• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

NVIDIA GeForce RTX 5050 8 GB

@1080p RTX 5070 Ti is 19% faster than RTX 5070 but costs 51% more.
@1440p RTX 5070 Ti is 23% faster than RTX 5070 but costs 51% more.
@4k RTX 5070 Ti is 27% faster than RTX 5070 but costs 51% more.

RTX 5070 Ti is not at the same level definattley not.

?


Screenshot 2025-07-06 at 10.57.42 AM.png
 
They can definitely get it to work, It's just performance would end up in a completely different tier of cards. Kinda like how a Mobile 5080 with 60 SM units ends up closer to a 36 SM desktop 5060 TI.



5070TI is also quite cost effective at MSRP and you're more or less getting more SM units per dollar relative to 5070. Card OC's extremely well.

If you don't OC, it has the highest efficiency on average. Better 4K card in regards to FPS/$.. esp since not being bottlenecked at 12 gig.

I have the opinion that both 70 variants effectively cover a wide range of gamers and exploit blackwell arch effectively per SM count.

5060 TI variants have obvious tradeoffs and down sides.
5080 isn't price effective, especially at 16G. Too close to 5070 TI.
5060 is "okay" but they really need a 12GB 3GB IC variant at a similar $300-330 price point.
5050 has the worse SM/$ ratio, but $200-230 would more or less balance it out with other NV cards. $200 price point would top effective 1080p performance in frame/USD @ 1080p.

You buy 5090 because you want the best.. no one is buying this card to be cost effective.
I implied a similar thing a few days ago.


It got laughs from the usual Nvidia fan suspects who don’t know how to measure a good buy due to how much they love Nvidia. For the rest of us, it is apparent that the best buys are the 060 and 070 cards from both AMD and nvidia. However even these are still a bit overpriced.
 
I'm looking at real prices right now!

RTX 5070 - 550€
RTX 5070 Ti - 830€

You can get a 5070 TI for $750 in the USA via BB PNY drops every few weeks. It's quite easy.

Cant comment on Europe.

"5070TI is also quite cost effective at MSRP"
 
Last edited:
You can get a 5070 TI for $750 in the USA via BB PNY drops every few weeks. It's quite easy.

Cant comment on Europe.
Looking at Newegg right now, the cheapest 5070 Ti is $840 and the cheapest 9070XT is $700. That $700 9070XT is from Asrock and its the best selling GPU right now on Newegg.
 
Looking at Newegg right now, the cheapest 5070 Ti is $840 and the cheapest 9070XT is $700. That $700 9070XT is from Asrock and its the best selling GPU right now on Newegg.
I mean.. I did say MSRP. $800+ isn't MSRP.

IE: 2 days ago.

These drop every 2 or so weeks consistently. I was able to get the PNY 5070TI and a ASUS prime 9070XT Prime for $599 via newegg the same june 6th.

I have no trackers or anything, but I did have Newegg notification when the 9070XT was stocked.
 
These drop every 2 or so weeks consistently. I was able to get the PNY 5070TI and a ASUS prime 9070XT Prime for 599 via newegg the same june 6th.
That's 25% more expensive for the same fps!
 
That's 25% more expensive for the same fps!

It's not that black and white.

NVIDIA tends to do better in CPU bound esports stuff in conjunction with my 12900K. Example: I get 100 FPS lower on the 9070XT when both are set to 1440p low via CS2. AMD's driver/scheulding just seems terrible unless youre on the latest X3D... This is also with the latest AMD driver.

The 9070XT is VERY good at raster though.. theres no difference in fps between the two cards in games like CP77. The biggest difference is honestly the upscaling.. FSR3 limitation blows in this game. but DLSS4 works a lot better.

Both cards have pros and cons.. I wouldn't recommend AMD if you play esports stuff.. NVIDIA tends to work better on most setups not using a 9800X3D. @W1zzard s CS2 benchmarks also kind of reflect this with cards like B580 and 9060XT next to a 2 gen old 3060 12..
 
