is radeon RX 6750 XT a steal ?

[Vayra86]

thank you all, i have made a choice - TOPIC CLOSED

You can ask a mod to close a topic if you want to. The mods online are on the front forum page. Because knowing TPU, this could go on for awhile
Glad you succeeded Care to tell what you got?

 

[#22]

Explanation:

View attachment 352140

Its godawfully slow memory. The card is neutered by its bandwidth, not even its 8GB.

Cut-down specs like narrow bus and low badwidth are mentioned reasons to hate. But I asked for something more sensible like maybe some analysis showing these affecting performance. More than obvious losing performance from narrower bus going higher with res or from PCIe combined with platform using older standard.

thank you all, i have made a choice - TOPIC CLOSED

So you can go instead of CLOSING THE TOPIC WITH CAPITAL LETTERS like you don't want us have fun talking further and not caring about you :*

Be more realistic. TPU tests are out of date and include many irrelevant games. The 7700xt is between 20-30% faster in most recent AAA, having performance to spare is the best thing for someone trying to keep a product for the long term. RT is totally irrelevant in low-end GPUs, for obvious reasons, the 4060ti can't even maintain stable 60fps without RT in all games. You're here recommending giving up 20-30% extra performance for futile gimmicks like a frame generator, it's a joke and it won't save you when your framerate is dropping to 30fps.

You don't put yourself in the best light with linking some likely fake shit benchmarks from one of many channels like that on yt. Maybe let's stick to more established and renowned sources of performance data

 

[Vayra86]

Cut-down specs like narrow bus and low badwidth are mentioned reasons to hate. But I asked for something more sensible like maybe some analysis showing these affecting performance. More than obvious losing performance from narrower bus going higher with res or from PCIe combined with platform using older standard.

There's no point debating the same benchmarks we can all find on the web. Every card you buy is a prediction on how long it'll last. And if you want it lasting long, you don't skimp on memory, its that simple, the only proof you can get of that is by looking back, not forward. Cards with high memory capacity / high bandwidth last much longer. This is because its easy to tweak game settings like heavy post processing to reduce strain on the core and not see or feel much of it. But its impossible to turn down texture detail and other VRAM intensive stuff without seeing it.

Now of course, all things should be balanced. And with GPUs this is doubly true. The 4060ti though is not a well balanced card. There is a lot of core power, and there is barely any bandwidth. It will do fine now and with Nvidia TLC it will do it for a while. But once that stops, it immediately gets relegated to obsolete territory. This is the story of every x60 to date, bar a few rare exceptions like the 1060 6GB, along with the rest of Pascal. On Nvidia's side, the 4070 has okay balance, the 4070ti Super has it, the 4080 and the 4090 have it. The rest? Avoid.

 

[Vya Domus]

You don't put yourself in the best light with linking some likely fake

Would be pretty amazed if someone went through the trouble of faking every one of those real time charts.

 

[dgianstefani]

There's no point debating the same benchmarks we can all find on the web. Every card you buy is a prediction on how long it'll last. And if you want it lasting long, you don't skimp on memory, its that simple, the only proof you can get of that is by looking back, not forward. Cards with high memory capacity / high bandwidth last much longer. This is because its easy to tweak game settings like heavy post processing to reduce strain on the core and not see or feel much of it. But its impossible to turn down texture detail and other VRAM intensive stuff without seeing it.

Seems to me, as has been stated many times in GPU reviews, the memory capacity/bandwidth is irrelevant against how much power the GPU core actually has, and the desired resolution. No, not everyone is going to upgrade to 4K or even 1440p. 1080p is still by far the most popular PC gaming resolution. Otherwise the Radeon VII with its HBM memory and massive bandwidth etc., would have done much better, but no, the 1080 Ti is still faster, despite being two years older and with less than half the memory bandwidth.

No, the RX580 didn't age better than the 1060, despite having 8 GB memory in a 256 bit bus vs the 6 GB 192 bit 1060.

Neither is the 3060 12 GB doing any better now, compared to when it was reviewed in 2021 against the 3070/Ti with 8 GB, they're still ~35% ahead. Oh, and the 3060 Ti with 8 GB is still 15% faster.

Benchmark numbers are compelling.

