Review Nvidia GeForce RTX 5070 review: $549 price and performance look decent on paper

[oofdragon]

This GPU ISNT EVEN ON PAR WITH THE 4070 SUPER!! You may actually buy a 5070 with missing ROPs, it's already missing 1000 cudas compared with the 4070 Super l, and it doesn't even have 32bit CUDA support lmao. This GPU is the worst product I've ever seen any company sell ever, last 30 years for real

 

[oofdragon]

Whoa is me (as someone who needs to buy a new GPU). I keep waiting for some light on the horizon, and yet between compounding tariffs, and near daily reveals of 50-series snafus and sheer mediocrity, even the used prices keep going up.

My GPU regrets, in order:
1) Not buying a brand new 4080Super for $1000 4 months ago. (currently averaging $1800 used)
2) Not buying a brand new 4090 for $1600 4 months ago (currently averaging $2500 used)
3) Not buying a brand new 7900XTX for $999 last week (currently averaging $1500 new)

I'm now staring down the potential of paying $1600+ for a 5080 once they become available when I could've gotten a superior and faster 4090 for the same price a few months ago.

First-world problems, to be sure, but I can't help but wonder what further compounding GPU mistakes I'm making right this minute by hoping things will change.

What are you regretting? 9070 XT is a better version of the 4080 for $600. Sure you can buy a 4090 for $1600 just to brag about or get the 9070XT with a OLED monitor and upgraded CPU and SSD for the same price to have exactly the same gaming experience

 

[helper800]

9070 XT is a better version of the 4080 for $600.

Thats some hopium if I have ever seen it. We shall see if the 9070 XT is as fast as a 4080 in the days to come, saying it is now is purely divining the future.

 

[Notton]

and as a reminder, this card doesn't have PhysX or 32-bit compute.

 

[helper800]

and as a reminder, this card doesn't have PhysX or 32-bit compute.

Or good drivers, or 100% of its ROPs on average, or stock. It's a 5060 in everything but name.

Nvidia really did the 4000 series crap with the "4080 12gb" all over again but for the entire 5000 series product stack. One tier lower performance at one tier higher cost compared to the tier of card they are named.

5070 = 60 tier card & at a 70 tier cost
5070 ti = 70 tier card & at a 80 tier cost
5080 = 70 ti tier card & at a 80 ti tier cost
5090 = 80 ti tier card & at an AI tier cost

 

[palladin9479]

Thankfully I am not in the market for a GPU right now. My RX 6800 is still sufficient for my needs. The whole market is a mess, but this card is still meh.

I'm ridiculously happy I bought my 7900 XTX for $850 a day after the 5080 embargo lifted. Now they are going for over $1000 if you can even find them in stock. With nVidia's supply being non-existent it's really up to AMD and possibly Intel to supply that demand. AMD seems to at least trying to be kind with it's pricing, now just to see the benchmarks and what the initial supply looks like.

 

[palladin9479]

5090 = 80 ti tier card & at an AI tier cost

This is fundamentally wrong, the 5090 is almost a datacenter GPU in size and power, far beyond anything they've offered before. It is ridiculously large and requires more power then the 12VHPWR connector can provide.

The rest is highly debatable because nVidia has subdivided to much they'll need an HOA soon. 5080 is definitely an 80 tier card the same as 4080 and 3080. 70 and below is wishy washy.

The issue is you, and most everyone else, was expecting between 20 and 100+ percent increase in generation to generation performance. People actually believed the marketing hype. Back here in reality, we've long since hit the limit for large architectural improvements in rasterization / raw FPS. Going forward it's going to be small architectural improvements with the gains mostly coming from fix function feature sets like RT / DLSS / Encoding / AI / etc.

4060 8GB vs 3060 12GB
https://www.tomshardware.com/pc-components/gpus/rtx-4060-vs-rtx-3060-12gb-gpu-faceoff

4070 Reviews vs 3070 and the rest
https://www.tomshardware.com/reviews/nvidia-geforce-rtx-4070-review/4

So we do see generational improvement that shows the 70 class is still 70 class and the 60 class still is the 60 class (mostly). What is bumming people out is not the MSRP but that the MSRP is completely fantasy and street pricing is insane due to low channel supply.

 

[Giroro]

RTX 5070 is great news for RTX 4070 Ti owners!

They couldn't even improve performance to the next highest card in their old stack. That's not 4 star performance, in my book. 4070 TI performance at 4070 MSRP would be the dictionary definition of 3 stars.

