RTX 3000 mobile TGP - early discussions

[FerrariVie]

Hey guys,

What do you think about the TGP on the new 3000 mobile GPUs? I found some trustable info online (as below) about overall TGPs, but no way of knowing for sure which are the TGPs that PCS (Clevo/Tongfang) will be using until they start shipping and people start reviewing it. So the best we have right now is this:

Variant	3060	3070	3080
Max-Q	60 - 80 W	80 - 90 W	80 - 90 W
Max-P	90 - 115 W	115 - 125 W	125 - 150 W

I think it's pretty much the same as the 2000 series and it's kind of what I imagined it would be, apart from the 3070 that I was not expecting 125W to be available. That is the GPU to get if it ships with 125W, imo.

 

[SpyderTracks]

Hey guys,

What do you think about the TGP on the new 3000 mobile GPUs? I found some trustable info online (as below) about overall TGPs, but no way of knowing for sure which are the TGPs that PCS (Clevo/Tongfang) will be using until they start shipping and people start reviewing it. So the best we have right now is this:

Variant	3060	3070	3080
Max-Q		80 - 90 W	80 - 90 W
Max-P	90 - 115 W	115 - 125 W	135 - 150 W

I think it's pretty much the same as the 2000 series and it's kind of what I imagined it would be, apart from the 3070 that I was not expecting 125W to be available. That is the GPU to get if it ships with 125W, imo.

From what I've heard there are going to be about 10 different models of each mobile card, each with different power ratings. It's utterly crazy!

NVIDIA will release up to 10 variants of mobile GeForce RTX 3080 with different frequencies and memory (phonemantra.com)

And the card number (3080), has no relation to the desktop variant chip. The Mobile 3080 is actually a modified 3070 chip. To make it even more confusing.

 

[barlew]

From what I've heard there are going to be about 10 different models of each mobile card, each with different power ratings. It's utterly crazy!

NVIDIA will release up to 10 variants of mobile GeForce RTX 3080 with different frequencies and memory (phonemantra.com)

And the card number (3080), has no relation to the desktop variant chip. The Mobile 3080 is actually a modified 3070 chip. To make it even more confusing.

Thats a great article mate thanks for sharing.

 

[FerrariVie]

omg, things are now going to get even more confusing:

NVIDIA eliminates Max-Q and Max-P labels from its GeForce Mobile products

Apparently, NVIDIA has decided to remove the Max-Q and Max-P labels from its GeForce Mobile products. Differential labels concerning lower consumption (Max-Q) and performance (Max-P) will be removed u...

www.guru3d.com

 

[Macco26]

Hey guys,

What do you think about the TGP on the new 3000 mobile GPUs? I found some trustable info online (as below) about overall TGPs, but no way of knowing for sure which are the TGPs that PCS (Clevo/Tongfang) will be using until they start shipping and people start reviewing it. So the best we have right now is this:

Variant	3060	3070	3080
Max-Q		80 - 90 W	80 - 90 W
Max-P	90 - 115 W	115 - 125 W	135 - 150 W

I think it's pretty much the same as the 2000 series and it's kind of what I imagined it would be, apart from the 3070 that I was not expecting 125W to be available. That is the GPU to get if it ships with 125W, imo.

Ionico models have the same chassis of Max-15 and Max-17 sold in USA. Plenty of information about their GPU TDP there. They are RTX Max-P. Defiance being Clevo should be Max-Q rated instead, like XMG Pro line.

 

[FerrariVie]

The first benchmarks started popping up today:

 

[Macco26]

Those are Max-Q though, as they are applicable to like PCS Clevo barebones. For the Tongfang, not yet specifically. But Jarrod's Tech reviewed a 15" AMD + RTX3070 Tongfang (and Notebookcheck started giving some benchmarks of the same machine but in 17" size) and OWNorDisown reviewed a Non-Tongfang laptop with Intel 8 cores + RTX3070, so might be pretty similar to the behaviour of our PCS Ionico, after all.
Anybody should encounter any review (video or website) for Max-15/Max-17/Prometheus with Intel + RTX30, they might be very very similar to our PCS Ionico, and I'd delighted to take a look at it.

 

[FerrariVie]

Official specs by Nvidia:

Comparison with RTX 20 and GTX 10 series:

 

[Tron1982]

I'm wondering, do you have the same graph for the desktop version of the 3000 and 2000 series ?
thanks in advance

 

[Macco26]

Well I don't but have seen in a lot of places, like Anandtech etc. They are astoundingly different, especially in the Power, topping at 320+ watts. Also CUDA cores are different despite the same number branding.

