RTX 4060 Ti vs RX 7700 XT faceoff: Which midrange graphics card is superior?

[Flayed]

The supply of the 6800 and 6700xt cards is waning, at least here in the UK so soon they will no longer be an option.

 

[35below0]

Basically, some people just cannot justify spending $500+ on a GPU, and cards like the ones you poo-poo, while I'm not arguing with you that they aren't the best for the money when money isn't the #1 object, they could be the absolute high end of what some buyers can justify.... and that's really what it comes down to.

But that's why i was poo poo-ing the 7700XT. If $500 is out of reach, take a 4060 or 3060 12Gb. Someone pushing their budget above that in order to buy a 7700XT is wasting their limited money. In my opinion anyway.
It's buried in that book i wrote above.

That GPU(4060) offers a ton of performance for weaker, cheaper PCs and it doesn't need too much power to do it. It's a weak GPU for demanding gaming, but for the low price it's worth it.

A 4060 exceeds recommended requirements for a lot of popular Steam games, and is good for 1080p and even 1440p gaming, but without a chance of running everything on Ultra. And it shouldn't able to because it's a budget GPU.

The rest of the PC paired to a 4060 can also be cheaper because the monitor, PSU and CPU don't need to be very powerful.
7800XT/7900GRE need a little bit more oomph to take advantage of their ability.

Sounds like you're like me - you like to get the best bang for your buck, but also have enough money not to really have to give a sh-- about spending an extra $200.

In a way, exactly yes. I buy a PC when i have the money, and have no set budget. Aiming for longevity more than bang/buck, and picking things i want and that i can justify. For example i could not justify more than a i5 13600K or 4060 even if i could technically have afforded it. No need.

My goal was to spend money without regret, not save money without regret.

And that's in sharp contrast to most folks who do have a set budget, sometimes a very tight one.
To them i would offer advice on how to avoid pitfalls such as dirt cheap PSUs, or those AsRock B660M-HDVs (they have since revised the model) going for $60. And i would suggest some of the components that are on the cheap side but still reliable.

It's none of my damn business how someone unwinds or spends their money. But those of us who obsess over PC components and prices have some affinity and understanding of what's good value at a given time.

So yeah, i hear you loud an clear. I was sharing my opinion that the 2 GPUs do not offer enough to justify the extra cost.

It really comes down to knowing what you have to spend and working within that. But on the other hand, I have been the guy in years past who didn’t have much cash and would try to get used parts etc and rig things to work. So I get it.

Been there myself. 4-5 very low end PCs. Today i'm in the "knowing what is needed/justifiable, and working to spend toward that" camp.
For example i never want to work with a cheaply made PC case again. That's a no go, so i'll pay the cost of a high quality one.

In my opinion regarding the 4060 and 7700xt, they are just in a weird place.

It's the 4060 Ti. The souped up 4060.
Like a souped up Mitsubishi Mirage, it's still a Mitsubishi Mrage.
If all you need is a car that gets you from A to B, don't look at the souped up one. And preferably not a Mitsubishi Mirage either.

On the other hand, the 7700xt at say 325 starts looking more interesting if it goes that low.

Absolutely. The card itself is not what i poo poo, it's the card and it's price.
4060 vanilla @ $285 vs 7700XT @ $325, the 7700XT would be the better option for anyone who could stretch their wallet that far. The 4060 still only draws 115w and is basically good enough so it's still a viable choice.

If you're planning on keeping the card for a while the 7700 XT does make sense as an option due to the 12GB VRAM and assuming price limited. There are already games that can break 8GB of VRAM at 1080/1440 and it's bound to get worse.

There aren't many games that need more than 8Gb on their own, but when modded the requirements increase. For $390 the 7700XT is still just barely above a 4060 that costs $300. It's not going to make a dent in demanding or modded games or at 4K if anyone tries that. Even though ray tracing is kind of an option, at this price level it's not really very good.
Maybe when the 50XX series launches.

