News Nvidia's new trimmed-down entry-level gaming GPU fails to outperform rivals — RTX 3050 6GB lags behind five-year-old GTX 1660 Ti

[Admin]

In its price bracket, Nvidia's new entry-level RTX 3050 6GB was reviewed and benchmarked by TechPowerUp and found extremely lackluster performance compared to other GPUs. Cards like the RX 6600 and A750 cost just $20-$30 more yet are 60% quicker.

Nvidia's new trimmed-down entry-level gaming GPU fails to outperform rivals — RTX 3050 6GB lags behind five-year-old GTX 1660 Ti : Read more

 

[AgentBirdnest]

Arc A580 - 68% faster for $10-15 less money.

 

[samopa]

Why 3050 ? What's happen with 4050 ?

 

[Neilbob]

Why 3050 ? What's happen with 4050 ?

Coming in a few weeks at a price of 299 I expect...

 

[Sleepy_Hollowed]

Wowza.

My RX580 is almost as good and perhaps better on games with demanding VRAM?!?!

Talk about nope.

 

[stonecarver]

Is this not the same marketing move Nvidia did when the RTX 2000's hit the stores. They went from the 1080's to the 2070/ 2080's and filled in the nitch gap with the GTX 1660's

Back than the 1660 was the new 1070 but less powerful.
So is the 3050 6Gb the new 1660 and again less powerful.

 

[King_V]

Arc A580 - 68% faster for $10-15 less money.

True, definitely a much better performance bang-per-buck.

But 164% greater power consumption to get that 68% better performance.

 

[Joseph_138]

Arc A580 - 68% faster for $10-15 less money.

Arc A580 also requires two 8-pin power connectors. The whole purpose of the 6gb 3050 is to put in systems that may not have any supplemental power connectors, like low end prebuilts, that may only come with a 300W PSU.

 

[Joseph_138]

1660 Ti is not a fair comparison for this card, as that was the top the range GTX 16 card, and also requires supplental power. You should be comparing it to the vanilla 1650.

 

[Joseph_138]

Is this not the same marketing move Nvidia did when the RTX 2000's hit the stores. They went from the 1080's to the 2070/ 2080's and filled in the nitch gap with the GTX 1660's

Back than the 1660 was the new 1070 but less powerful.
So is the 3050 6Gb the new 1660 and again less powerful.

No, it's the new 1650, only more powerful.

 

[bit_user]

TechPowerUp reviewed Nvidia's new entry-level GeForce RTX 3050 6GB and found its performance to be good and bad, depending on how you view it.

Wow... so Toms is reduced to reporting on other site's reviews? C'mon, guys!

 

[bit_user]

Why 3050 ? What's happen with 4050 ?

My theory is that it got renamed to the RTX 4060. Nvidia couldn't justify calling that a x50 card, based on the price increase, so they just shifted the names.

 

[Notton]

There is an RTX 4050.
It's a laptop part using the same AD107 chip as the desktop 4060, but cut down to 96bit memory bus, 2560 shaders, 80TMUs, 48 ROPs, and way less core clocks.

I think 4050m might be faster than the 3050 6GB, even when power and thermal restricted inside a laptop.

 

[Vardant]

For the best 70W TDP card, there is the RTX 4000 SFF Ada Generation

 

[TerryLaze]

True, definitely a much better performance bang-per-buck.

But 164% greater power consumption to get that 68% better performance.

RX 7000-series GPUs fixed idle power draw this december so one year after launch, and AMD has a lot of experience with high end gpus...
So it might get fixed?

Realistically we should look at the max power draw, look what GPUs are close to the power draw of the intel GPUs and logically, if drivers ever get fixed, they should end up in the same region as those in the other categories.
They all use the same node after all.
But until it is fixed it is what it is.

 

[bit_user]

RX 7000-series GPUs fixed idle power draw this december so one year after launch, and AMD has a lot of experience with high end gpus...
So it might get fixed?

What that AMD issue was concerning is idle power draw. In this area, Intel dGPUs are also pretty bad, and what I've read is that this is largely due a hardware issue of some sort. They already delivered one software improvement on that and the word was that no more improvements would be possible, in the current generation hardware.

However, what @King_V is talking about is active power usage, and here's where Intel's Alchemist is really behind the pack. In general, it gets much worse perf/W than its competitors.

The best hope we can have is for more continued performance improvements, in newer driver revisions. However, you have to ask how much room is left, and whether the driver team is even still working on Alchemist, with Battlemage set to launch later this year.