There are many tests of gaming performance.
There are few tests of encoders and encoding quality.
Decoder tests simply do not exist.
1. I don't see any gpu video encoding tests / reviews on TPU.
2. A decoding test would include: test what codecs including frame rate and bit rate the gpu can decode. Test how many concurrent video decodes the gpu can do of 4k30fps, 4k60fps, and 8k videos.

This 5050 has the same nvenc/nvdec chip as the 5090 except the 5090 has 3 nvenc and 2 nvdec chips compared to this gpu being 1/1. It really seems like it is one of the major selling points of the gpu. In my experience, it takes 8c/16t zen5 cores to saturate just one nvenc encoder chip.

If we can get commercial grade ssd reviews I don't see why we can't get a review of the different gpu vendors encoding performances.
 
Last edited:
I implied a similar thing a few days ago.


It got laughs from the usual Nvidia fan suspects who don’t know how to measure a good buy due to how much they love Nvidia. For the rest of us, it is apparent that the best buys are the 060 and 070 cards from both AMD and nvidia. However even these are still a bit overpriced.

I agree with you, but ultimately this turns out to be true every single generation. The midrange and performance segment models have the highest return on investment and the highest performance per dollar.
 
I agree with you, but ultimately this turns out to be true every single generation. The midrange and performance segment models have the highest return on investment and the highest performance per dollar.

I don't know if the base 5060 would be considered midrange or low end, but it would actually end up fairly decent if they bumped VRAM to a 3GB IC without changing pricing. Has the highest SM/$ ratio in the current Blackwell lineup. 5070 TI is second to this card.

I personally think the 16GB 5060TI isn't that great at current MSRP. Treading into 5070 price territory, which ends up more versatile across all resolutions. Super variants will offset this more.

----

Either way, we can all agree that the 5050 is too close to the 5060 per MSRP and VRAM config. +10SM, higher TDP, and better bandwidth to compete more effectively at 1440p is worth the extra $50 for most people.

RTX5050 really needs to drop to a $200-230 range, would also effectively make it the strongest performance per dollar 1080p card. That is, assuming my math is right.
 
Last edited:
RTX5050 really needs to drop to a $200-230 range, would also effectively make it the strongest performance per dollar 1080p card. That is, assuming my math is right.
The review does the math for you.


At $200-$225, this card would have the best perf/$ as long as all other cards stay at current pricing.
 
I don't know if the base 5060 would be considered midrange or low end, but it would actually end up fairly decent if they bumped VRAM to a 3GB IC without changing pricing. Has the highest SM/$ ratio in the current Blackwell lineup. 5070 TI is second to this card.

I personally think the 16GB 5060TI isn't that great at current MSRP. Treading into 5070 price territory, which ends up more versatile across all resolutions. Super variants will offset this more.

----

Either way, we can all agree that the 5050 is too close to the 5060 per MSRP and VRAM config. +10SM, higher TDP, and better bandwidth to compete more effectively at 1440p is worth the extra $50 for most people.

RTX5050 really needs to drop to a $200-230 range, would also effectively make it the strongest performance per dollar 1080p card. That is, assuming my math is right.

Base 5060 is already well within the realm of low end, IMO. The 5060 Ti (16 GB model at least) barely qualifies as a midranger in my eyes. It has the performance and isn't at the minimum benchmark for memory capacity.
 
I think he meant that they are not enough to show the whole picture.
No, he specifically said...
Even if the current benchmarks lie.
..that.
If so I agree.
Rubbish and nonsense. The benchmarks performed test a very wide range of capabilities. If you think there's something being missed, by all means, do explain.

They can definitely get it to work, It's just performance would end up in a completely different tier of cards.
True, but then again, the 3050 8GB and the slot powered 3050 6GB were in different performance tiers. And that's ok. Still want one because it would be a massive upgrade from what most are running now and would be a compelling upgrade from a 3050. So again, it's all good, bring it on I say.