If modding heavily or pushing higher resolutions or fps, then you need a more powerfull GPU, and monitor/PSU/CPU to go with it.
You need a 7800XT or 7900GRE in that case. A $300 GPU is just as weak as a $400 one. But for $500 you can actually get a good GPU.

Trouble is all of them are going to take a step towards obsolescence within a year as the 50XX RTXs are coming.
Which makes spending extra on a budget GPU even more of an own goal.

 

[Flayed]

Hopefully, RDNA 4 cards will improve the sub $500 options.

 

[ohio_buckeye]

Yes with a lot of what was said in the previous post, depending on what happens in the 5000 series in the next 2 years, I’d say the 7700xt for example will become what the 5700xt became and what the 6700xt is becoming.

If you think back to those 2 cards they came out as 1440p cards. As time went on the 5700xt became a 1080p card effectively. About the performance of a 6600xt. I remember owning a 6700xt and it was fine at the time for 1440p, but id argue it’s on the way to becoming a fast 1080p card. That is how I see the 7700xt especially with only 12gb of vram in the next couple of years.

Regarding the 4060ti, I do remember seeing benchmarks for the 4060ti 8gb at least, and the 6750xt was basically tying it or sometimes beating it. So I don’t really regard that gpu much.

As far as rdna 4, not sure I see those cards being lower priced but I think what you may see is prices on the rdna 3 cards coming down as those release. Look at rdna 2 currently and you’ll get an idea what I mean. Where those cards started out And the prices they went to. I picked up a new 6800xt in November/december last year for $439 from Newegg which at the time was a great deal imo. So I’d expect prices on what’s there now sitting to come down.

 

[phxrider]

Newegg is selling the Sapphire Pulse versions of each card for $389.99 and $479.99 respectively with the GRE at $539.98 these are all basically the best prices for each card.

Now THAT seems like a sensible spread - none of them are priced so close together that it makes no sense not to get the higher model.

 

[thestryker]

There aren't many games that need more than 8Gb on their own, but when modded the requirements increase. For $390 the 7700XT is still just barely above a 4060 that costs $300.

There indeed aren't many, but a couple of years ago there were basically zero (unless you went to 4K, but that's definitely not what I'm talking about). A lot of games also dynamically reduce quality which means you might not see the problem in just the benchmark numbers unless the reviewer is watching for it. When you say the 7700 XT is barely above the 4060 what do you mean? It's a significantly faster GPU and the only time they're close is ray tracing which isn't really going to be playable on either one without a bunch of upscaling.

 

[ohio_buckeye]

I’d have to agree. The 6700xt/6750xt were pretty close to the 4060ti 8gb. I remember seeing video where they showed a 6750xt hanging with a 4060ti or beating it at times in Jedi survivor at the time. The competition to the 4060 non ti is the 7600.

The 7700xt is closer to an rx 6800, so rtx 3070 level performance really.

 

[phxrider]

For example i never want to work with a cheaply made PC case again. That's a no go, so i'll pay the cost of a high quality one.

A little off-topic, but depending on what spending for a high quality case means to you, have you worked with the Lian Li Lancool 216 case, for about $100? My current build is in one, and it's one of those cases that makes you constantly say to yourself, "I can tell these people build PCs themselves". It's just ingenious, and the ventilation is excellent as well. Tom's seems to agree with my assessment, as they gave it a glowing review.

 

[JarredWaltonGPU]

You deserve nothing but here is a price check anyway:

AsRock Phantom Gaming OC Radeon RX 7700XT - $449-557
Sapphire PULSE Radeon RX 7800 XT - $479-615

Price difference between lowest prices is $30. For highest price it's $55

What kind of comparison is that? It's like you intentionally picked the WORST priced 7700 XT to prove a point. How about the Sapphire RX 7700 XT Pulse? That's $399–$496 (but the $496 price is meaningless, as it's $399 at both Newegg and Amazon, and that's going to cover 95% or more of PC hardware buyers). Now, with Pulse cards on both, the price difference is $80, and at the highest price it's $119.