"8 GB bad", or "narrow bus bad" have some merit, in a vacuum, but the choices here are: new 4060 Ti or 7700XT, or used 6750XT/6800 or 3080, and for the OP's specs, preference for new, budget and 1080p60/old games use case, there really is no contest. It's also not a fair comparison to just talk about bus widths, when for example, cache sizes are massively larger in Ada vs Ampere.

The 4060 Ti is more than twice as fast as the 1660S, in the same power envelope. Quite an upgrade for 1080p60.

 

[Vayra86]

Seems to me, as has been stated many times in GPU reviews, the memory capacity/bandwidth is irrelevant against how much power the GPU core actually has, and the desired resolution. No, not everyone is going to upgrade to 4K or even 1440p. 1080p is still by far the most popular PC gaming resolution. Otherwise the Radeon VII with its HBM memory and massive bandwidth etc., would have done much better, but no, the 1080 Ti is still faster, despite being two years older and with less than half the memory bandwidth.

No, the RX580 didn't age better than the 1060, despite having 8 GB memory in a 256 bit bus vs the 6 GB 192 bit 1060.

Neither is the 3060 12 GB doing any better now, compared to when it was reviewed in 2021 against the 3070/Ti with 8 GB, they're still ~35% ahead. Oh, and the 3060 Ti with 8 GB is still 15% faster.

Benchmark numbers are compelling.

"8 GB bad", or "narrow bus bad" have some merit, in a vacuum, but the choices here are: new 4060 Ti or 7700XT, or used 6750XT/6800 or 3080, and for the OP's specs, preference for new, budget and 1080p60/old games use case, there really is no contest. It's also not a fair comparison to just talk about bus widths, when for example, cache sizes are massively larger in Ada vs Ampere.

View attachment 352142

We disagree, because I base myself on personal experience rather than a bar chart. I've had my share of cards and I've seen them run out of juice, and how. These bar charts do not account for the options you have with higher GB cards in terms of IQ vs FPS gain/loss. An aspect that becomes highly relevant as cards age and you just don't want to or cannot upgrade. If money is no object, sure, buy a shitty card and 2 years later buy another.

The 4060ti is E waste. Simple as.

 

[dgianstefani]

We disagree, because I base myself on personal experience rather than a bar chart. I've had my share of cards and I've seen them run out of juice, and how. These bar charts do not account for the options you have with higher GB cards in terms of IQ vs FPS gain/loss.

You're projecting. I've used many generations of GPUs too.

Despite the people who cry out "you need 16 GB or more VRAM for 1440p Ultra", unsurprisingly, I still find my 12 GB 3080 Ti to be more performant than all the RDNA 2 16 GB cards.

I'd also like to know whether you think DLAA/DLDSR or having excess VRAM past 8 GB is more important for IQ at 1080p60. Answer is fairly obvious to me, but perhaps to an AMD GPU owner it's not as clear.

 

[Vayra86]

You're projecting. I've used many generations of GPUs too.

Despite the people who cry out "you need 16 GB or more VRAM for 1440p", unsurprisingly, I still find my 12 GB 3080 Ti to be more performant than all the RDNA 2 16 GB cards.

I'm not saying any of this. I'm talking about cards that have aged and how more VRAM makes them last longer because they give you more wiggle room. I always have by the way. Nowhere will I tell anyone you can't game on 12GB today or you can't game on 8GB today. But 8GB is not the right purchase today, that is true just the same. You're buying into a world that is flying past us, not the one coming at us.

 

[dgianstefani]

I'm not saying any of this. I'm talking about cards that have aged and how more VRAM makes them last longer because they give you more wiggle room. I always have by the way. Nowhere will I tell anyone you can't game on 12GB today or you can't game on 8GB today. But 8GB is not the right purchase today, that is true just the same. You're buying into a world that is flying past us, not the one coming at us.

Why?

You can't just throw out statements like that.

For 1440p, I'd probably agree with you, despite the evidence still suggesting, that even with 2024 games, 1440p Ultra is still doable with 60 FPS or more with 8 GB.

For 1080p?

People work within budgets and power/space constraints.

Like the example I just gave, the 1060 6 GB 192 bit is still faster than the RX 580 8 GB 256 bit. So the "over time" argument isn't clear to me either.

The reality seems to be "you can't future proof GPUs".

You're buying into a world that is flying past us, not the one coming at us.