But considering AMD's performance is already known (but not not published) - it was probably taken into account when the decision was made to give this card such a high score.

People need to stop losing their minds over every stupid pointless product launch. I don't even want these cards at all, but the panic buying is getting super annoying.

If it's a choice between a new GPU and a used car: BUY THE OLD CAR, DUMMY. THEY HOLD VALUE BETTER, AND YOU CAN'T DRIVE A GPU TO WORK.
Worst case, taking that car to a demolition derby is going to be way more fun than modern gaming.

 

[shady28]

I would say the entire 30-series in late 2020 throughout 2021 was, so far, worse than what we've had from the 50-series. RTX 3080 selling for $2000–$2500? RTX 3090 going for up to $4500? Yeah. And you know what? None of that was the fault of Nvidia or AMD.

The current supply restrictions are much more in Nvidia's control, because it's deciding to prioritize AI over consumer. But I can't fault a company for choosing to do more of the thing that accounted for 88% of its revenue last year.

Is four stars too high? 🤷‍♂️ That's based on the theoretical MSRP, because GOK what the actual prices are going to be throughout 2025! On paper, everything looks decent. In practice, everything is fubar — and I mean that about all GPUs right now. So writing emotionally vapid comments blaming Nvidia for lack of stock just isn't something I'm going to bother doing. Yes, the supply situation sucks right now. Prices suck right now. You can't buy these at $549 right now (unless you win the lottery). But if you could buy one at that price? Sure, it's a 4-star card, maybe 3.5-star. And getting bent out of shape about a half a star difference of opinion isn't worth the effort.

Put another way: Read the review, look at all the pretty charts, decide for yourself how good/bad/whatever the card is. But don't get hung up on one number that tries (and always fails) to encapsulate way too much information.

I don't know a way out for reviewers, but I disagree that Nvidia and AMD have no blame here.

Any manufacturer should do a market analysis before they set the MSRP. That analysis means they do a supply / demand curve to determine the *correct market price* for their product. I guarantee you that both AMD and Nvidia have done this. They know more or less where these cards will *actually* sell at based on how much supply they provide.

But they are not setting the MSRP to that number, which pads the pockets of AIBs and retailers who later mark up the price well beyond MSRP. This is a trick that they seem to have 'learned' during the Covid year(s). They don't want to make a lot of these cards with limited TSMC allocation, so they are setting the MSRP artificially low and will then talk about unprecedented demand for their cards.

It's all basically a big lie, that pads the pockets of AIBs and retailers who they don't want to see leave them in the event there is a future AI market implosion. Think back to the implosion of network infra providers during .com, the likes of 3com, Cisco, Novell. It's likely to happen. They know this.

This is their hedge.

 

[byteme1]

How is this decent? Just wow , look at the stripping down and gen on gen uplift this is more like a 5060 , also shouldn't be 12gb at this price point now

 

[Jabberwocky79]

Can you run it on air until a block appears on the markets for whatever GPU you get, or are you building in an SFF that cannot fit any of the air cooled cards?

I can definitely run it on air, but: 1) I have every single other component needed to finish a custom loop, so I'd like to go ahead and finish the project, and 2) Not every iteration of every GPU has (or will have) a water block solution, so it's a gamble if I plunk money down on a certain card and no water cooled solutions come to market for it. Generally you have to enter the enthusiast-tier cards to find options, so who knows about the 9070.

 

[ThisTechie]

First time commenting on a Tom's Hardware article. I will say, I've always thought of Tom's Hardware as a trustworthy, very thorough review site, and I make extensive use of their hierarchy charts and benchmarks for my own research and for recommending hardware to my friends and family.

But despite my very positive perception of this site's content, the exclusion of the 4070 Super from this review has me genuinely worried about the quality of journalism here. Tom's Hardware is not the only reviewer I've seen that has skipped the 4070S, but it is the only reviewer I've seen where the absence of the 4070S seems to have artificially boosted the 5070's alleged value proposition.

I understand there is a lot of work that goes into redoing hundreds of benchmarks for a new test bench setup, but the end result of skipping the 4070S on the excuse that its "not too far off the performance of the RTX 4070 Ti" is shaky at best when it results in dubious claims of "decent generational performance increase" (which is the first Pro listed on the first page of this review!!!), a claim which I have not seen any other third party reviewer support when compared to the 4070S.

