Question Qualcomm's first Nuvia based SoC - Hamoa

[poke01]

Qualcomm's working on a 2024 PC chip codename "Hamoa" with up to 12 (8P+4E).

Said to the same cache layout as M1. large private L1$, per-cluster L2$ (cluster = 4 cores, 12MB for every cluster) and a lot of LLC.

source:

https://twitter.com/x/status/1589405172979339264

 

[poke01]

Jon hypes up everything( he is great). The fact that no one can match Gerard is insane. It's all about what your design goals and hiring capable people.

 

[SpudLobby]

Hmm, I don't think it refers to device TDP cause that includes RAM and SSD???

It does. Andrei said as much in another forum. It’s the full platform with the RAM. The whole point of efficient chips is to reduce DRAM access, have an efficient power delivery so just measuring the TDP in software/SoC sensors isn’t enough.

 

[FlameTail]

[Deleted]

 

[Tigerick]

It does. Andrei said as much in another forum. It’s the full platform with the RAM. The whole point of efficient chips is to reduce DRAM access, have an efficient power delivery so just measuring the TDP in software/SoC sensors isn’t enough.

If device TDP is included only RAM, then it is more believable cause RAM has only consumed 2-3W

 

[SpudLobby]

Yours is a fair take, and it sounds reasonable. It's just you know, I've been thinking about those charts published by Nuvia in 2020, with that big blue Phoenix rising above everything else, and yet it's three years later and they are only within the lower border of that (admittedly ambiguous) prediction. I suppose we will have to wait and see what they will come up with in the future.

Keep in mind what they went through between the lawsuits with Apple, acquisition and building for mobile on Qualcomm’s uncore IP, and then the Arm lawsuit.

I suspect the following iterations will have some improvement, nothing huge but something. Probably E Cores too.

 

[FlameTail]

I think it’s the result of different design priorities over the years. Apple wants general-purpose GPU cores that perform well on complex workloads. Qualcomm has traditionally prioritized mobile graphics over compute ability.

DAT IS CONCERNING.

Qualcomm has said they are not going after Gamers with the X Elite. They are aiming the product at enterprise/business, office and general usage.

In that case the GPU compute performance needs to be strong

 

[SpudLobby]

Beating the 7940HS when you control device TDP — doing so by almost 50% iso-power.

Qualcomm do have 4 more cores, but AMD have SMT.


Great showing and proves this thing will probably excel with lower power and that other 80W figure is when pushed especially hard. I don’t think AMD are going to be able to compete with this for the average user in a thin and light category.

And we can’t pull the “muh process node” BS. This is just better design.

 

[FlameTail]

Jon hypes up everything( he is great). The fact that no one can match Gerard is insane. It's all about what your design goals and hiring capable people.

Is it crazy that the person we are talking about is literally in this forum? o_0

 

[FlameTail]

It does. Andrei said as much in another forum. It’s the full platform with the RAM. The whole point of efficient chips is to reduce DRAM access, have an efficient power delivery so just measuring the TDP in software/SoC sensors isn’t enough.

Wait. So does that also include Display power consumption.

If so, that's significant isn't it?

The display is one of the most powr hungry components in a laptop.

 

[Doug S]

Now if that team stayed at Apple they likely would have advanced with way more daring changes in that same time, the base already existing and all. But Apple apparently put too little effort in retaining them. It's quite telling that they all thought leaving Apple to form a startup was worth the risk.

I'm not sure it is fair to say "Apple put too little effort into retaining them". Reportedly they wanted to do a server CPU, and Apple had no interest in pursuing that. They left to form their own company that would design an ARM based server CPU.

No doubt Apple paid them well enough they were quite comfortable. It would be impossible to pay them as much as they could have and actually did end up making from starting their own company and having it purchased by Qualcomm - though of course there was some risk involved on their end if it had failed and gone under.

What I find odd is they wanted to do a server CPU and were well along on that path, then Qualcomm bought them out. It seems that ambition has been shelved as there are apparently no plans to release a server CPU. They are back to designing for the exact same market segment they were designing for at Apple. Will they get an itch to go back out on their own and try again? This time with enough money in their pockets they could truly be their own bosses (i.e. don't have to be under the thumb of VCs looking for a quick exit)

Or are they happy to be designing the same mobile/PC CPUs at Qualcomm they did at Apple, and the whole "we want to design server CPUs" was just a cover story for some underlying reason they were unhappy at Apple? If so we will probably never know, but if they stick around at Qualcomm for much longer (they may have made a contractual commitment to remain at Qualcomm for some period as part of the acquisition) one has to wonder.

 

[Doug S]

Wait. So does that also include Display power consumption.

If so, that's significant isn't it?

The display is one of the most powr hungry components in a laptop.

It is, but that may not apply when running benchmarks if the screensaver turns off the display. What we should pay attention to instead is the huge difference between the TDP for the two specs - nearly 60W higher. That's an absolutely massive difference for what is a fairly minor sacrifice in clock speed - from 3.8/4.3 to 3.4/4.0. It seems that high spec version is running beyond where the frequency/power curve starts to take a turn for the worse.