Except OP clearly stated his resolution will not be changing from 1080p/60, and that he plays old single player games, so is this statement you've written true?

 

[Vayra86]

Why?

You can't just throw out statements like that.

For 1440p, I'd probably agree with you, despite the evidence still suggesting, that even with 2024 games, 1440p Ultra is still doable with 60 FPS or more with 8 GB.

For 1080p?

People work within budgets and power/space constraints.

Like the example I just gave, the 1060 6 GB 192 bit is still faster than the RX 580 8 GB 256 bit. So the "over time" argument isn't clear to me either.

The reality seems to be "you can't future proof GPUs".

Your reality, not mine. Perspectives and all. I've told you why, a sufficient number of times.

I've also told the OP not to spend just under what is required to make a card last. I've bought my share of x60's. They didn't last. The end.

Except OP clearly stated his resolution will not be changing from 1080p/60, and that he plays old single player games, so is this statement you've written true?

We don't know, as OP didn't share his decision with us I understand why, at this point. God almighty

If 400 was spent on a 4060ti, 400 was spent. I hope he has fun with it for a long time.

 

[dgianstefani]

Your reality, not mine. Perspectives and all. I've told you why, a sufficient number of times.

I've also told the OP not to spend just under what is required to make a card last.

You've "told" me why, sure, but that telling isn't backed up by any evidence, therefore there is no "sufficiency" to the argument without evidence.

We disagree, because I base myself on personal experience rather than a bar chart.

I've just demonstrated to you how a VRAM and bandwidth excess made zero difference over the long term with multiple examples, yet you're still trying to tell me that having excess on purchase is necessary, even if that isn't reflected in performance graphs either on release, nor six-eight years later. Apparently it's supposed to start mattering later? Because of "personal experience"?

 

[Vayra86]

Because of "personal experience"?

Yes. There is more to gaming than bar charts at max settings and a %.

 

[dgianstefani]

You're buying into a world that is flying past us, not the one coming at us.

Except OP clearly stated his resolution will not be changing from 1080p/60, and that he plays old single player games, so is this statement you've written true?

We don't know, as OP didn't share his decision with us I understand why, at this point. God almighty

You don't know? It seems to me OP knows.

no i will never play those modern games tested by forum, most are shooters which i deslike... im into skyrim, the elder scrolls, enshrouded type of survival farming open world discovery more slow paced peaceful games LOL --- i will upgrade my psu to 750 or 850 with 100~130€ ive kept alongside 400 for gpu is my 550w not enough

 

[Vayra86]

You don't know? It seems to me OP knows.

Who are you talking to now?

 

[dgianstefani]

Yes. There is more to gaming than bar charts at max settings and a %.

The 4060ti is E waste. Simple as.

Ah, so subjectively you prefer bigger numbers (VRAM and bus width), but objective performance testing both on release, and long term (eight years after for the 1060 6GB/580 8GB comparison) is irrelevant, and the current offering from the most popular tier of GPU (by far) from the most popular manufacturer (by far, 90%+ market share), designed for the most popular resolution (by far) - 1080p, is "E-waste".

Cool.

Thanks for explaining, I was confused.

thank you all, i have made a choice - TOPIC CLOSED

Great, enjoy your new GPU!

 

[#22]

There's no point debating the same benchmarks we can all find on the web. Every card you buy is a prediction on how long it'll last. And if you want it lasting long, you don't skimp on memory, its that simple, the only proof you can get of that is by looking back, not forward. Cards with high memory capacity / high bandwidth last much longer. This is because its easy to tweak game settings like heavy post processing to reduce strain on the core and not see or feel much of it. But its impossible to turn down texture detail and other VRAM intensive stuff without seeing it.

E.g. it was popular to claim that 4060Ti won't be able to use it's whole memory due to such low bandwidth. If I recall, Hardware Unboxed demistified it. I originally meant things like that (instead of hating just cut-down specs numbers). These numbers make it awful, shameless product, but on the other hand offering quite many fps with support of superior upscaler and amount of VRAM you rather won't worry about. Cut-down specs make it losing more perf when going higher with res or lower with PCIe of platform, but I find it far from making this card a definite no go. Especially in case of this girl playing FHD and e.g. liking Skyrim mods being potentially heavy on VRAM.