 

[Jabberwocky79]

What are you regretting? 9070 XT is a better version of the 4080 for $600. Sure you can buy a 4090 for $1600 just to brag about or get the 9070XT with a OLED monitor and upgraded CPU and SSD for the same price to have exactly the same gaming experience

As already mentioned, that's a pretty big assumption about the 9070XT.

I won't be buying any card for "bragging rights." I have certain needs based on my gaming preferences AND work.

While the overall value of the 9070XT seems promising, it won't necessarily be the level of performance I am looking for.

 

[King_V]

Not to mention the 50 series is probably the wrost GPU launch ever.

Spoiler: This seems like it fits perfectly here:

 

[atomicWAR]

The word disappointing comes to mind first and foremost. This is basically a 4070 Super with MFG at a 50 dollar 'discount' assuming you hit the lottery and are able to buy one at msrp. Nvidia needs to start giving its non-flagship cards more VRAM and performance. The gap is just so large and growing between 90 and 80 class it doesn't leave enough room for gen on gen up lift in its lower tier cards. Personally I'd like to see Nvidia hobble AI or compute features like they did in the Titan years on the lower down skus while preserving close to flagship raster performance as much as possible. I understand some of this is upsell tactics. But things are getting stoopid at this point. End of day the 50 series just doesn't seem worth it in most situations, even IF you can get something at MSRP. I'll be curious to see how the RX 9070/9070 XT pan out. AMD really has an opening for a change. I hope they can excute a solid mid range coup and not bungle it...my fingers are crossed.

Great review @JarredWaltonGPU though I do have a singular complaint. The lack of the 4070 Super in the charts or at least heavily mentioned in text. I absolutely understand it is easy to extrapolate 4070 Super numbers from the included cards but considering just how close or better the 4070 Super is to a 5070 in some games, it feels like an oversight that could be perceived as neferious to hide how weak the uplift in performance is gen on gen. To be clear I do not believe that is your reasoning as a regular reader. I know it's just a lot of extra work for an extra card but I do think it easier to nip some things in the bud so to speak. Heck I even saw some posts eluding as such which is typical. Everything else looks great though, information wise. Thank you for all your hard work!

 

[helper800]

First time commenting on a Tom's Hardware article. I will say, I've always thought of Tom's Hardware as a trustworthy, very thorough review site, and I make extensive use of their hierarchy charts and benchmarks for my own research and for recommending hardware to my friends and family.

But despite my very positive perception of this site's content, the exclusion of the 4070 Super from this review has me genuinely worried about the quality of journalism here. Tom's Hardware is not the only reviewer I've seen that has skipped the 4070S, but it is the only reviewer I've seen where the absence of the 4070S seems to have artificially boosted the 5070's alleged value proposition.

I understand there is a lot of work that goes into redoing hundreds of benchmarks for a new test bench setup, but the end result of skipping the 4070S on the excuse that its "not too far off the performance of the RTX 4070 Ti" is shaky at best when it results in dubious claims of "decent generational performance increase" (which is the first Pro listed on the first page of this review!!!), a claim which I have not seen any other third party reviewer support when compared to the 4070S.

Your mistake was not making the correct comparisons. The 4070 is directly compared to the 5070 because both of them have the same name in different generations. Why would he compare generational performance of the 5070 to the 4070S? He is not reviewing a 5070S...

 

[Giroro]

I don't know a way out for reviewers, but I disagree that Nvidia and AMD have no blame here.

Any manufacturer should do a market analysis before they set the MSRP. That analysis means they do a supply / demand curve to determine the *correct market price* for their product. I guarantee you that both AMD and Nvidia have done this. They know more or less where these cards will *actually* sell at based on how much supply they provide.

But they are not setting the MSRP to that number, which pads the pockets of AIBs and retailers who later mark up the price well beyond MSRP. This is a trick that they seem to have 'learned' during the Covid year(s). They don't want to make a lot of these cards with limited TSMC allocation, so they are setting the MSRP artificially low and will then talk about unprecedented demand for their cards.

It's all basically a big lie, that pads the pockets of AIBs and retailers who they don't want to see leave them in the event there is a future AI market implosion. Think back to the implosion of network infra providers during .com, the likes of 3com, Cisco, Novell. It's likely to happen. They know this.

This is their hedge.

I think it's less about padding the pockets of the AIB partners, and more about shifting blame, while Nvidia pads their own pockets.