See, as opposed to the naysayers, someone of us have understanding of proper perspective and scope(not a jab at you jaszy). I'm not saying the 5050 will blow ANYONES doors off compared to the rest of the RTX5000 lineup. However, it will blow the doors off previous offerings 2 or more generations back. Context and perspective are important.
 
Last edited:
The review does the math for you.


At $200-$225, this card would have the best perf/$ as long as all other cards stay at current pricing.

Oh.. I usually ignore market price and scroll down to MSRP. lol
 
This thing might have stood a chance if they'd reduced the power consumption to go with the reduced core count. The 4060 was a massive disappointment with less VRAM and barely more performance than the 3060 it replaced - yet the ONE redeeming feature was its low power draw at just 115W.

As it stands, the 5050 is both 20W hungrier than the 4060 and it's slower. What is the point? This thing doesn't even get GDDR7 so it's bandwidth-choked, too. The only reason it has a place in the market at all is because Nvidia killed 4060 production several months ago, making an artificial hole for this inferior replacement.

Should have been cheaper without GDDR7.
Should have been more efficient with fewer shaders than the 115W 4060
Should have served the slot-powered market segment as the slowest/cheapest SKU in the 50-series lineup.
 
Oh.. I usually ignore market price and scroll down to MSRP. lol

Ignoring market price is unfortunately ignoring what the consumer actually pays as MSRP is a problem when no cards are selling for it. I'd be an RTX 3080 owner if it sold for MSRP even 18 months after release but when I bought something else a few months after that, the 3080 was still at 1.5-2x MSRP.

Market conditions are everything when you actually pay out, MSRP is irrelevant.
 
Last edited:
Ignoring market price is unfortunately ignoring what the consumer actually pays as MSRP is a problem when no cards are selling for it. I'd be an RTX 3080 owner if it sold for MSRP even 18 months after release but when I bought something else a few months after that, the 3080 was still at 1.5-2x MSRP.

Market conditions are everything when you actually pay out, MSRP is irrelevant.

Yet, I got both a 5070TI and 9070XT at retail US MSRP... Sure it was 3-6 months post launch, but... It's not like I'm sitting at my PC 24/7 or using trackers.

Literally just eshop notifications and browsing reddit.
 
Last edited:
Because it is. 1080p has roughly twice the pixel count of 720p, but both are well within the capabilities of budget hardware today.
The difference in VRAM usage from 720p to 1080p is ridiculous compared to the one notch increase in texture quality.
There are many tests that demonstrate this also done on this site, try to understand the context, before badly defending someone else's post...

Textures do not decrease the amount occupied in relation to resolution, and since it is the parameter that occupies the most, usually deservedly so... it is not clear why we always have to talk only about resolution when mentioning the small amount of VRAM.

You obviously think that the 3060 can never surpass the 3060 Ti in the texture parameter? You said never, when there are cases in which even the 3080 drops dramatically thanks to the lower amount of VRAM. So one case would be enough to refute you, and from 12GB to 8GB there are many, and it shouldn't even be necessary to underline them, if you know how it ends when the VRAM is saturated.

There is a 10% difference between the 5050 and the 3060, it is insane to talk about "A LOT" (even in capital letters), while minimizing the +50% of VRAM present in the 3060.

It seems that you limited yourself to reading halfway, without understanding the thing that was quite clear: When you need more VRAM.
On the 3060 you just need to lower the shadow parameter from ultra to high to get the same frame rate as the 5050, in some cases you even gain.
What is better to keep higher? What improves the overall performance of the upscaler, thanks to the source with superior textures?

You are carrying forward short-sighted stereotypes that have aged badly, increasing the quality of the textures weighs almost ZERO, when you can afford it, you are treating it as a setting that weighs even more.