 

[FlameTail]

It is imperative that Qualcomm can iterate strongly an quickly on it's Oryon CPU.

While the product has previewed and is competitive against what's in thr market right now, the competitors are not sitting ducks. Apple, Intel and AMD are all launching successors to their existing gen soon.

Thus it is essential that they strongly iterate on the hardware front, while also making improvements on the software front (drivers, native apps etc...)

 

[FlameTail]

How much better is N3E compared to N4P?

Asking since X Elite is on N4P and one can guess the succesor might be on N3E

 

[moinmoin]

I'm not sure it is fair to say "Apple put too little effort into retaining them". Reportedly they wanted to do a server CPU, and Apple had no interest in pursuing that.

If the interests of a company and that of its employees mismatches the company still has several ways to go about that. If Apple indeed shut down any efforts to work on server chips even if that's what the employees that left absolutely wanted to work on, then it didn't value those employees high enough. It's really not a new thing nowadays to allow highly valued employees to work at least part time on stuff that current company proper see no value in. If that's really how Apple lost most of its previously formidable CPU team, tough luck and zero commiseration (especially considering where they ended up at).

 

[SpudLobby]

Wait. So does that also include Display power consumption.

If so, that's significant isn't it?

The display is one of the most powr hungry components in a laptop.

no lmao.

He measures from the mobo and all, opens the device up. It’s the power draw minus the static draw for these. You don’t care about the display or SSD even. You care about the package+power delivery and the RAM. Those are the independent and dependent variables going into the power draw from the choice of an SoC (because even with the same RAM, that SoC will influence how often it’s used via e.g. cache).

 

[FlameTail]

Also speaking of which, I wonder what the succesor of the Snapdragon X Elite will be called?

I hope it's not Snapdragon X Elite Gen 2.

Don't want them to go back to that chungus naming scheme

 

[Doug S]

How much better is N3E compared to N4P?

Asking since X Elite is on N4P and one can guess the succesor might be on N3E

Per TSMC N4P is 22% better on power or 11% better on performance than N5, while N3E is 34% better on power or 18% better on performance than N5. From the math you can see it is an improvement, but not all that large of an improvement.

 

[SpudLobby]

How much better is N3E compared to N4P?

Asking since X Elite is on N4P and one can guess the succesor might be on N3E

Notably better but primarily via logic density. FinFLEX would come in handy though for power/perf on a given block so we’ll see.

 

[uzzi38]

@SpudLobby Slight correction: it's not full platform power, it's the cooling capacity of the chassis. Significant difference between the two tbh.

From his explanation over on Discord, it sounds a lot closer to what STAPM power indicates on Ryzen laptops.

 

[FlameTail]

He

Geekerwan?

measures from the mobo and all, opens the device up. It’s the power draw minus the static draw for these. You don’t care about the display or SSD even. You care about the package+power delivery and the RAM. Those are the independent and dependent variables going into the power draw from the choice of an SoC (because even with the same RAM, that SoC will influence how often it’s used via e.g. cache).

But we are talking about the Reference Devices that Qualcomm showed. Are you telling me dude opened up that device and touchy touchy the motherboardy to see the powerr consumptiony?

 

[Doug S]

Also speaking of which, I wonder what the succesor of the Snapdragon X Elite will be called?

I hope it's not Snapdragon X Elite Gen 2.

Don't want them to go back to that chungus naming scheme

Unfortunately that is probably one thing Gerard Williams has no control over

 

[FlameTail]

Also speaking of which, I wonder what the succesor of the Snapdragon X Elite will be called?

I hope it's not Snapdragon X Elite Gen 2.

Don't want them to go back to that chungus naming scheme

How abou Snapdragon X2 Elite?

Sounds nice. Reminiscent of how Apple names the M chips.

 

[FlameTail]

Unfortunately that is probably one thing Gerard Williams has no control over

😭😭

 

[FlameTail]

How abou Snapdragon X2 Elite?

Sounds nice. Reminiscent of how Apple names the M chips.

The lower tier version can simply be called Snapdragon X2.

 

[FlameTail]

Also the reference device Qualcomm showed for the X Elite apparently had a whopping 64 GB of RAM, gleaning from the Task Manager Geekerean showed in his video.

Samsung unleashes industry's first LPDDR5X DRAM

Built using the 14 nm technology, Samsung's first 16 Gb LPDDR5X memory chips promise to drive further growth beyond the mobile sector, also targeting 5G and AI applications, as well as the metaverse. The new 16 Gb LPDDR5X comes three years after Samsung introduced the industry's first 8 Gb...

www.notebookcheck.net

Samsung, the speeds of the LPDDR5X DRAM can hit 8.5 Gbps and the 16 Gb LPDDR5X chips can be used in packages of up to 64 GB.

So this means the entire 64 GB RAM is one module? That would be impressive.

For reference if you look at the motherboard of an M2 Max macbook, the M2 Max chip has 4 'modules' of RAM around it.