AIB partners are forced by Nvidia to pay such high prices for the chips that they can't really make any money at MSRP, Nvidia wants to keep all the money for themselves.
Even when stock is good, AIBs have no choice but to charge over MSRP, so consumers blame the AIB for high prices. Nvidia sits back to enjoy good publicity for releasing a "good value" product that is in "high demand" (relative to their nonexistent supply). Nvidia knows high prices are inevitable, they want prices high, and they engineered the whole situation. I would go one further and say I think Nvidia intentionally wants customers to balk at the high prices. It's like Rolex or Rolls Royce advertising to people they know will never, ever be able to afford their pointless, overpriced luxury products. Elite luxury is all about brand perception. This paper launch isn't about selling $1000 gaming GPU's to kids too young to understand tech influencers make their money through sales commission paid by their advertisers. It's all about generating enough buzz to make billionaires feel important when they buy enough $100,000 GPUs to fill a warehouse. So many cards that no company will be able to use them to turn a profit - not even if they were so good that they stayed on the bleeding edge of performance for the next 10 years.
At least, no profit if they try to use their AI for a consumer-facing product. I think some of these AI projects are more about trying to really good auto traders that can break/manipulate the stock market/crypto currency/world economies/etc. The first one to do that will probably make all the money in the world (maybe Nvidia's success is because they already reached that breakthrough)....

Point being, Nvidia just doesn't want customers to realize they are acting as a literal monopoly - and one of the largest companies in the history of earth... despite AI currently being a niche and and arguably inessential fad market. Nvidia wants to avoid recognition for choosing to be the root cause of evil. They want customers to blame those mean ol' AIB partners who are always bullying underdog Nvidia and telling them what to do.

At least, that's what it looks like what Nvidia hopes their perception will be. That hinges on us gross little consumers thinking that $550 is a good price for any gaming GPU - even a mediocre one. Personally I strongly disagree. New PC games just are not good enough to be worth that price of admission. Nobody cares what I think though, not with all the automated scalper bots trading these cards back and forth with each other are probably helping Nvidia with their cause. Whipping people into a panic for no good reason. Or at least they will, right up until the bubble bursts. Nothing about this system seems stable.

 

[Penzi]

Just want to make sure I got this. Compared to the 4070, there’s a 17.1% gain in ray tracing, 19.1% gain in non-RT games. So far, so decent! To achieve this gain over the 4070 they increased power usage 22.8%. This generates an additional 6.8ºC of heat and 9.9 dB(A) more noise… (?!) My goodness, nVidia engineers must have been burning the candle at both ends for the past 2+ years…

 

[palladin9479]

Just want to make sure I got this. Compared to the 4070, there’s a 17.1% gain in ray tracing, 19.1% gain in non-RT games. So far, so decent! To achieve this gain over the 4070 they increased power usage 22.8%. This generates an additional 6.8ºC of heat and 9.9 dB(A) more noise… (?!) My goodness, nVidia engineers must have been burning the candle at both ends for the past 2+ years…

Performance increase from more power is not linear unfortunately. There was a small architectural increase over last generation and they supplemented the rest by just increasing power / units. Expect this to be the norm going forward.

 

[mhmarefat]

Which, sadly, has a going price of basically $1000 or so new, or you can take your chances with eBay where prices over the past 30 days are averaging $789.55. Not that I expect the 5070 to be any better in the near term.

Wait, in 5070 Ti comments you previously said you can't predict what will happen in 2 weeks later, that in the future 5070 Ti will be available for $745 yet it seems, when it suits Nvidia, you indeed CAN predict the future: $1000 4070 will never go back to its MSRP so this makes 5070 a 4 stars GPU!

Why can you predict the future when it benefits Nvidia but cannot do so when it is against it?

So writing emotionally vapid comments blaming Nvidia for lack of stock just isn't something I'm going to bother doing.

Since the 30 series launch, you've been whitewashing Nvidia's predatory prices, blaming inflation, hard times for industry due to Covid (after the Covid era Nvidia's value was worth "only" $1 Trillion btw), etc... but you never stood up to Nvidia for it ugly greedy predatory prices except once, where you rightfully called 3090 Ti $2k launch price "a joke", and we are saying the same! The 50 series GPUs are a joke too!! Because they are not priced fairly, not even greedily, rather explosively greedily and have the worst generational uplift yet.

