RTX 2080 mobile VS RTX 3060 mobile.

[Csupati]

HI Guys!
Im thinking about that what aspect makes the 3060 weaker than the 2080. 

If we look into the 2 GPU-s specs it can clearly see that the 3060 should be more powerful that the 2080 .

Some specs:

RTX 2080 has 2944 cuda cores VS RTX 3060 has 3840

RTX 2080 core freq 16-1700mhz VS RTX 3060 17-1800mhz 

The only thing that the 2080 is better is 256bit bus VS 196 bit bus.

But if we compare to the 4050 , it has 96 bit bus 2560 cuda cores and same performance as the 2080.
Where Nvidia plays with performance ? Its makes no sense with these specs....

[ryan]

ineteresting, but the nanometer process and efficiency plays a role. the 3060 should be more effecient so yeah it should be doing better than a 2080. I don't know if the cuda cores in the 2080 are more effective or not but somethings at play and I also would like an answer

[Reciever]

Core frequency isnt static, its also dictated by power envelopes and thermal capacities of the systems they are installed to.

 

RTX 3060 TGP's 
TGP Frequencies
60W = 817-1282 MHz
65W = 975-1357 MHz
70W = 1050-1402 MHz
80W = 900-1425 MHz
85W = 1035-1485 MHz
90W = 1163-1530 MHz
95W = 1215-1567 MHz
100W = 1267-1605 MHz
105W = 1305-1642 MHz
110W = 1342-1680 MHz
115W = 1387-1702 MHz

 

It wouldnt surprise me to see many 3060 based systems not adequately cooled either since its unlikely that you would see them installed into hefty systems that wouldnt have much issue cooling them with some attention to detail.

 

For the 4050 I am not seeing that it performs similarly to the 2080, but I'm just looking at NBC before work and havent compared the 2 in a meaningful way before.

[Csupati]

1 hour ago, Reciever said:

Core frequency isnt static, its also dictated by power envelopes and thermal capacities of the systems they are installed to.

 

RTX 3060 TGP's 
TGP Frequencies
60W = 817-1282 MHz
65W = 975-1357 MHz
70W = 1050-1402 MHz
80W = 900-1425 MHz
85W = 1035-1485 MHz
90W = 1163-1530 MHz
95W = 1215-1567 MHz
100W = 1267-1605 MHz
105W = 1305-1642 MHz
110W = 1342-1680 MHz
115W = 1387-1702 MHz

 

It wouldnt surprise me to see many 3060 based systems not adequately cooled either since its unlikely that you would see them installed into hefty systems that wouldnt have much issue cooling them with some attention to detail.

 

For the 4050 I am not seeing that it performs similarly to the 2080, but I'm just looking at NBC before work and havent compared the 2 in a meaningful way before.

Yep I know, the G15 has a 115+15w 3060 its boosting to 1900-2000mhz on core.

Its doing 65C at full load 

 

1 hour ago, Reciever said:

For the 4050 I am not seeing that it performs similarly to the 2080, but I'm just looking at NBC before work and havent compared the 2 in a meaningful way before.

Average Gaming NVIDIA GeForce RTX 2080 Mobile → 100%

Average Gaming NVIDIA GeForce RTX 4050 Laptop GPU → 95%

Source

[Reciever]

1 hour ago, Csupati said:

Yep I know, the G15 has a 115+15w 3060 its boosting to 1900-2000mhz on core.

Its doing 65C at full load 

 

Average Gaming NVIDIA GeForce RTX 2080 Mobile → 100%

Average Gaming NVIDIA GeForce RTX 4050 Laptop GPU → 95%

Source

Posts merged, please do not double post. Thx! 

Unless I am misreading, your 3060 is pulling 750w? Impressive! lol After looking at more data it does look like the 4050 is punching up to the 2080 in some cases beating it in 1080p or lower, and losing out at larger resolutions though thats hardly surprising considering the power envelop and memory bus. In those circumstances it seems the performance hit can be anywhere from 5% to 50%, while in titles where it has superior performance can be up to 15% Just depends on what you intend to play at the time.

 

 

[Csupati]

3 hours ago, Reciever said:

Posts merged, please do not double post. Thx! 

Unless I am misreading, your 3060 is pulling 750w? Impressive! lol After looking at more data it does look like the 4050 is punching up to the 2080 in some cases beating it in 1080p or lower, and losing out at larger resolutions though thats hardly surprising considering the power envelop and memory bus. In those circumstances it seems the performance hit can be anywhere from 5% to 50%, while in titles where it has superior performance can be up to 15% Just depends on what you intend to play at the time.