With the 5050 you are tight to start with, it is not difficult to predict that in the future the situation will only get worse, because it has always been like this.

I didn't understand what you were going to answer about Stellar Blade. With 8Gb you can't set the texture parameter to the maximum, which is what I was talking about.

"Tight" benchmarks are not made to highlight the cons of having a lower quantity, I said it in closing, among the many things you willingly skipped...

The low quantity of VRAM is a con that you can't fill with other technologies, indeed, in the case you amplify that con, since these increase the use of VRAM, so it is a con that in some way also dirties those that would otherwise be only pros.

He wasn't talking about specific and rare cases, he was framing it as if that should be the rule, instead it's already a lot if it's the exception.
The rule is that if you need VRAM when you don't get there you pay the price, and nowadays paying the price with 8GB is anything but exceptional.

Try to better understand the specific context in which I was moving, if you don't want to go from doctor to patient...
 
Last edited:
I hope to see the price settle under MSRP in coming months. Already seen multiple Blackwell SKU's selling under MSRP now that the launch hype has settled, after all SMRP is meaningless and street price is king right, so like most other items it's still region dependent where the best value picks are.

And the giant coolers make no sense, these cards could all easily be single fan or LP and stay relatively cool and quiet like the rest of the Blackwell GeForce line up.
 
Last edited by a moderator:
if by any chance some of you live in europe and have access to alza shipping (poland,czech,slovakia,austria,hungary)
than look no further - this is a banger: https://www.alza.sk/inno3d-geforce-rtx-3080-x3-oc-lhr-12-g-d7082095.htm

brand new , 2 year warranty , having the better performance than both rtx 5060Ti and rtx 4070 and has enough VRAM
something like rtx 5060Ti 8gb selling for the same (or even higher) price and 20% weaker in rasterization looks kinda ridiculous next to this card right now .
 
Last edited:
if by any chance some of you live in europe and have access to alza shipping (poland,czech,slovakia,austria,hungary)
than look no further - this is a banger: https://www.alza.sk/inno3d-geforce-rtx-3080-x3-oc-lhr-12-g-d7082095.htm

brand new , 2 year warranty , having the better performance than both rtx 5060Ti and rtx 4070 and has enough VRAM
something like rtx 5060Ti 8gb selling for the same (or even higher) price and 20% weaker in rasterization looks kinda ridiculous next to this card right now .
I you don't mind that it runs hot and is loud.
103545_h5dgkz3kovro3bgx_factory.jpg
 
I you don't mind that it runs hot and is loud.
I am not sure why we are talking about the RTX 3080. The 5050 is an entry level card, part of its market will be people with older systems upgrading from 4GB and 6GB cards. Limited power supply capacity will probably be a factor. According to TPU maximum power consumption of the 5050 is 131W and the 3080 335W.
 
1. I don't see any gpu video encoding tests / reviews on TPU.
2. A decoding test would include: test what codecs including frame rate and bit rate the gpu can decode. Test how many concurrent video decodes the gpu can do of 4k30fps, 4k60fps, and 8k videos.

This 5050 has the same nvenc/nvdec chip as the 5090 except the 5090 has 3 nvenc and 2 nvdec chips compared to this gpu being 1/1. It really seems like it is one of the major selling points of the gpu. In my experience, it takes 8c/16t zen5 cores to saturate just one nvenc encoder chip.

If we can get commercial grade ssd reviews I don't see why we can't get a review of the different gpu vendors encoding performances.

This and I would also love to see realtime streaming encoding latency and image quality comparisons.
 
I am not sure why we are talking about the RTX 3080. The 5050 is an entry level card, part of its market will be people with older systems upgrading from 4GB and 6GB cards. Limited power supply capacity will probably be a factor. According to TPU maximum power consumption of the 5050 is 131W and the 3080 335W.
Don't ask me, i'm not the one who derailed the topic.
 
Back
Top