Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > sci.physics > #896261 > unrolled thread

Light Tuning of LLMs Locally (Was: Micro penis walking around with a rucksack)

Back to article view | Back to sci.physics

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  Light Tuning of LLMs Locally (Was: Micro penis walking around with a rucksack) Mild Shock <janburse@fastmail.fm> - 2026-07-13 13:28 +0200
    PCIe 6.0 x16 versus LPDDR5X-8000 latency (Re: Light Tuning of LLMs Locally) Mild Shock <janburse@fastmail.fm> - 2026-07-13 15:56 +0200

#896261 — Light Tuning of LLMs Locally (Was: Micro penis walking around with a rucksack)

Hi,

Interestingly  "Strix" (Strix Point) is no longer
the top news, as "Krackan" (Krackan Point) and the
massive "Halo" (Strix Halo) have officially stepped
into the spotlight.

The New AMD Ryzen Ai Halo Mini PC Has The Most Powerful iGPU!
https://www.youtube.com/watch?v=AZpCPX_EJPg

My tester Acer Swift Go was already "Krackan" (Krackan
Point. So a shift from local inferencing, to limited
local machine learning is possible. And AMD
Ryzen is hitting the drums:

AMD Ryzen™ AI Halo: Build the AI You Want. Your Stack. Your Rules.
https://www.youtube.com/watch?v=lxTaJ5LDY20

Pre-orders for the platform open in June, featuring
128GB of unified memory which enables developers to
run intensive workloads with models up to 200
billion parameters.

LoL

Have Fun!

Bye

Mild Shock schrieb:
> Hi,
> 
> So we might see more mobile grade GPUs,
> on a finger nail, without any cooling fans
> in the future? Yes my Redmi smart phone doesn't
> 
> have any cooling fans, I do not intend to put
> a RTX 5090 in a rucksack and carry a 600 W battery
> pack (*), to have some AI on my phone.
> 
> Have Fun!
> 
> Bye
> 
> (*) The Bluetti Premium 30 V2 is 4.3 Kg heavy:
> https://www.galaxus.ch/de/s4/product/bluetti-premium-30-v2-320-wh-430-kg-power-station-65489839 
> 
> 
> Mirco penis was loosing his marbles:
>> you stinking sack of rocks, you can't read your own redundant links. 
>> Watch the prices for the
>  > Ryzen AI laptops. Fucking idiot. Not worth the
>  > price for what they give in AI. Those are embedded
>  > cpu/gpu cretin, you don't need to read more.
> 
> Mild Shock schrieb:
>> Hi,
>>
>> For marketing purposes people
>> typically look at the race towards
>> 2nm, and we find:
>>
>> A month ago, Apple lost its exclusivity
>> on 3 nm smartphone processors with
>> MediaTek’s Dimensity 9400 chip, integrated
>> in the Vivo X200 Pro smartphone. Qualcomm
>> is also in the race with its recently
>> unveiled Snapdragon 8 Elite and set to
>> power the Xiaomi 15 Pro in 2025. However,
>> Apple should regain its position as innovation
>> leader in 2026 with the release of the
>> iPhone 18, which should feature the A20
>> chip built on TSMC’s 2 nm process."
>>
>> But there is a vertical vias revolution
>> going on as well, some SOCs typically
>> being at 18 layers now:
>>
>> Zooming Into a CPU (It's Incredible)
>> https://www.youtube.com/watch?v=Bez-2cvYja0
>>
>> imec has coined the term CMOS 2.0:
>>
>> LEUVEN (Belgium), MARCH 12th, 2026 — Imec,
>> a world-leading research and innovation hub
>> in advanced semiconductor technologies, has
>> launched a first-of-its-kind consortium with
>> 26 European university groups that will jointly
>> work on the technology roadmap beyond
>> CMOS scaling (CMOS 2.0).
>> https://www.imec-int.com/en/press/imec-launches-university-consortium-around-next-generation-chips 
>>
>>
>> So we might see more mobile grade GPUs.
>>
>> Bye
> 

[toc] | [next] | [standalone]

#896263 — PCIe 6.0 x16 versus LPDDR5X-8000 latency (Re: Light Tuning of LLMs Locally)

Hi,

Its amazing how Micro Penis slumps in
his own turf, when he argues against iGPU
via nonsense like:

 > to the embedded gpu, blocked by the speed of the cpu

Facts are that dual channel RAM has some
advantage over PCIe x16. Take for example latency:

Latency Comparison

Data Pathway  Approximate Latency   Why It Matters
iGPU to LPDDR5X-8000  ~30 to 50 ns
Direct on-die access. The iGPU accesses the system
RAM directly through the CPU's internal fabric,
bypassing the PCIe bus completely.