BTW do I understand you correctly that you claim that lower bandwidth will make card loosing more perf with a time, meaning future games? Not that I doubt, it's like with this not being able to utilize 16GB above, I would like to see something covering it.

Would be pretty amazed if someone went through the trouble of faking every one of those real time charts.

I more amazed seeing somebody relying on such sources or defending them like above and assuming they are legit when really likely not even bothering to check it

 

[Sithaer]

Well that escalated huh.
All I have to say is since the OP has made the choice aleady but its something that I've noticed more than a few times around here.

OP has clearly stated the use case and plans for the future upgrades/gaming so why even argue about moot points 'hardware in this case'?

If the OP is staying at 1080P/60 with mainly older games then hes getting exactly no value out of a stronger and more expensive yet more power hungry card anyway.
'This is the main reason why I pretty much never ask about my upgrade ideas around here or on any tech forum cause its like some ppl pretend to know MY use case/preferences better than myself an then argue about it..'

 

[Vayra86]

BTW do I understand you correctly that you claim that lower bandwidth will make card loosing more perf with a time, meaning future games? Not that I doubt, it's like with this not being able to utilize 16GB above, I would like to see something covering it.

A bandwidth limitation is just a hard cap and it can hit you hard. Having a lack of bandwidth or VRAM can escalate into stuttery behaviour. Its similar to a critical lack of CPU processing power, you will get a hang. On the other hand, a core limitation hits you consistently in terms of FPS. Heavy scenes, will still be heavy scenes with low FPS. But low is not stutter. Its just low.

And that's where my budget argument comes in - if a card 'forces' you to upgrade because you simply run into this problem in the games you want to play (you're not telling me someone with a new card is not playing any recent content, Enshrouded being one example mentioned, so there will be others - talking about perspective ), you've made a bad purchase. And with a mere 288GB/s, that is a real scenario. Even at 1080p.

 

[dgianstefani]

Well that escalated huh.
All I have to say is since the OP has made the choice aleady but its something that I've noticed more than a few times around here.

OP has clearly stated the use case and plans for his future upgrades/gaming so why even argue about moot points 'hardware in this case'?

If hes staying at 1080P/60 with mainly older games then hes getting exactly no value out of a stronger and more expensive yet more power hungry card anyway.
'This is the main reason why I pretty much never ask about my upgrade ideas around here on or any tech forum cause its like some ppl pretend to know MY use case/preferences better than myself an then argue about it..'

Great to see other people understanding OP's requirements, instead of writing about their own.

A bandwidth limitation is just a hard cap and it can hit you hard. Having a lack of bandwidth or VRAM can escalate into stuttery behaviour. Its similar to a critical lack of CPU processing power, you will get a hang. On the other hand, a core limitation hits you consistently in terms of FPS. Heavy scenes, will still be heavy scenes with low FPS. But low is not stutter. Its just low.

So these low FPS, one would assume they show up on a 1% low FPS chart? Perhaps take a look at the one where the 4060 Ti 8 GB performs better than the 2080 Ti 11 GB, 6700XT 12 GB, or 3060 12 GB in 1080p 1% lows.

ASUS GeForce RTX 4060 Ti Dual with M.2 Slot Review - Gen 5 Supported

The ASUS GeForce RTX 4060 Ti Dual with M.2 Slot is a unique graphics card that showcases innovation by combining an RTX 4060 Ti GPU with an M.2 NVMe SSD slot. In our review we test the M.2 interface with a Gen 5 and Gen 4 SSD to find out whether there's any performance compromises to accept.

www.techpowerup.com

 

[Vya Domus]

I more amazed seeing somebody relying on such sources or defending them like above and assuming they are legit

I don't know what you mean by legit, you can search for other comparison videos like that, if the differences between the cards are consistent I don't see any reason to believe they're all "fake".

 

[#22]

A bandwidth limitation is just a hard cap and it can hit you hard. Having a lack of bandwidth or VRAM can escalate into stuttery behaviour. Its similar to a critical lack of CPU processing power, you will get a hang. On the other hand, a core limitation hits you consistently in terms of FPS. Heavy scenes, will still be heavy scenes with low FPS. But low is not stutter. Its just low.

