AMD Radeon RX 9060 XT Reportedly Capable of Boosting Up To 3.3 GHz, New Leak Suggests "Navi 44 XT" GPU

[Vayra86]

Understand what I am asking:

Two cards equal in spec except one is 192-bit, GDDR6 and 448GB/s, the other is 128-bit, GDDR7 and 448GB/s. Which performs better?

They both perform identically. So why the obsession with bus width? It's boring.

Agreed, in that context its irrelevant.

 

[Deleted member 187435]

Depends on what you're using it for, as usual, the fact remains, this kind of bandwidth was normal 10 years ago in a midrange card, but x60 got stuck on it indefinitely.

Perhaps AMD and nvidia know that more isn't always needed relative to the processing power of the cards in this class. It would be like having a low powered car with a 6th gear which can do 250 mph. It's moot because it will never be able to use all that potential.

We have a real world example of this in the GTX 1660 and 1660 Super, the latter has 75% more bandwidth but performance only went up by a fraction of that.

 

[Vayra86]

Perhaps AMD and nvidia know that more isn't always needed relative to the processing power of the cards in this class. It would be like having a low powered car with a 6th gear which can do 250 mph. It's moot because it will never be able to use all that potential.

Part of that is true but its no different from what I just said: sometimes you don't need all the VRAM nor bandwidth. The bottom line remains; if there is some content the core can easily crunch and the VRAM can't, you've got a badly balanced card, that could have done more with just $5-10,- spent on a few more chips. Be mindful of what you're defending here

The other part is that having a tight VRAM capacity and bandwidth is the perfect way to force upgrades in a timely manner. Additionally, VRAM is actively used now to sell/not sell a card due to people using LLMs. That's why we're not seeing this segment move to 12GB. Its not because the cards can't use 12GB even for gaming. Its because there is a profit strategy behind this.

AMD and Nvidia know that the x60 segment is for people with tighter budgets and clearly a shorter view on their future expenses. Its like buying shoes. You can buy plastic shoes for $10,- and replace them every few months because they've gone to shit, or you can spend $60,- and walk on them for years. Both allow you to walk on more than bare feet... but its likely the $60,- shoes also happen to be better for your feet and more comfy during all this time. GPUs are no different these days, since the x60 segment has begun stalling (Ampere onwards). Don't fool yourself. You're getting ever less GPU for your money, while the higher end propositions offer more in a relative sense. Its literally the worst possible time to buy these cards, and has been, for a while now.

Everything else is cognitive dissonance.

 

[Deleted member 187435]

Part of that is true but its no different from what I just said: sometimes you don't need all the VRAM nor bandwidth. The bottom line remains; if there is some content the core can easily crunch and the VRAM can't, you've got a badly balanced card, that could have done more with just $5-10,- spent on a few more chips. Be mindful of what you're defending here

The other part is that having a tight VRAM capacity and bandwidth is the perfect way to force upgrades in a timely manner. Additionally, VRAM is actively used now to sell/not sell a card due to people using LLMs. That's why we're not seeing this segment move to 12GB. Its not because the cards can't use 12GB even for gaming. Its because there is a profit strategy behind this.

I agree about capacity which is why I got a 16GB card despite all the people saying it was largely pointless. Actually there are cases where the 8GB card really chokes and then the lower bandwidth does very much hurt it in ways it doesn't hurt the 16GB one. But having said that, 8GB is perfectly adequate still, if you are on a tighter budget, game at 1080p and don't mind turning some settings down so I'm not going to criticise those cards too much.

 

[Vayra86]

I agree about capacity which is why I got a 16GB card despite all the people saying it was largely pointless. Actually there are cases where the 8GB card really chokes and then the lower bandwidth does very much hurt it in ways it doesn't hurt the 16GB one. But having said that, 8GB is perfectly adequate still, if you are on a tighter budget, game at 1080p and don't mind turning some settings down so I'm not going to criticise those cards too much.

You do you, enjoy a few more generations of similar products then, I guess. You're just supporting your own stagnation. You even acknowledged it - 8GB chokes content the card should easily run. Its not very relevant there is a great amount of stuff that doesn't choke on 8GB. Its the stuff you CANT run properly that will force earlier upgrades and more money out of your pocket.

Its interesting to look back on a series of x60 purchases over a longer period of time, and do the math: what did you spend over that time, and how much net performance have you got compared to the latest gen? I can guarantee you the x60 will show a pretty frightening number that cannot weigh up against a well timed high end GPU bought, say, 6 years ago (and considering you've never had to dial down anything during that time, you only might do it in the last couple years, much like you would on every x60 from the start).

