Discussion Qualcomm Snapdragon Thread

[DrMrLordX]

Apologies if it was linked here as well and I missed it:

No Brother print drivers? y0wza, that's bad. Brother is maybe the only printer brand worth a darn anymore.

 

[MarkizSchnitzel]

I hate Brother. Wireless never worked anyway. Only on Galaxy phones it does for some reason. And on Surface laptop.

 

[FlameTail]

According to JustJosh, AutoCAD works on Snapdragon X Elite.

Source:

https://www.reddit.com/r/snapdragon/comments/1djxyeu/snapdragon_x_elite_laptops_what_works_and_what

 

[CouncilorIrissa]

The only relevant benchmark data has arrived.

https://twitter.com/x/status/1804086107212255310

 

[Nothingness]

The only relevant benchmark data has arrived.

https://twitter.com/x/status/1804086107212255310

This gives this (I hope I did not make errors...)


So Oryon is below x3 in int work/GHz.

 

[poke01]

Can no one beat Firestorm?

No one can do it, not even the person who created it, lol. What a beast of an architecture!!

Fours years and still kicking

 

[okoroezenwa]

This gives this (I hope I did not make errors...)
View attachment 101577

So Oryon is below x3 in int work/GHz.

Weird that M1 is listed at 3GHz frequency. Wasn’t it 3.2?

 

[CouncilorIrissa]

Can no one beat Firestorm?

No one can do it, not even the person who created it, lol. What a beast of an architecture!!

Fours years and still kicking

My 16 inch M1 Pro laptop is looking better by the day. So glad I picked it up a year ago with a discount.

 

[FlameTail]

I mean those numbers are funny. M1 -> M2 is an IPC regression??

 

[poke01]

I mean those numbers are funny. M1 -> M2 is an IPC regression??

M1 should be at 3.2GHz.

 

[FlameTail]

I did this math, and I discovered that Firestorm in M1 has higher PPC than Firestorm in A14 Bionic, and even Avalanche in A15 Bionic.

 

[CouncilorIrissa]

View attachment 101580
I did this math, and I discovered that Firestorm in M1 has higher PPC than Firestorm in A14 Bionic, and even Avalanche in A15 Bionic.

That's entirely possible. Even the same uarch that's clocked higher can score lower per clock in a memory bandwidth-bound scenario as additional clock cycles aren't doing anything.

 

[Nothingness]

M1 should be at 3.2GHz.

Weird that M1 is listed at 3GHz frequency. Wasn’t it 3.2?

Yeah I guess the Baidu guy made a mistake. With 3.2 GHz the M1 score/GHz goes from 2.47 to 2.31, placing it behind M2 as I would expect.

 

[Hitman928]

Yeah I guess the Baidu guy made a mistake. With 3.2 GHz the M1 score/GHz goes from 2.47 to 2.31, placing it behind M2 as I would expect.

View attachment 101582

Unfortunately I think there are multiple errors for the frequencies. 7700x has a 5.4 GHz listed max boost, 13700k is also 5.4 GHz. Technically the 7700x has an Fmax of 5.55 GHz but it's unlikely to actually run that high in SPEC. For sure it's not getting 5.65 GHz. M2 clocks at 3.5 GHz (maybe 3.49 technically), not 3.45 GHz. I believe the X3 goes up to 3.36 GHz as well but someone could check me on that.

These results look a little suspect to me but I guess it's the best we have for now, but for sure, the listed frequencies are not reliable.

 

[Nothingness]

Unfortunately I think there are multiple errors for the frequencies. 7700x has a 5.4 GHz listed max boost, 13700k is also 5.4 GHz. Technically the 7700x has an Fmax of 5.55 GHz but it's unlikely to actually run that high in SPEC. For sure it's not getting 5.65 GHz. M2 clocks at 3.5 GHz (maybe 3.49 technically), not 3.45 GHz. I believe the X3 goes up to 3.36 GHz as well but someone could check me on that.

These results look a little suspect to me but I guess it's the best we have for now, but for sure, the listed frequencies are not reliable.

Yeah that seems dubious. Modern CPU frequency variations are a real pain for that kind of cross CPU uarch comparisons. I guess the only acceptable way would be to lock CPU frequency the closest possible to boost frequency, but I'm afraid that's not possible for all platforms.

I'm glad we don't have this issue when doing CPU simulations, we leave the pain to the implementation guys 😀

 

[SpudLobby]

Stop testing exclusively desktop Intel/AMD chips on Spec for christ’s sake. Anyways, here’s another datapoint with MTL, Raptor Mobile, Phoenix, M3.

 

[Nothingness]

Stop testing exclusively desktop Intel/AMD chips on Spec for christ’s sake. Anyways, here’s another datapoint with MTL, Raptor Mobile, Phoenix, M3.
View attachment 101600

As far as "IPC" goes, this is not massively different from what we have above.
But I agree we should only compare laptop vs laptop chips.

 

[CouncilorIrissa]

Any clue as to why it's mostly Chinese sources that are SPEC licensees? We have Geekerwan, David Huang and now whoever this is.

 

[FlameTail]

Have you guys forgotten this claim?

 

[gdansk]

Any clue as to why it's mostly Chinese sources that are SPEC licensees? We have Geekerwan, David Huang and now whoever this is.

It's possible some people don't take the terms of the license very seriously.

 

[CouncilorIrissa]

Have you guys forgotten this claim?
View attachment 101605

That's the mythical 4.3 GHz devkit I guess.

 

[FlameTail]

That's the mythical 4.3 GHz devkit I guess.

Not that.

From the numbers we have now current X Elite SKUs, there is no way the 4.3 GHz SKU is going to match the M2 Max at 30% less power.

 

[CouncilorIrissa]

Not that.

From the numbers we have now current X Elite SKUs, there is no way the 4.3 GHz SKU is going to match the M2 Max at 30% less power.

Power is of course unrealistic, but the 1T score is likely realistic. It's just that the slide was done before QCOM got RDNA3'd.

 

[The Hardcard]

Oops. I have too many threads open, not sure how I missed that SPEC17 post.

 

[Nothingness]

Somebody ran SPEC2017 on X Elite:

This has been posted above and there are some doubts about the frequencies in the graph.