 

 

Sorry for doubling!
Yes it has that range . But the original concept was that why the 3060 weaker than the 2080 if it has more shading units and higher core freq.

[Reciever]

1 hour ago, Csupati said:

Sorry for doubling!
Yes it has that range . But the original concept was that why the 3060 weaker than the 2080 if it has more shading units and higher core freq.

I would surmise power envelope and thermals for individual systems. Also, its worth noting that they arent the same architecture either so not all things are equal when looking at pure specifications (Turing vs Ampere iirc)

[KING19]

A full wattage RTX 3060 has similar performance to a RTX 2070 Super depending on the wattage.

 

https://www.notebookcheck.net/NVIDIA-GeForce-RTX-3060-Mobile-GPU-Benchmarks-and-Specs.497453.0.html

 

Also a RTX 2080 uses 150w and can be boosted up to 200w with a VBIOS flash if possible. The bit bus matters a lot because no matter if a RTX 3060 has higher cuda cores and clock speeds the lower bit bus is going to slow it down.

[Csupati]

12 hours ago, KING19 said:

The bit bus matters a lot because no matter if a RTX 3060 has higher cuda cores and clock speeds the lower bit bus is going to slow it down.

Ok, but as I mentioned, The RTX 4050 has 96bit bus, 2560 cuda cores, and its capable to do same performance as the 2080. So theres no logic ....

[ryan]

haha, man. its nvidia. im sure theres a good reason(cough money) but I dont see why the 3060 doesnt have a 256bit bus im sure theres a reason but I think thats the culprit

[Reciever]

Its not the same architecture, comparing the 2 as if they are 1:1 is one of the reasons it doesnt add up.

[KING19]

On 12/9/2023 at 5:50 AM, Csupati said:

Ok, but as I mentioned, The RTX 4050 has 96bit bus, 2560 cuda cores, and its capable to do same performance as the 2080. So theres no logic ....

 

It cant because a RTX 4050 is weaker than a RTX 3060 in raw performance, about 5-10% on 1080p but and gap widen on 1440p because it has 96bit bus but it can beat it if a game has Frame Gen which are only available for 40 series GPUs. A RTX 2080 is still stronger than both GPUs

[ryan]

I still think its not a big deal being slower than a laptop 2080. the 3060 laptop is more or less equal to a desktop 3060 and the fact you can get thin and light laptops with decent temps and 3060 desktop performance at a fair price its a good thing. the 4090 on the other hand costs an arm and leg to get in your hands. the cheapest 4090 laptop ive seen in canada is over 4000. and now with geforce now being efficient enough to use i think eventually these dedicated systems are dying a slow death.

[Reciever]

2 hours ago, ryan said:

I still think its not a big deal being slower than a laptop 2080. the 3060 laptop is more or less equal to a desktop 3060 and the fact you can get thin and light laptops with decent temps and 3060 desktop performance at a fair price its a good thing. the 4090 on the other hand costs an arm and leg to get in your hands. the cheapest 4090 laptop ive seen in canada is over 4000. and now with geforce now being efficient enough to use i think eventually these dedicated systems are dying a slow death.

Until they can get GPU streaming to be faster than response times on the monitor for an affordable price, that aint happening.

[ryan]

it already is! geforce now was sold out not long ago.

 

when a monitor has 0.2ms vs 20ms on paper its like 100 times faster but in use you are not going to notice it. once you hit 20-30ms its not noticeable but you will notice it at around 40ms up. its similar to the 60hz vs 120hz argument. some will notice, most will not.

[Reciever]

1 hour ago, ryan said:

it already is! geforce now was sold out not long ago.

 

when a monitor has 0.2ms vs 20ms on paper its like 100 times faster but in use you are not going to notice it. once you hit 20-30ms its not noticeable but you will notice it at around 40ms up. its similar to the 60hz vs 120hz argument. some will notice, most will not.

most dont notice it until they experience it. Now I use 144hz for all monitors save for my 40" HDTV. Some will never know that 144hz when dont correctly resolves headaches and nausea as the mind detects incongruity with whats on display and interaction of thereof.

[ryan]

Yeah I know Linus did a video on it and the results were they noticed it. I personally don't notice or care once latency hits 30ms but some people are sensitive to it so I understand

[ryan]

well if you compare the max q varients the 2080 super max q is similar to a 3060 so theres that. if your comparing a 2080 desktop to a laptop part often times the desktop is bigger and better, the TDP also plays a role is thermals and performance from pascal gen and up