$550 for 3090 Ti performance would be a win for gamers if this had 16 GB VRAM. But 12 GB is DOA. If people buy 12 GB GPU for a midrange 70 series they are misinformed and are victims of Nvidia.

 

[valthuer]

3) Not buying a brand new 7900XTX for $999 last week (currently averaging $1500 new)

I'm now staring down the potential of paying $1600+ for a 5080 once they become available when I could've gotten a superior and faster 4090 for the same price a few months ago.

First-world problems, to be sure, but I can't help but wonder what further compounding GPU mistakes I'm making right this minute by hoping things will change.

As a guy who sold his 4090 for roughly 1,500$, i totally get you.

At one time, i was so frustrated with the availability of 5090, that i was seriously contemplating the purchase of either 7900XTX or 5080.

Fortunately, my 5090 arrived last week - albeit for a scalper price.

 

[valthuer]

I'm ridiculously happy I bought my 7900 XTX for $850 a day after the 5080 embargo lifted. Now they are going for over $1000 if you can even find them in stock. With nVidia's supply being non-existent it's really up to AMD and possibly Intel to supply that demand. AMD seems to at least trying to be kind with it's pricing, now just to see the benchmarks and what the initial supply looks like.

You made a good choice picking 7900XTX for $850. The cheapest one here in Greece, is currently selling for no less than $986.

 

[helper800]

This is fundamentally wrong, the 5090 is almost a datacenter GPU in size and power...


It is ridiculously large and requires more power then the 12VHPWR connector can provide.


The rest is highly debatable because nVidia has subdivided to much they'll need an HOA soon.

The above is irrelevant to my claim.

far beyond anything they've offered before.

False. They have offered farm more of the die before.

The 5090 uses 88.54% of the GB202 die.
The 4090 uses 88.9% of the AD102 die.
The 3090 Ti uses 100% of the GA102 die.
The 3090 uses 97.6% of the GA102 die.

5080 is definitely an 80 tier card the same as 4080 and 3080. 70 and below is wishy washy.

False, again they have generally offered more in 80 class cards compared to their top cards.

The 5080 has 10752 shaders or 49.4% the shaders of the 5090
The 4080 has 9728 shaders or 59.3% the shaders of the 4090
The 3080 has 8704 shaders or 80.95% the shaders of the 3090 ti
The 2080 has 2944 shaders or 63.88% the shaders of the Titan RTX
The 1080 has 10752 shaders or 66.66% the shaders of the TITAN Xp

Even using the most charitable comparisons with the old Titan cards, every single past generation has given you more as a percentage of shaders compared to the same generation pinnacle card than the 5000 series.

The issue is you, and most everyone else, was expecting between 20 and 100+ percent increase in generation to generation performance. People actually believed the marketing hype. Back here in reality...

Really now? You know what myself and everyone else was expecting specifically? No. I can only speak for myself, but I personally expected the new gen 60 tier card being the same performance or better than the old 70 tier card at a similar cost to the old 60 tier card. The new gen 70 tier card being the same performance or better than the old 80 tier card at a similar cost to the old 70 tier card. The new gen 80 tier card being at 15-30% faster than the old 80 tier card at a similar cost to the old 80 tier card. The new gen 90 tier card being 30%+ or better performance at a similar cost to the old 90 tier card.

 

[valthuer]

The word disappointing comes to mind first and foremost. This is basically a 4070 Super with MFG at a 50 dollar 'discount' assuming you hit the lottery and are able to buy one at msrp. Nvidia needs to start giving its non-flagship cards more VRAM and performance. The gap is just so large and growing between 90 and 80 class it doesn't leave enough room for gen on gen up lift in its lower tier cards. Personally I'd like to see Nvidia hobble AI or compute features like they did in the Titan years on the lower down skus while preserving close to flagship raster performance as much as possible. I understand some of this is upsell tactics. But things are getting stoopid at this point. End of day the 50 series just doesn't seem worth it in most situations, even IF you can get something at MSRP.

We are basically being witnesses to the same s**tshow as the RTX-40 series; the halo product, is the only one worth it and everything else is complete garbage.

 

[Jabberwocky79]

We are basically being witnesses to the same s**tshow as the RTX-40 series; the halo product, is the only one worth it and everything else is complete garbage.

The 4090 price was shockingly expensive at the time. This time around, everyone was bracing for the shock, only to be somewhat pleasantly surprised at MSRP announcements.... only to suddenly have whiplash from the degree of price inflation over the MSRP. - At least that's the way I saw it.