• We should now be fully online following an overnight outage. Apologies for any inconvenience, we do not expect there to be any further issues.

Question x86 and ARM architectures comparison thread.

Page 18 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Doug S

Diamond Member
Feb 8, 2020
3,633
6,419
136
I wonder how Rosetta wasn't sued by Intel

For what? Intel didn't sue Transmeta and their entire business model was essentially a Rosetta type strategy to directly compete with Intel - they even sued Intel themselves (something to do with low power) and Intel still didn't sue them back over the emulation.

The only argument Intel could make would have been Rosetta 2 "using" Intel's patented instructions, which was supposedly why Rosetta 2 only supported SSE 4.2 at release; since SSE* was so old it was out of patent. They later updated Rosetta 2 to support AVX 2 which was still under patent, but even if Intel had grounds for a suit it would have been pointless since Apple's transition off x86 was complete.
 

gdansk

Diamond Member
Feb 8, 2011
4,589
7,724
136
It would be a tenuous case. As far as I know they didn't sue DEC for FX!32 either.
When was the last time Intel was successful in court preventing another x86 manufacturer even? Was it UMC's 486?

Cyrix/National, AMD, VIA. All won enough to keep making x86 after their legal battles.
 
Last edited:

johnsonwax

Senior member
Jun 27, 2024
407
609
96
For what? Intel didn't sue Transmeta and their entire business model was essentially a Rosetta type strategy to directly compete with Intel - they even sued Intel themselves (something to do with low power) and Intel still didn't sue them back over the emulation.

The only argument Intel could make would have been Rosetta 2 "using" Intel's patented instructions, which was supposedly why Rosetta 2 only supported SSE 4.2 at release; since SSE* was so old it was out of patent. They later updated Rosetta 2 to support AVX 2 which was still under patent, but even if Intel had grounds for a suit it would have been pointless since Apple's transition off x86 was complete.
My understanding is that Rosetta2 doesn't even touch x86 instructions. It lifts x86 binaries to LLVM IR, runs it through LLVM optimizers and compiles to ARM64. Not sure what basis Intel would have to sue.
 
  • Like
Reactions: 511

MS_AT

Senior member
Jul 15, 2024
884
1,777
96
My understanding is that Rosetta2 doesn't even touch x86 instructions. It lifts x86 binaries to LLVM IR, runs it through LLVM optimizers and compiles to ARM64. Not sure what basis Intel would have to sue.
Do you have any links to share about this? I would like to read up further on the topic.
 

poke01

Diamond Member
Mar 8, 2022
4,334
5,654
106
1761088275322.png

1761088357911.png


M5 is out, heres some real world cpu tests. Looking foward to Panther Lake, should be big uplift ffrom Lunar
 

Thunder 57

Diamond Member
Aug 19, 2007
4,080
6,811
136
Wow I think he may be the same guy that runs loserbenchmark. His Twitter is insane! I don't have an account so I can't read the replies to see if anyone calls him out. Some highlights include:

- Royal core is back with "super core" which is the old "reverse hyperthreading" joke.
- Mobile phone SoC's are going to use 18A
- Nova Lake should easily surprass Zen 6 and M4 in ST and match M5, while easily surpassing them both in MT.
- Nova Lake should easily match/surpass Zen 6 in gaming because of bLLC.
- Claims that no one made regarding Intel/Nvidia vs AMD only to call them out and say that it's all BS "heavily sugar coated by their excellent PR and vocal fan base".

Excellent PR? That's a good one lol.
 

Saylick

Diamond Member
Sep 10, 2012
4,060
9,486
136
Wow I think he may be the same guy that runs loserbenchmark. His Twitter is insane! I don't have an account so I can't read the replies to see if anyone calls him out. Some highlights include:

- Royal core is back with "super core" which is the old "reverse hyperthreading" joke.
- Mobile phone SoC's are going to use 18A
- Nova Lake should easily surprass Zen 6 and M4 in ST and match M5, while easily surpassing them both in MT.
- Nova Lake should easily match/surpass Zen 6 in gaming because of bLLC.
- Claims that no one made regarding Intel/Nvidia vs AMD only to call them out and say that it's all BS "heavily sugar coated by their excellent PR and vocal fan base".

Excellent PR? That's a good one lol.
I'm on Xitter, and the usual replies are from Intel sympathizers and/or investors. Ever since Intel stock doubled from their 52-week lows, a ton of people came out of the woodwork to hype up the company and hope for its success.
 

Thunder 57

Diamond Member
Aug 19, 2007
4,080
6,811
136
I'm on Xitter, and the usual replies are from Intel sympathizers and/or investors. Ever since Intel stock doubled from their 52-week lows, a ton of people came out of the woodwork to hype up the company and hope for its success.

About what I would expect. They've pretty much been given a blessing so not surprised about the stock. That's all I'll say about that as it's already a bit OT and certainly no need to get in to politics.
 

OneEng2

Senior member
Sep 19, 2022
874
1,129
106
The most radical thing that has happened to Intel is the government subsidy which should, indeed, have a long term effect on their competitiveness.

In the short run, NVL's biggest potential improvement IMO is if they can lower the latency on the ring buss. I have a feeling that NVL could post some impressive improvements if they give the cores some breathing room on the interface.

As for the ARM vs x86 debate,

Can NovaLake surpass this?

View attachment 132514
From here: https://www.notebookcheck.net/Arrow...-threaded-performance-benchmark.898105.0.html

Seems like 285K is more like 5268. Still a far cry from 6371, but it makes me wonder about the other numbers.
 

Thunder 57

Diamond Member
Aug 19, 2007
4,080
6,811
136
The most radical thing that has happened to Intel is the government subsidy which should, indeed, have a long term effect on their competitiveness.

In the short run, NVL's biggest potential improvement IMO is if they can lower the latency on the ring buss. I have a feeling that NVL could post some impressive improvements if they give the cores some breathing room on the interface.

As for the ARM vs x86 debate,


From here: https://www.notebookcheck.net/Arrow...-threaded-performance-benchmark.898105.0.html

Seems like 285K is more like 5268. Still a far cry from 6371, but it makes me wonder about the other numbers.

Intel also needs to get thier L3 up to par. It's been pretty bad for a few generations now, probably since around the same time they went with a heterogeneous design. If they can't fix that the bLLC version may not live up to the hype.
 

Geddagod

Golden Member
Dec 28, 2021
1,544
1,630
106
Intel also needs to get thier L3 up to par. It's been pretty bad for a few generations now, probably since around the same time they went with a heterogeneous design. If they can't fix that the bLLC version may not live up to the hype.
I think it was ICL where the problems started to get bad. Maybe a result of the dual ring design. RKL was a small regression vs CML/SKL IIRC? But yes, very much agree with your point.
 

Doug S

Diamond Member
Feb 8, 2020
3,633
6,419
136
About what I would expect. They've pretty much been given a blessing so not surprised about the stock. That's all I'll say about that as it's already a bit OT and certainly no need to get in to politics.

Eh I'm happy about the stock going up. I bought a fair sized chunk last year at just under $30 since I figured it didn't have much further to fall. Bad call on my part, but now it is well above what I paid for it so I'm debating whether to get out now and see if I can ride it a little higher.
 

OneEng2

Senior member
Sep 19, 2022
874
1,129
106
Intel also needs to get thier L3 up to par. It's been pretty bad for a few generations now, probably since around the same time they went with a heterogeneous design. If they can't fix that the bLLC version may not live up to the hype.
Fair point.

Still, even latent L3 would likely be a huge step up over going out to main memory.