sm_1024

w zeszłym rokudiscuss

Interesting part is they have actually been open for a week now. They did a quiet opening on the 25th and it was still busy that day

sm_1024

2 lata temudiscuss

I see what's going on, they have two HX370 laptops:

  Laptop  MC score  Avg Power
     P16      1213      113 W
     S16       921       29 W
  M3 Pro      1059    (30 W?)

They don't have M3 Pro power numbers, but I assume it is somewhere around 30W, seems like S16 has similar power efficiency as HX 370 at 30 W.

Any more power, and the CPU is much less power efficient, 300% increase in power for 30% increase in performance.

sm_1024

2 lata temudiscuss

I think the OC might have mis-read the power numbers, 110 W is well into desktop CPU power range. Here is a excerpt from Anand Tech:

> In our peak power test, the Ryzen AI 9 HX 370 ramped up and peaked at 33 W.

https://www.anandtech.com/show/21485/the-amd-ryzen-ai-hx-370...

sm_1024

2 lata temudiscuss

119W for hx370 looks extremely sus, seems to me more like the system level power consumption and not CPU-only.

According to phoronix [1,2], in their blender CPU test, they measured a peak of 33W.

Here max power numbers from some other tests that I know are multi-threaded:

--

Linux 6.8 Compilation: 33.13 W

LLVM Compilation: 33.25 W

--

If I plug in 33W into your equation, that would give us score of HX 370: 104 PPA

This supports the HX 370 being pretty power efficient, although still not as power efficient as M3.

[1] https://www.phoronix.com/review/amd-ryzen-ai-9-hx-370/3

[2] https://www.phoronix.com/review/amd-ryzen-ai-9-hx-370/4

sm_1024

2 lata temudiscuss

That is fair, I was taught that decoders for x86 are less efficient and more power hungry than RISC ISAs because of their variable length instructions.

I remember being told (and it might be wrong) that ARM can decode multiple instructions in parallel because the CPU knows where the next instruction starts, but for x86, you'd have to decode the instructions in order.

sm_1024

2 lata temudiscuss

I have heard that part of the reason for little coverage of ryzen mobile CPUs is their limited availability as AMD was focussing on using the fab capacity for server chips.

sm_1024

2 lata temudiscuss

IMO, the most interesting thing about this line is the battery life---within an hour of MBP3 and within 2 hours of Asus's Qualcomm. Making it comparable to ARM architectures.

Which is a little surprising because ARM is commonly believed to be much more power efficient than x86.

[1] https://youtu.be/Z8WKR0VHfJw?si=A7zbFY2lsDa8iVQN&t=277

sm_1024

2 lata temudiscuss

These are interesting results and make a strong case for AMD over Qualcomm, especially in battery life:

  ASUS S16 (Qualcomm):    13h 39m  
  Apple MacBook Pro (M3): 12h 35m  
  ASUS S16 (AMD):         11h 10m  
  ASUS S15 (Intel):        9h 15m

Cinebench Multicore:

  ASUS S16 (Qualcomm):   1,138  
  ASUS S16 (AMD):          997  
  Apple MacBook Pro (M3):  716

Cinebench Single core:

  Apple MacBook Pro (M3): 141  
  ASUS S16 (AMD):         113  
  ASUS S16 (Qualcomm):    108

sm_1024

2 lata temudiscuss

They do make 10 TB+ SSDs:

https://www.allhdd.com/kioxia-kcmyxrug15t3-15.36tb-ssd

Infact, Kioxia's enterprise line goes up to 30 TB/drive.

sm_1024

2 lata temudiscuss

Doesn't microsoft support eBPF on Windows?

https://github.com/microsoft/ebpf-for-windows

sm_1024

3 lata temudiscuss

I'm guessing H100 has 2x host energy overhead for connecting those GPUs? That might offset some of the perf/W benefits of nvidia's offering.

sm_1024

3 lata temudiscuss

Why should I link the wikipedia page? The current link perfectly rebuts OP's claim.

sm_1024

3 lata temudiscuss

> There is no HSR happening.

https://hsr.ca.gov/

SF-LA high-speed rail

sm_1024

3 lata temudiscuss

Modern Android processors are pretty competitive with iPhone's silicon:

S23, 1.4k (Single core), 4.8k (Multi core), 79.3 (3d mark/GPU)

14 pro max, 1.8k (Single core), 5.3k (Multi core), 74 (3d mark/GPU)

Looks like gen2 is already on-par in multi-core and GPU perf. Single core still is 28% faster in iphone.

---

https://www.tomsguide.com/news/galaxy-s23-ultra-vs-iphone-14...

sm_1024

3 lata temudiscuss

If you are using a US SIM card, e.g., all your traffic is tunneled to the US. That could be the reason.

sm_1024

3 lata temudiscuss

I wonder how hard it would be to replace FUSE with userspace block driver. That seems to be focused on performance.

https://lwn.net/Articles/903855/

sm_1024

3 lata temudiscuss

I'd expect bloatware to be unnecessary software. To a user who is not familiar with google's offering, the bloat doesn't exist, they don't see two different softwares. Samsung phones don't have two messaging apps, gallery, or phone app

Not including google's version doesn't equate to bloat. In fact, the opposite is true.

sm_1024

3 lata temudiscuss

That hasn't been my experience at all. If anything, they are more userfriendly than other phones I have used. One thing in particular that is very nice is that everything is accessible on the larger screen sizes, action centers, call logs, messages, everything is designed to be accessible when holding your phone with one hand without having to awkwardly reach the top.

sm_1024

3 lata temudiscuss

> While 243GB/s is massive, and overshadows any other design in the industry, it’s still quite far from the 409GB/s the chip is capable of. More importantly for the M1 Max, it’s only slightly higher than the 204GB/s limit of the M1 Pro, so from a CPU-only workload perspective, it doesn’t appear to make sense to get the Max if one is focused just on CPU bandwidth.

https://www.anandtech.com/show/17024/apple-m1-max-performanc...

sm_1024

3 lata temudiscuss

Latest qualcomm SoCs seem pretty competitive in mobile market (3d mark):

Snapdragon 8 Gen 2: 12726 (76 fps)

A16 Bionic: 9827 (58 fps)

https://nanoreview.net/en/soc-compare/qualcomm-snapdragon-8-...

Submissions

[untitled]

FCC approves Starlink first generation upgrade plan

Ten years after the Huygens landing: The story of its images (2015)

Network Transparency with Wayland

OpenAI releases ChatGPT's hyper-realistic voice to some paying users

Disruptor: High performance alternative to bounded queues

comments