Here's an example.

Here's another:

Now consider the fact you can only buy, at BEST, two x60's for the price of one 1080ti.
And that's not even considering the potential resale value. You can still sell a 1080ti for a good $100,-

 

[Deleted member 187435]

You do you, enjoy a few more generations of similar products then, I guess. You're just supporting your own stagnation. You even acknowledged it - 8GB chokes content the card should easily run. Its not very relevant there is a great amount of stuff that doesn't choke on 8GB. Its the stuff you CANT run properly that will force earlier upgrades and more money out of your pocket.

Its interesting to look back on a series of x60 purchases over a longer period of time, and do the math: what did you spend over that time, and how much net performance have you got compared to the latest gen? I can guarantee you the x60 will show a pretty frightening number that cannot weigh up against a well timed high end GPU bought, say, 6 years ago (and considering you've never had to dial down anything during that time, you only might do it in the last couple years, much like you would on every x60 from the start).

Here's an example.

View attachment 395153

Yes on paper the 4060 Ti is bad but then despite having less than half the bandwidth and (in the case of the 8GB card) less RAM it still beats the 2080 Ti while costing less and using less power. Isn't that impressive really? And the 5060 Ti + 9060 XT will be cheaper and faster still. I don't think we can complain too much.

 

[Vayra86]

Yes on paper the 4060 Ti is bad but then despite having less than half the bandwidth and (in the case of the 8GB card) less RAM it still beats the 2080 Ti while costing less and using less power. Isn't that impressive really? And the 5060 Ti + 9060 XT will be cheaper and faster still. I don't think we can complain too much.

No I'm not impressed by the 4060 Ti, lol. Its a very poor product, commercially. Technically, its a poorly balanced GPU, and it is far too expensive for its given hardware. After the x70ti non super its the worst positioned card in the Ada stack. There is no single metric on which the 4060ti is great, not even within the Ada stack. Its 16GB sibling is even known to perform worse than the 8GB counterpart because its struggling for power that can't be directed to the core boost.

Ada's efficiency gain is impressive, but it doesn't save me money either. Before you start about an energy bill... you're still maxing out the card to get the desired performance, its still 200 160odd watts, despite the watt/frame metric being better. You'll just want more frames as a result.

I'm sorry. It is what it is, Ada was largely a clusterfuck, and Blackwell is even worse. The only good Ada card I could detect was the 4090 and the 4070ti Super given their price/performance ratios and overall balance. The rest should be skipped.

 

[Mrgravia]

If the hit the price point I think it will sell well.

Also, praying someone gives us a low profile option this gen.

 

[Deleted member 187435]

No I'm not impressed by the 4060 Ti, lol. Its a very poor product, commercially. Technically, its a poorly balanced GPU, and it is far too expensive for its given hardware. After the x70ti non super its the worst positioned card in the Ada stack. There is no single metric on which the 4060ti is great, not even within the Ada stack. Its 16GB sibling is even known to perform worse than the 8GB counterpart because its struggling for power that can't be directed to the core boost.

Ada's efficiency gain is impressive, but it doesn't save me money either. Before you start about an energy bill... you're still maxing out the card to get the desired performance, its still 200 160odd watts, despite the watt/frame metric being better. You'll just want more frames as a result.

I'm sorry. It is what it is, Ada was largely a clusterfuck, and Blackwell is even worse. The only good Ada card I could detect was the 4090 and the 4070ti Super given their price/performance ratios and overall balance. The rest should be skipped.

First I've heard of that. Probably not significant? You can easily increase the power to 109% if you want anyway. More often it's the other way around, and I assume this applies to the AMD cards too (to keep it on topic of the 9060 XT which also comes in 8+16 versions).

Video shows dramatically better performance in the 16GB 4060 Ti in multiple games.

 

[Vayra86]

First I've heard of that. Probably not significant? You can easily increase the power to 109% if you want anyway. More often it's the other way around, and I assume this applies to the AMD cards too (to keep it on topic of the 9060 XT which also comes in 8+16 versions).

Video shows dramatically better performance in the 16GB 4060 Ti in multiple games.

Not very significant no, but the higher power target doesn't change this either because you're also capable of doing that on the 8GB version. Its very logical though, more chips cost more power that can't go elsewhere, and VRAM power usage is fixed relative to its (current) speeds.

Either way it goes to show what a strange product this is, the 16GB clearly being an afterthought.

You can browse through this review to get an idea; the margins are small, but they are there. Note how the higher resolution creates the opposite effect here to what you'd expect on a higher vram card.