I know about bandwidth limitation, so you don't have to explain me that and I asked about something covering it. Like some analysis comparing older hardware from this perspective. Like I said, I don't doubt it happening, but find it interesting and would like to see examples and numbers - everything is theoretical until it gets practical enough to hurt or it never happens

I don't know what you mean by legit, you can search for other comparison videos like that, if the differences between the cards are consistent I don't see any reason to believe they're all "fake".

Legit means correct, reliable, simply agreeing with other, more serious sources. You should check if Testing Games' numbers are legit instead of assuming they are, because you find fake benchmarks hard to produce. It was a mistake in your way of thinking, but to make things easier I would just screw it and stick to reputable sources instead


"Hi, my name is Andrei, I work in a computer store where I take various components for tests in my videos. On advertising please contact : (Сarefully check the email address, they can write fakes on my behalf with one character changed!!!) You can also order advertising through a convenient service PinTop - I don't have a telegram channel, anyone who offers to go to telegram on my behalf is a scammer!!!!"

It's most likely fake garbage making money on ads and naive people seeking performance numbers from random popping yt channel - the thing is, there's no need to watch or check it.

EDIT: removed links of damn Andrei lol

 

[Vayra86]

I know about bandwidth limitation, so you don't have to explain me that and I asked about something covering it. Like some analysis comparing older hardware from this perspective. Like I said, I don't doubt it happening, but find it interesting and would like to see examples and numbers - everything is theoretical until it gets practical enough to hurt or it never happens



Legit means correct, reliable, simply agreeing with other, more serious sources. You should check if Testing Games' numbers are legit instead of assuming they are, because you find fake benchmarks hard to produce. It was a mistake in your way of thinking, but to make things easier I would just screw it and stick to reputable sources instead


"Hi, my name is Andrei, I work in a computer store where I take various components for tests in my videos. On advertising please contact : frak61822@gmail.com (Сarefully check the email address, they can write fakes on my behalf with one character changed!!!) You can also order advertising through a convenient service PinTop - https://pin.top/TestingGames I don't have a telegram channel, anyone who offers to go to telegram on my behalf is a scammer!!!!"

It's most likely fake garbage making money on ads and naive people seeking performance numbers from random popping yt channel - the thing is, there's no need to watch or check it.

The most recent personal experience I've had in both Oblivion and in Skyrim with mods; Elder Scrolls Online on GTX 660, both single and in SLI; swapping to less core better VRAM 770 was a straight up win there on all aspects, even with it being a 2GB card just the same. . Part of that was due to funky memory config on the 660 (a similar debacle as the GTX 970 3.5GB). After that I didn't fail on my GPU purchases anymore. The 4060ti has every look of being another such weak horse in the Nvidia stable. We can also recall the 960 which was an utterly pointless release.

For the good cards, the 780ti 3GB ran out of juice in TW WH2 and the 1080 8GB did so in part 3 at 3440x1440. But at that point the core perf was also abysmal and only with heavy tweaking would you get the thing playable - but still sitting at 7.x GB VRAM

A lot of this often depends on when you purchase a card, too. There's going to be movement on the console front soon and there IS movement in the engine space with UE and heavier shading tech. The stars align for the 4060ti in that respect for not having a long life.

 

[Vya Domus]

You should check if Testing Games' numbers are legit instead of assuming they are, because you find fake benchmarks hard to produce.

Dude, all of this is based on faith anyway, you have to make the same assumption for any other website. It's not like I can go grab 50 video cards and reproduce all of TPUs benchmarks, I assume they're probably legit.

 

[#22]

Dude, all of this is based on faith anyway, you have to make the same assumption for any other website. It's not like I can go grab 50 video cards and reproduce all of TPUs benchmarks, I assume they're probably legit.

So using your way of thinking, I recommended you and og yt benchmarks enthusiast to put your damn faith in data coming from more reputable sources lol

It's this kind of situation when I explained you why you are wrong for obvious reasons, so you can only agree with it, but won't admit it and instead you try hard to defend yourself making you look even worse than in the beginning. Not counting it being total waste of time and this thread, so maybe better go listen to Justin Bieber or something.

 

[Ruru]

Dude, all of this is based on faith anyway, you have to make the same assumption for any other website. It's not like I can go grab 50 video cards and reproduce all of TPUs benchmarks, I assume they're probably legit.

Personally I find TPU the go-to site what it comes to reviews.