Also note that AMD has officially dropped the RX 7700 XT MSRP to $419, so anything more than that is above MSRP. Which for a custom card is fine, I suppose, but there's no way anyone sane would buy the ASRock Phantom OC 7700 XT at $550 when you could get a far superior 7900 GRE for the same price.

 

[JarredWaltonGPU]

Is this really what we need for AMD to squeak out a win? 50% more VRAM and noticable better gaming performance? You could count RT if you want but at this level that is more of a "bonus" as the performance hit is significant.

I am also tired of hearing the 8GB Nvidia = 10GB AMD nonsense. Show me a game with muddy textures on AMD 8GB that plays fine on Nvidia 8GB. Nvidia has some wind like efficiency, no need to try to gift them another.

I'll give you four examples from the test suite, at higher resolutions, but that's sort of the point as it's difficult to exceed 8GB VRAM use at 1080p in most games.

Diablo IV at 1440p and 4K RT Ultra:
4060 Ti 8GB: 82.6/42.5 fps (avg/1%low) at 1440p, 46.7/22.8 fps at 4K
7700 XT: 78.1/48.0 fps at 1440p, 44.3/19.9 fps at 4K

Forza Horizon 5 at 1440p and 4K extreme:
4060 Ti 8GB: 97.5/84.3 fps at 1440p, 57.8/48.8 fps at 4K
7700 XT: 103.2/81.8 fps at 1440p, 71.3/58.3 at 4K

Spider-Man: Miles Morales at 1440p/4K "Very High + RT max":
4060 Ti 8GB: 43.5/32.3 fps at 1440p, 23.8/17.1 fps at 4K
7700 XT: 51.0/41.2 fps at 1440p, 26.4/15.2 fps at 4K

Total War: Warhammer 3 at 1440p and 4K ultra:
4060 Ti 8GB: 43.5/32.3 fps at 1440p, 23.8/17.1 fps at 4K
7700 XT: 51.0/41.2 fps at 1440p, 26.4/15.2 fps at 4K

Keeping in mind that running out of VRAM generally means 1% lows will dip more than average fps, you'll notice that in Diablo IV, the 7700 clearly wins on 1% lows at 1440p, but then falls well behind at 4K — even though it has 50% more VRAM. That's weird and suggests Nvidia is doing better with memory management.

Forza Horizon 5 favors AMD overall, but the 1% lows are better at 1440p. The 4K results clearly hit the memory limit on the 4060 Ti however, which is why AMD opens up a 20%+ gap.

Spider-Man greatly favors the 7700 XT at 1440p, especially in 1% low fps, indicating yet again that the 4060 Ti ran out of VRAM. But at 4K, it strangely shifts and while AMD wins on average fps by 11%, it loses on 1% lows by 11%.

Total War: Warhammer 3 has the 7700 XT barely winning at 1440p (7% faster avg, 3% faster 1% lows), but at 4K there's a big reversal and the 4060 Ti 8GB wins by 6% on average fps, and by 60% on the 1% lows.

Those tests have been run many, MANY times (at least three per resolution, but I've also retested both cards at least four times since they launched...), and the results have generally remained the same, at least in the games that have been around a while. I'm not saying that Nvidia always behaves like it has more VRAM, but if sometimes behaves like it has more VRAM.

And we're looking at 8GB vs 12GB. If it were 8GB Nvidia vs. 10GB AMD (and I do have numbers for the 6700 10GB and 4060 as an example), the AMD GPU definitely doesn't act like it has more VRAM. At all.

 

[35below0]

What kind of comparison is that? It's like you intentionally picked the WORST priced 7700 XT to prove a point. How about the Sapphire RX 7700 XT Pulse? That's $399–$496 (but the $496 price is meaningless, as it's $399 at both Newegg and Amazon, and that's going to cover 95% or more of PC hardware buyers). Now, with Pulse cards on both, the price difference is $80, and at the highest price it's $119.

They were two cards close together on the list. I didn't pick the cheapest one, it's true.
The price is only part of it. And i wrote in my original post and later confirmed that if the 7700XT is cheaper, it is more attractive and justifiable.