 

[FerrariVie]

I'm wondering, do you have the same graph for the desktop version of the 3000 and 2000 series ?
thanks in advance

I found this one for Desktop cards:

 

[Tron1982]

Thanks a lot !
So, i think i can say "Here i come 3080 GPU" (you expensive piece of chip )

 

[Grumpywurzel]

It's literally a minefield out there with the new 3000 mobile chips, although as Linus says in the vid below, the 2000 series were exactly the same-

 

[FerrariVie]

Thanks a lot !
So, i think i can say "Here i come 3080 GPU" (you expensive piece of chip )

Yeah, the laptop 3080 is the most expensive one available, but at the same time is the one that also has the biggest difference against its desktop counterpart (memory type & size, as well as cuda core count). I think the 3070 is the sweet spot between price and performance

It's literally a minefield out there with the new 3000 mobile chips, although as Linus says in the vid below, the 2000 series were exactly the same-

I've seen that video. The problem is that it's just worse now, as the max-p and max-q naming are not going to be used anymore. Hopefully it will force manufacturers to post the TGP on the spec sheet, however I can also foresee some manufacturers hiding this on purpose to foul people into buying something more expensive.

 

[barlew]

Yeah, the laptop 3080 is the most expensive one available, but at the same time is the one that also has the biggest difference against its desktop counterpart (memory type & size, as well as cuda core count). I think the 3070 is the sweet spot between price and performance


I've seen that video. The problem is that it's just worse now, as the max-p and max-q naming are not going to be used anymore. Hopefully it will force manufacturers to post the TGP on the spec sheet, however I can also foresee some manufacturers hiding this on purpose to foul people into buying something more expensive.

Fully agree with this. The 2000 series were not the same, you new what you were getting because of the naming convention. There was some variation in performance depending usually on cooling performance and factory OC but nothing like what we are about to see.

Now two identically specced 3070 laptops from two separate OEM's could perform completely differently depending upon how many Watts the OEM decides to bang into the GPU.

This is a huge minefield for the laymen user as they are going to see GPU's named the same and assume they perform the same without the technical knowledge or understanding to do the research to work out which GPU is the better deal.

Lets also remember that the delta in performance between the low wattage 3000 and a higher wattage chip will far exceed the delta between the 2000 Max-P and Q models.

 

[FerrariVie]

Fully agree with this. The 2000 series were not the same, you new what you were getting because of the naming convention. There was some variation in performance depending usually on cooling performance and factory OC but nothing like what we are about to see.

Now two identically specced 3070 laptops from two separate OEM's could perform completely differently depending upon how many Watts the OEM decides to bang into the GPU.

This is a huge minefield for the laymen user as they are going to see GPU's named the same and assume they perform the same without the technical knowledge or understanding to do the research to work out which GPU is the better deal.

Lets also remember that the delta in performance between the low wattage 3000 and a higher wattage chip will far exceed the delta between the 2000 Max-P and Q models.

Indeed. I've seen a Jarrod'sTech video on youtube where a 3070 125W (140W with dynamic boost) outperformed the 3080 90W (105W with DynBoost) on basically all games tested, while still being way cheaper. The CPUs were different as well (R7 on the 3070, i7 on the 3080), but I don't believe that the difference is due to CPUs only.

 

[SpyderTracks]

Indeed. I've seen a Jarrod'sTech video on youtube where a 3070 125W (140W with dynamic boost) outperformed the 3080 90W (105W with DynBoost) on basically all games tested, while still being way cheaper. The CPUs were different as well (R7 on the 3070, i7 on the 3080), but I don't believe that the difference is due to CPUs only.

This is where it gets really dicey! In the past system integrators have never been very clear on what version of GPU is installed, with this situation, there's going to NEED to be far more transparency to aid the buyer into getting the right unit for their needs.

 

[FerrariVie]

This is where it gets really dicey! In the past system integrators have never been very clear on what version of GPU is installed, with this situation, there's going to NEED to be far more transparency to aid the buyer into getting the right unit for their needs.

I hope that PCS starts adding the TGP of their chassis on the configurator from now on, or at least on the detailed specs page.

 

[Macco26]

You're right. I had to make the decision on pre-ordering my Ionico based exclusively on a competitor product, which on the opposte, details very carefully that this chassis has full RTX3070 Max-P up to 125W+15W. But it sounds weird to rely on third party sellers.

 

[TheMash]

I hope that PCS starts adding the TGP of their chassis on the configurator from now on, or at least on the detailed specs page.

I hope so.
They should start customising (and optimising) a little bit more their bioses, as it start to fall short in terms of optimisation and features compared to competition.