof of ECC-purchase. It takes a software to run the
>> error injector.
>>
>> I would not blame you for selecting your current mobo choice,
>> if it meant less work getting ECC to work.
>>
>> Always double-check, that the handful of components you are using,
>> really are "ECC ReadY" and it is not s sham. I got screwed out of
>> ECC on my X48, via various press releases (will-she-or-wont-she).
>> I had ECC DIMMs but the BIOS gives the bad news "it won't work".
>> Design-error in chipset (X48).
>>
>> These are purchase patterns for the RAM.
>>
>> 4x32GB DDR5 ECC (suited to a highly reduced clock rate of operation)
>> (slower speed, caused by bus loading of two DIMMs per channel)
>> or
>>
>> 2x48GB DDR5 ECC (the "odd" sized DIMMs can be used and ECC ones are available)
>> (One DIMM-Per-Channel allows the highest operating clock)
>>
>> plus
>>
>> 1x8GB non-ECC DDR5 (this is the single DIMM you put in the motherboard while
>> flashing up the BIOS. You also use this stick, when the
>> motherboard won't start any more, despite resetting the CMOS,
>> and the last ditch effort, is to put the non-ECC single stick
>> back in by itself).
>>
>> Always have that teaser-DIMM ready, on these high tech builds. I didn't need that
>> on my builds - I spent the first evening working in the dark, nothing would
>> run, the BIOS flasher LED was flashing (but it wasn't doing anything). But
>> I eventually got it running after "trying all permutations and combinations"
>> as a technique :-/ On one combination, suddenly the LED frequency was different...
>>
>> Paul
>
> Hi Paul,
>
> Thank you!
>
> This board has a W680 chipset:
> https://www.asus.com/us/motherboards-components/motherboards/workstation/pro-ws-w680m-ace-se/helpdesk_qvl_cpu?model2Name=Pro-WS-W680M-ACE-SE
>
>
> i5-14600K
> https://www.intel.com/content/www/us/en/products/sku/236799/intel-core-i5-processor-14600k-24m-cache-up-to-5-30-ghz/specifications.html
>
> Max Turbo Frequency 5.3 GHz
> Performance-core Max Turbo Frequency 5.3 GHz
> Efficient-core Max Turbo Frequency 4 GHz
> Performance-core Base Frequency 3.5 GHz
> Efficient-core Base Frequency 2.6 GHz
>
>
> i5-14600T (Yes, I did notice they are hard to find: found
> one on FleeBay)
> https://www.intel.com/content/www/us/en/products/sku/236782/intel-core-i5-processor-14600t-24m-cache-up-to-5-10-ghz/specifications.html
>
> Max Turbo Frequency 5.1 GHz
> Performance-core Max Turbo Frequency 5.1 GHz
> Efficient-core Max Turbo Frequency 3.6 GHz
> Performance-core Base Frequency 1.8 GHz
> Efficient-core Base Frequency 1.3 GHz
>
> I can live with the difference. I am after long life
> and low heat means longer life. And I always over cool
> the processor. And this is not a high performance
> gaming design anyway. Just a work horse multitasking
> workstations -- virtual machines and all.
>
>
> I will let Kingston figure out the memory for me:
> 2 x 16GB for 32 GB dual interleaved
>
> https://www.kingston.com/datasheets/KSM48E40BS8KI-16HA.pdf
> KSM48E40BS8KI-16HA, 16GB 1Rx8 2G x 72-Bit, PC5-4800 CL40 288-Pin DIMM
>
> Is dual interleave still a thing?
>
>
> I intend to run Fedora 42+ on it with Windows in qemu-kvm
> Virtual machines. I may retire when W12 comes out.
> This new design is meant last me till I get my
> heavenly reward.
>
> -T
>
DDR5 is two channels per stick. Two DDR5 channels is four channels total.
Yes, there is bound to be page interleaving and better
performance on a certain stride through memory.
You're a technical user, and should have enough memory for technical user things.
For example, I can run an AI on the machine across from me,
because the machine has 128GB of RAM. For your current arch pursuit,
2x48GB is the best compromise setup (to retain a good memory clock
plus gobs of RAM).
*******
The evidence in the room, makes me a "steam powered" guy.
We just can't seem to get the snappy performance, by using
any old thing. Only my 5GHz box does some things "with authority",
while the other boxes are more of a "you'll have to wait" thing.
I want my computers, to run "as intended", not perpetually be
"below the median".
Your two processor specifications, suggest 92W to 253W makes the diff
between 5.1GHz and 5.3GHz. I don't know enough about the architecture
of these, the die split, to know whether some have the "per-instruction overclock"
while the others do not. The 14900K was one of the ones, being degraded
by the choices made when the microcode was done -- this can be patched
via a BIOS update.
14600K
Intel Speed Shift Technology Yes (Uses P-States and clock multiplier)
Intel Turbo Boost Technology 2.0 ("Turbo Boost 2.0 boosts all cores") Essentially custom SpeedStep
Intel Hyper-Threading Technology Yes
Enhanced Intel SpeedStep Technology Yes (Used to use two specific P-States...)
14600T
Intel Speed Shift Technology Yes (same as 14600K - expect the top clocks to differ...)
Intel Turbo Boost Technology 2.0
Intel Hyper-Threading Technology Yes
Enhanced Intel SpeedStep Technology Yes
14700K
Intel Speed Shift Technology Yes
Intel Turbo Boost Max Technology 3.0 Yes (binned cores, custom "best core" identified, 1 or 2 turbo cores [SuperPI benchie])
Intel Turbo Boost Technology 2.0
Intel Hyper-Threading Technology Yes
Enhanced Intel SpeedStep Technology Yes
14900K
Intel Speed Shift Technology Yes
Intel Adaptive Boost Technology Yes (all-core Turbo frequency, like if you locked the cores together)
Intel Thermal Velocity Boost Yes (+1 clock multiplier, if CPU temp 70C or less, as if that will happen)
Intel Turbo Boost Max Technology 3.0 Yes
Intel Turbo Boost Technology 2.0
Intel Hyper-Threading Technology Yes
Enhanced Intel SpeedStep Technology Yes (Maybe when running WinXP native :-) )
285K (different socket, just to see the evolution -- same as 14900K)
Intel Speed Shift Technology Yes
Intel Thermal Velocity Boost Yes
Intel Turbo Boost Max Technology 3.0 Yes
Intel Turbo Boost Technology 2.0
Intel Hyper-Threading Technology No (I added this line, because they would not admit it :-) )
Enhanced Intel SpeedStep Technology Yes
# Handy table of terminology-explainers
https://www.intel.com/content/www/us/en/gaming/resources/how-intel-technologies-boost-cpu-performance.html
Summary: If any advantage is to be had, we can get it in 14700K, Windows Update
finishes 30 seconds sooner.
My opinion, just go by the clock numbers in each table, and... move on :-)
The marketing department buys us a small advantage, but not hundreds of dollars worth.
My 4930K was Turbo Boost 2 behavior, 3.4 to 3.9GHz. Yes, it's faster,
but a good OS (Win10 is getting slower), can easily smother that and
make it seem totally irrelevant. Which is why the "steam powered" part of
me, still wants a good top-clock. It's compensating for the software degradations.
And just because a processor runs a top clock, does not mean the internal
counter-rotating rings or mesh interconnect for cores, feeds memory accesses
quickly enough for the top clock to be useful. I first observed this on my AthlonXP,
where boosting the clock did absolutely nothing. It was the FSB that sucked donkey balls.
It was the same thing with the Celeron 300 running at 450, a pointless exercise
because the memory was running at 300MB per second (slower than an SSD).
Paul