I disliked both GPUs because they are in between the $300 tier (cheap & cheerful) and the $500 tier (not very good but decent and enough for most willing to spend more).
These $400 GPUs do not offer enough . They're still cheap and cheerfull, only difference is they cost more.

Will you or anyone else tell a gamer looking to buy a GPU that a $400 4060 Ti is a great, fast GPU? No. It's cheap and good luck running the latest and most demanding games at high settings. Is it good for the money? Again no. Is the 7800XT/7900GRE fast? Eh... for 1440p and for most except the most demanding games, yes it is. Is it good for the money? Hell yes.

7700XT and 4060 Ti get a lukewarm recommendation at best, and for good reason.

Also note that AMD has officially dropped the RX 7700 XT MSRP to $419, so anything more than that is above MSRP.

RTX 4060 MSRP is $285, but the majority of them sell above $300.
Agreed that a lower price makes it worth considering, but it's still a $400 GPU. Good for AMD and nvidia, bad for consumers.

Which for a custom card is fine, I suppose, but there's no way anyone sane would buy the ASRock Phantom OC 7700 XT at $550 when you could get a far superior 7900 GRE for the same price.

I'm not sure i follow. Maybe prices changed after you posted the reply (happens). The 7800XT Challenger is listed for $480.
AsRock had a few Challengers on sale recently, so maybe the price comparisons are unreliable.

Again, it's not the case that the 7700XT is a garbage product. I have never said that nor attacked AMD. I said all along that i see spending more than ~$300 for a 4060/3060 12 as a mistake unless we're going up to a 7800XT/7900GRE. Though more expensive, it is seriously better.
For those who cannot afford the expense, i'd advise sticking with a 4060.

 

[thestryker]

Again, it's not the case that the 7700XT is a garbage product. I have never said that nor attacked AMD. I said all along that i see spending more than ~$300 for a 4060/3060 12 as a mistake unless we're going up to a 7800XT/7900GRE. Though more expensive, it is seriously better.
For those who cannot afford the expense, i'd advise sticking with a 4060.

Based on the raster numbers from TPU's 7900 GRE review:
7700 XT is ~47% faster than a 4060 at 1080p and ~50% at 1440p with ~35% price difference between the cheapest models. Given the current state of quality PSUs I doubt one could save a meaningful amount of money here by going with the 4060. So if you can afford the 7700 XT, but that's your limit cost wise it sure seems like a better buy than the 4060 to me.

 

[35below0]

Based on the raster numbers from TPU's 7900 GRE review:
7700 XT is ~47% faster than a 4060 at 1080p and ~50% at 1440p with ~35% price difference between the cheapest models. Given the current state of quality PSUs I doubt one could save a meaningful amount of money here by going with the 4060. So if you can afford the 7700 XT, but that's your limit cost wise it sure seems like a better buy than the 4060 to me.

Admittedly those numbers seem compelling.

Sadly for AMD, it doesn't seem the 7700XT sold so well. Steam survey says the % of 4060s and 4060 Ti is 2.33% and 1.99% respectively, so i know a lot of people opted for the 4060 Ti at least.

https://imgur.com/usXx3Hj

View: https://i.imgur.com/usXx3Hj.png

I'm not sure if AMD GPUs are buried under generic names, or what. The 7700XT as such doesn't even feature on the list. Of the 7000 series only the 7900 XTX is listed.

https://imgur.com/UVwYfsv

View: https://i.imgur.com/UVwYfsv.png

 

[thestryker]

Admittedly those numbers seem compelling.

Sadly for AMD, it doesn't seem the 7700XT sold so well.

While I put very little stock in the steam hardware survey numbers due to their collection practices gamers have still shown over the last two generations they'll pay more for less from nvidia.

 

[35below0]

While I put very little stock in the steam hardware survey numbers due to their collection practices gamers have still shown over the last two generations they'll pay more for less from nvidia.