NVIDIA GeForce RTX 4060 Ti 16 GB Review - Twice the VRAM Making a Difference?

Finally! Our GeForce RTX 4060 Ti 16 GB review is live. We're investigating what performance gains you can expect from doubling the VRAM size to 16 GB. Our review of the 16 GB version also looks at power consumption and how overclocking is affected by the additional four memory chips on the back...

www.techpowerup.com

NVIDIA GeForce RTX 4060 Ti 16 GB Review - Twice the VRAM Making a Difference?

Finally! Our GeForce RTX 4060 Ti 16 GB review is live. We're investigating what performance gains you can expect from doubling the VRAM size to 16 GB. Our review of the 16 GB version also looks at power consumption and how overclocking is affected by the additional four memory chips on the back...

www.techpowerup.com

Some games lean heavily into VRAM and assets taken directly out of it, using cache far less efficiently; which is where the 16GB shines despite lack of bandwidth (it 'fixes' the bandwidth issue):

NVIDIA GeForce RTX 4060 Ti 16 GB Review - Twice the VRAM Making a Difference?

Finally! Our GeForce RTX 4060 Ti 16 GB review is live. We're investigating what performance gains you can expect from doubling the VRAM size to 16 GB. Our review of the 16 GB version also looks at power consumption and how overclocking is affected by the additional four memory chips on the back...

www.techpowerup.com

Anyway... the point stands, I hope you understand where its coming from now a little better.

 

[Deleted member 187435]

Not very significant no, but the higher power target doesn't change this either because you're also capable of doing that on the 8GB version. Its very logical though, more chips cost more power that can't go elsewhere, and VRAM power usage is fixed relative to its (current) speeds.

Either way it goes to show what a strange product this is, the 16GB clearly being an afterthought.

Presumably raising the power limit equalises the 16GB's performance to the 8GB card at least. I mean, if you raised the 8GB card's power limit then it wouldn't make that card any quicker, because it wasn't limited by power to begin with. But that's interesting anyway, it could explain why in some of TPU's charts that the 16GB card is a couple of tenths slower sometimes.

For me at least, in games where I use vsync at 1080p60, the card doesn't come close to its default power limit anyway, it's only during very GPU intensive games, benchmarks or stress tests where you'd find the limit. Usually I'd adjust graphics settings so the card isn't running flat out like that.

 

[Vayra86]

Presumably raising the power limit equalises the 16GB's performance to the 8GB card at least. I mean, if you raised the 8GB card's power limit then it wouldn't make that card any quicker, because it wasn't limited by power to begin with. But that's interesting anyway, it could explain why in some of TPU's charts that the 16GB card is a couple of tenths slower sometimes.

This depends on the content; ONLY the content that benefits from the extra 8GB will find better performance on the 16GB card. In every other situation, the 'leaner' lower VRAM card has the potential to do better. It doesn't matter at all whether you power limit the cards or not; if they run at the same % power budget, whether thats 100% or 109%, they will show a similar behavior, UNLESS you didn't need that power to begin with and are running into another limiting factor.

And yes, that's precisely why those benches are marginally below the 8GB versions there.

 

[morn14150]

Where is the 12GB VRAM sweet spot? I swear both companies are intentionally handicapping their cards to 128-bit bus and only gives us 8GB (which is DOA already in 2025) or 16GB (potentially overpriced af)

 

[medi01]

to be able to compete with the 5060s

Leaks show that 5060s suck more than even 5070, which is on record levels of sucking. I doubt 9060s would have problems competing with that crap.

Sometimes I think people would be happier if the bus was 192 or 256, even if the performance was exactly the same. It would just make them happier that the bus number is bigger.

I personally do not give a flying F about <insert tech>.
Performance, power consumption, price. Maybe "ficherz" is what matters.
What bazinga was used internally this time - why does it even matter?

 

[Deleted member 187435]

I personally do not give a flying F about <insert tech>.
Performance, power consumption, price. Maybe "ficherz" is what matters.
What bazinga was used internally this time - why does it even matter?

Not sure if you are agreeing with me but of course that is the point I was trying to make.

 

[john_]

Leaks show that 5060s suck more than even 5070, which is on record levels of sucking. I doubt 9060s would have problems competing with that crap.

Pricing and how much real that pricing is will have the final say. 9060 XT might end up close to a 7800XT, so 5060 Ti will need to be at around a 7700XT performance level to not have any kind of problem to sell. Why? because of Nvidia's stronger name and brand. Now, if only AMD could offer the 9060XT at $379 and the 9060 non XT at $299, that would have been great. And 9070 GRE at $479.