GPU over PCIe 6.0 x4 Link ~80 to 100 ns
Bus transition penalty. This is just the time it
takes for a packet to leave the GPU chip, travel
down 4 traces, and reach the system controller.

For PCIe 6.0 x16 latency is a little lower,
but still higher than LPDDR5X-8000.

Bye

Mild Shock schrieb:
> Hi,
> 
> Interestingly  "Strix" (Strix Point) is no longer
> the top news, as "Krackan" (Krackan Point) and the
> massive "Halo" (Strix Halo) have officially stepped
> into the spotlight.
> 
> The New AMD Ryzen Ai Halo Mini PC Has The Most Powerful iGPU!
> https://www.youtube.com/watch?v=AZpCPX_EJPg
> 
> My tester Acer Swift Go was already "Krackan" (Krackan
> Point. So a shift from local inferencing, to limited
> local machine learning is possible. And AMD
> Ryzen is hitting the drums:
> 
> AMD Ryzen™ AI Halo: Build the AI You Want. Your Stack. Your Rules.
> https://www.youtube.com/watch?v=lxTaJ5LDY20
> 
> Pre-orders for the platform open in June, featuring
> 128GB of unified memory which enables developers to
> run intensive workloads with models up to 200
> billion parameters.
> 
> LoL
> 
> Have Fun!
> 
> Bye
> 
> Mild Shock schrieb:
>> Hi,
>>
>> So we might see more mobile grade GPUs,
>> on a finger nail, without any cooling fans
>> in the future? Yes my Redmi smart phone doesn't
>>
>> have any cooling fans, I do not intend to put
>> a RTX 5090 in a rucksack and carry a 600 W battery
>> pack (*), to have some AI on my phone.
>>
>> Have Fun!
>>
>> Bye
>>
>> (*) The Bluetti Premium 30 V2 is 4.3 Kg heavy:
>> https://www.galaxus.ch/de/s4/product/bluetti-premium-30-v2-320-wh-430-kg-power-station-65489839 
>>
>>
>> Mirco penis was loosing his marbles:
>>> you stinking sack of rocks, you can't read your own redundant links. 
>>> Watch the prices for the
>>  > Ryzen AI laptops. Fucking idiot. Not worth the
>>  > price for what they give in AI. Those are embedded
>>  > cpu/gpu cretin, you don't need to read more.
>>
>> Mild Shock schrieb:
>>> Hi,
>>>
>>> For marketing purposes people
>>> typically look at the race towards
>>> 2nm, and we find:
>>>
>>> A month ago, Apple lost its exclusivity
>>> on 3 nm smartphone processors with
>>> MediaTek’s Dimensity 9400 chip, integrated
>>> in the Vivo X200 Pro smartphone. Qualcomm
>>> is also in the race with its recently
>>> unveiled Snapdragon 8 Elite and set to
>>> power the Xiaomi 15 Pro in 2025. However,
>>> Apple should regain its position as innovation
>>> leader in 2026 with the release of the
>>> iPhone 18, which should feature the A20
>>> chip built on TSMC’s 2 nm process."
>>>
>>> But there is a vertical vias revolution
>>> going on as well, some SOCs typically
>>> being at 18 layers now:
>>>
>>> Zooming Into a CPU (It's Incredible)
>>> https://www.youtube.com/watch?v=Bez-2cvYja0
>>>
>>> imec has coined the term CMOS 2.0:
>>>
>>> LEUVEN (Belgium), MARCH 12th, 2026 — Imec,
>>> a world-leading research and innovation hub
>>> in advanced semiconductor technologies, has
>>> launched a first-of-its-kind consortium with
>>> 26 European university groups that will jointly
>>> work on the technology roadmap beyond
>>> CMOS scaling (CMOS 2.0).
>>> https://www.imec-int.com/en/press/imec-launches-university-consortium-around-next-generation-chips 
>>>
>>>
>>> So we might see more mobile grade GPUs.
>>>
>>> Bye
>>
> 

[toc] | [prev] | [standalone]

Back to top | Article view | sci.physics

csiph-web