Yeah but AMD drivers terribl /s

Steam survey is not authoritative, no. Gives a glimpse at least.
What do you mean by collection practises?

 

[JarredWaltonGPU]

Yeah but AMD drivers terribl /s

Steam survey is not authoritative, no. Gives a glimpse at least.
What do you mean by collection practises?

Valve has never publicly stated how it actually samples PCs for the Steam Hardware Survey. Simply put, if it's not purely random, then the statistics are not valid. There are indications that the SHWS may sample "unknown" GPUs more frequently than "known" GPUs, and there are enough questions that I wouldn't take it as gospel truth. But I do think that, in general, the SHWS gives a decent indication of trends.

(I'd really like to know why only the 7900 XTX shows up right now for RDNA 3, though, as I can't believe there haven't been a modest number of 7900 XT, 7800 XT, 7700 XT, and 7600 purchases.)

On a different note, the RX 7700 XT had far less going for it when it launched at $449. A $70 price cut makes it a far more useful GPU. The 7800 XT ends up too close to the 7900 GRE in price these days, so that if you want more than a 7700 XT, that's the next best step. Compared to the previous RX 6700 XT, the 7700 XT performs about 30% faster, justifying it's current (street) pricing.

 

[35below0]

(I'd really like to know why only the 7900 XTX shows up right now for RDNA 3, though, as I can't believe there haven't been a modest number of 7900 XT, 7800 XT, 7700 XT, and 7600 purchases.)

Agreed. If it's the case that other RDNA 3s are statistically insignificant, then something is really weird in the Steam universe. I hardly think the 7900 XTX is the leader of that pack?

This has been mentioned before though. With some accusing Valve of bias.

 

[thestryker]

Steam survey is not authoritative, no. Gives a glimpse at least.

It does give an idea.

What do you mean by collection practises?

As Jarred mentioned there isn't any transparency, but just looking at some of the numbers you can tell where it's wrong. The only RDNA 3 being the 7900 XTX is certainly a big one, but things like more 3080 Ti than RX 6600 etc. They also have statistical anomalies which you can see pop up in both hardware/software. With the scale of the platform you really shouldn't see gigantic swings but it's happened at least 2-3 times in the last 18 mo.

 

[JarredWaltonGPU]

It does give an idea.

As Jarred mentioned there isn't any transparency, but just looking at some of the numbers you can tell where it's wrong. The only RDNA 3 being the 7900 XTX is certainly a big one, but things like more 3080 Ti than RX 6600 etc. They also have statistical anomalies which you can see pop up in both hardware/software. With the scale of the platform you really shouldn't see gigantic swings but it's happened at least 2-3 times in the last 18 mo.

I could actually understand 3080 Ti outselling the RX 6600, just because it came out during that crazy mining period. I think a lot of miners may have purchased it, or gamers who wanted to mine in their off hours. Granted, it had an Ethereum lock for a while, though that was eventually cracked.

There's also the difficulty with the RX 6600 effectively matching the RTX 2060 Super in rasterization performance and the RTX 3050 in RT performance, so anyone with a 2060 Super or better (or really, even an RTX 2060 or GTX 1080 or similar) would have little reason to upgrade to the 6600. So basically, if you were an AMD user, you'd need to have an RX 580/570 or below (or maybe an RX Vega 56) in order to think the 6600 was worth the upgrade. And it wasn't very good for mining back when it launched, so it didn't get a lot of sales in that area.

In general, though, AMD's dedicated GPU representation on the SHWS looks terrible — much worse than I'd expect. I tend to think that a lot of the "other" or "unknown" GPUs in the survey are from AMD and don't get separated or classified correctly for whatever reason.

The fact that the RTX 4070 Super, RTX 4070 Ti Super, and RTX 4080 Super all show up on the Steam survey less than three months after launch, while only the RX 7900 XTX shows up from AMD's RDNA 3 series, is still the most damning evidence.

My pseudo code for what SHWS sampling feels like:
Is this an unknown RTX or GTX GPU? If yes, greatly increase sampling rate.
Any other unknown GPU? Don't sample at all (which eliminates Arc and most of RX 7000-series).
Otherwise sample the PC about once every year.