Realistically we should look at the max power draw,

Oh, okay. So, max power draw when it's Intel dGPUs and typical power draw when it's Intel CPUs? Nice try.

They all use the same node after all.

The don't. Intel's Alchemist is on TSMC N6. Nvidia's RTX 4000 is all on TSMC 4N. AMD's RX 7000 is on TSMC N5 (GCD), except that the RX 7600 is still on N6.

So, you're carrying water for Intel GPUs now, too? That's a change.

 

[TerryLaze]

Oh, okay. So, max power draw when it's Intel dGPUs and typical power draw when it's Intel CPUs? Nice try..

What is this even supposed to mean, I mean besides the obvious pure flamebaiting, inside your mind did it have some reasoning?

Since the GPUs are made on TSMC and are impossible to run at higher power than default, then the ARC GPUs that have similar max power than other GPUs, made on the same node as them, should also have similar power draw as them at other tasks.

 

[bit_user]

Since the GPUs are made on TSMC and are impossible to run at higher power than default, then the ARC GPUs that have similar max power than other GPUs, made on the same node as them, should also have similar power draw as them at other tasks.

This statement depends on several invalid assumptions.

I doubt there's anyone here who would, but I'd strongly caution anyone against buying an Alchemist dGPU, under the assumption that any amount of driver updates are going to significantly close the efficiency gap vs. AMD or Nvidia. Even if it's possible, which I seriously doubt, Intel is quite likely focusing their driver improvements on Meteor Lake's iGPU and working on support for upcoming dGPUs and iGPUs.

BTW, Intel had some recent layoffs which affected their GPU software team.

https://www.tomshardware.com/tech-i...-totalling-700-layoffs-in-the-state-this-year
​

 

[George³]

but I'd strongly caution anyone against buying an Alchemist dGPU

You didn't support 3rd GPU player to compete with duopoly of AMD/Nvidia?

 

[bit_user]

You didn't support 3rd GPU player to compete with duopoly of AMD/Nvidia?

The part you edited out of my statement was a key qualifier.

Basically, I was saying that you should be okay with their current efficiency, if you choose to buy one.

Personally, I'm interested in Intel dGPUs. However, I have decided to sit this generation out. I was hoping Alchemist+ (i.e. Alchemist refresh) was going to fix enough of the issues to make it more compelling. Since that seems to have been cancelled, I guess I'll be waiting for Battlemage.

 

[Pierce2623]

Arc A580 - 68% faster for $10-15 less money.

You’re literally comparing a 200w card to a 75w card. I can get the nvidia in a true low profile for $175 whereas the previous best was the rtx a2000. It actually performs like a standard 3050 but it cost $300 minimum, too. A reasonably powerful low profile card will have zero problems finding a niche. The low profile models have already sold out.

 

[Pierce2623]

This statement depends on several invalid assumptions.

I doubt there's anyone here who would, but I'd strongly caution anyone against buying an Alchemist dGPU, under the assumption that any amount of driver updates are going to significantly close the efficiency gap vs. AMD or Nvidia. Even if it's possible, which I seriously doubt, Intel is quite likely focusing their driver improvements on Meteor Lake's iGPU and working on support for upcoming dGPUs and iGPUs.

BTW, Intel had some recent layoffs which affected their GPU software team.

​

https://www.tomshardware.com/tech-i...-totalling-700-layoffs-in-the-state-this-year​

​

I agree 100%. The drivers aren’t going to greatly improve efficiency. There’s something about Alchemist which keeps it from scaling properly to higher shader counts. That’s why their lower low midrange, upper low midrange, and true midrange cards all have extremely similar power draw and performance even though the a780 has 70% more fp32 lanes than the a580.

 

[tgrsbrks]

The 4060 really is where the 4050 should have been and the 3050 refresh with 6gb is a waste of time and only shows everyone how little Nvidia cares for for the market below the 4070 series GPU's.

 

[bit_user]

The 4060 really is where the 4050 should have been and the 3050 refresh with 6gb is a waste of time

The only thing I like about this refresh is that it provides a newer card that runs off only the 75 W PCIe slot power and can be low-profile. That's a huge win for compact desktop PCs, where that's all the space they have and there's no auxillary power cable for an extra 6-pin or 8-pin connector.

However, what I wish Nvidia had done is used the AD107 instead of the GA107. They could've done the same thing of cutting it down to 96-bit / 6 GB and reducing clock frequencies to fit the RTX 4000 chip into a 75 W power envelope. I guess the reason they didn't is probably they thought such a card would be too expensive for what that market is prepared to spend on it.