 

[Sol_Badguy]

Again going with the assumption that the provisional specs turn out to be real.

9060 XT specs / 9070 XT specs / 5060 Ti specs / 5070 Ti specs

What is immediately apparent is that the 9060 XT is half of 9070 XT in everything but with significantly higher clock speed where we have 220 Mhz (+9.2%) extra Game Clock and 260 Mhz (+8.8%) extra Boost Clock.
Knowing from the 9070 vs 9070 XT power draw behavior that the 9070 XT is factory overclocked beyond the effciency curve to the point of having higher power draw than the 5070 Ti and power spikes comparable to higher tier cards makes me believe that the 9060 XT will be the same or worse for its category -> basically higher power draw than the 5060 Ti.

About the performance, as I said the 9060 XT is half of 9070 XT but the 5060 Ti is a little more than half of 5070 Ti when it comes to some compute units, others are exactly half just like the bandwidth. It too is higher clocked than the 5070 Ti, 112 Mhz (+4.9%) extra Base Clock and 120 Mhz (+4.9%) extra Boost Clock.
So if everything scales perfectly the difference between the 9060 XT and the 5060 Ti will be very similar to the difference between the 9070 XT and 5070 Ti. The 5060 Ti has some extra compute units to be a pinch over 50% of 5070 Ti but the 9060 XT has some extra Mhz to compensate.
If we look at the bandwidth differences (322 vs 448 & 644 vs 896) the ratios are exactly the same between the small cards and their bigger brothers. Again, if everything scales perfectly...
In my opinion, if these specs turn out to be true, there is no way the 9060 XT is going to beat the 5060 Ti, it just doesn't have the meat to do it, but nevertheless it's going to be very close, so close that in raster one cannot tell the difference while playing. In raytracing we know the drill.

Now about the price, if the 5060 Ti 16GB is getting a $430 MSRP then it would be absolutely reasonable to expect the 9060 XT 16GB at $350 at the most, after all the deficit in feature set is still there. $350 is the price that fits perfectly in this scheme:
5070 Ti - $750 ....................................... 9070 XT - $600
5060 Ti 16GB - $430 ............................. 9060 XT 16GB - X
Some people mentioned that the 5060 Ti 16GB will be a good card for light productivity workloads, this adds value whereas with AMD it's hit or miss, until everything is on par AMD must undercut nVidia in prices, after all you are getting less for your money.

 

[RejZoR]

Any videocard you see out there starts with the best slice of the wafer (high end) towards lower end. The lower end are cut down chips, that did not meet Quality guidelines. It is a waste to throw half of a wafer away because they are not matching up to the best ones.

So your a bit complaining while all fabs are doing the same for years really: it's only a good thing. Less E-waste and such. More for us.

3.3Ghz is nothing spectacular. Some 9070XT models have reached that. I'd say that that card is going to be a perfect all round 1440p card.

Um, at what point I've complained about anything? This has been standard practice for decades at this point. I was just telling that sometimes it's beneficial to have a cutdown rejected chip of a higher tier card because it has larger chip with fewer operational cores/shaders which means it cools easier. That's all I said. If you make a dedicated smaller chip it'll have smaller die surface area which is always harder to cool.

 

[lexluthermiester]

Agreed, in that context its irrelevant.

Exactly. 448GB a sec? Are we kidding? That is amazing performance for a budget range product so who cares what bit-rate the memory bus operates on.

So for a card with 128bit GDDR7 to be able to perform on the same level as GDDR6 at 192bit is a showing of a 50% increase in performance. from one gen to the next. Not bad at all.

 

[The Shield]

With those 8 Gb, they went full Nvidia mode.

 

[Dwarden]

really hoped AMD will offer higher bandwidth on those cards than just 320GB/s
... sigh , the lack of competition makes all the offering bad or not good,
Intel didn't even bother to go with 2nd gen to do Battlemage with 16GB,
with good memory and lil faster bus it would be great card for the price

starts to feel like corporations are stuck in era of AI generated product lines

 

[lexluthermiester]

With those 8 Gb, they went full Nvidia mode.

No, they're doing what makes sense.

Contrary to popular notions and opinions, 8GB of VRAM is still a good amount of RAM.

 

[Dawora]

Tsk tsk DOA just like Nvidia's 5060's series but somehow they'll sell even with gimped memory configs, should of been simply 12GB all around, no confusion for customers.

at least Nvidia Ti have 16GB Vram