 

[jordanbuilds1]

theres a few mistakes in the article i wanted to point but heres a major one most people missed: preformance.
i dont know what test bench hes using but theres something def wrong with it (Atleast in some of the games). minecraft is a fairly easy to run game at ultra but if a 4060 ti is barley pushing 90fps and the 7700 doing 50 fps, there is def something fishy. i know they said this was running dxr but thats still unrespectable preformance with the bench. (keep in mind, this is 1080p WERE TALKING HERE!)

 

[thestryker]

I could actually understand 3080 Ti outselling the RX 6600, just because it came out during that crazy mining period. I think a lot of miners may have purchased it, or gamers who wanted to mine in their off hours. Granted, it had an Ethereum lock for a while, though that was eventually cracked.

I know they sold a lot, but it still doesn't seem likely given that it has consistently been the best choice at $200 (or less) for about a year (thanks nv/amd ditching budget offerings). The 3080 Ti seemed to be the least common (well the 12GB 3080 which I assume was 3080 Ti binning) of the primary cards using GA102. Of course Valve doesn't publish detailed historicals so it's hard to see what sort of changes appear long term unless someone saved them.

The fact that the RTX 4070 Super, RTX 4070 Ti Super, and RTX 4080 Super all show up on the Steam survey less than three months after launch, while only the RX 7900 XTX shows up from AMD's RDNA 3 series, is still the most damning evidence.

I agree there's no way to explain this other than some sort of flawed dataset.

My pseudo code for what SHWS sampling feels like:
Is this an unknown RTX or GTX GPU? If yes, greatly increase sampling rate.
Any other unknown GPU? Don't sample at all (which eliminates Arc and most of RX 7000-series).
Otherwise sample the PC about once every year.

it absolutely feels that way for sure

 

[JarredWaltonGPU]

theres a few mistakes in the article i wanted to point but heres a major one most people missed: preformance.
i dont know what test bench hes using but theres something def wrong with it (Atleast in some of the games). minecraft is a fairly easy to run game at ultra but if a 4060 ti is barley pushing 90fps and the 7700 doing 50 fps, there is def something fishy. i know they said this was running dxr but thats still unrespectable preformance with the bench. (keep in mind, this is 1080p WERE TALKING HERE!)

You apparently haven’t ever tried running Minecraft RTX without upscaling. It’s almost purely GPU limited, so the RTX 4090 gets 270 fps while the 4060 Ti gets 90 fps. Not coincidentally, the 4090 has three times as much VRAM, running at higher speeds (21 Gbps vs. 18 Gbps), and over three times the core count (128 SMs vs. 34 SMs).

 

[jordanbuilds1]

You apparently haven’t ever tried running Minecraft RTX without upscaling. It’s almost purely GPU limited, so the RTX 4090 gets 270 fps while the 4060 Ti gets 90 fps. Not coincidentally, the 4090 has three times as much VRAM, running at higher speeds (21 Gbps vs. 18 Gbps), and over three times the core count (128 SMs vs. 34 SMs).

tried it on a friends system using an amd card, thats why im confused.
edit: i asked him and he said he was using shaders. also why use minecraft rt on a gpu without eather using an upscaler or shader? i know this is a test but rtx cards come with dlss out of drivers.

 

[bolshevik420]

Echoing the sentiment that the first graphs showing FPS at certain resolutions are illegible, awful things. Why would you choose dark blue and black?

 

[JarredWaltonGPU]

Echoing the sentiment that the first graphs showing FPS at certain resolutions are illegible, awful things. Why would you choose dark blue and black?

Because everything else can be quite garish looking. I've suggested in the past that if you REALLY care so much, please show me something better and let the forums sound off. I can change colors, but no one has actually said anything useful. (Hint: Red AMD and green Nvidia looks like Christmas vomit and we're not going to do that, not to mention red/green color blindness wouldn't do well with that either.)