AMD Radeon RX 9060 XT to Roll Out 8 GB GDDR6 Edition, Despite Rumors

[Mrgravia]

No, it doesn't. Small Form Factor has been half height, by design since it's inception and release. Certain companies are trying to redefine a standard they have no right or authority to redefine.

And who exactly has the right to declare that? What authority defines such a standard?

How exactly do you define a "SFF" pc then? Volume of the case is the most common, so whats the cutoff in your opinion?

Edit: You updated your previous post, but you seem to be doing exactly what you claim the companies are doing: Declaring by fiat details of a format, despite the common use of said terminology.

No one gives a damn how it was initially created, it's based on how the PC buying public uses the term in question. SFF has come to mean the volume of the case, rather then individual parts selected. Industry seem to agree and markets it as such.

 

[lexluthermiester]

And who exactly has the right to declare that? What authority defines such a standard?

Go look it up. Reference OEM makers like HP, Dell, Lenovo and the like.

How exactly do you define a "SFF" pc then? Volume of the case is the most common, so whats the cutoff in your opinion?

I've already explained this above and with photo's in the thread. SFF = LP = half height expansion cards.

 

[AusWolf]

Low Profile is single fan cards that are powered only by the PCIE connector traditionally, yea, but SFF cards are not. They are usually dual fan cards. I think of SFF as a middle ground between LP and regular full size cards.

SFF cards were intended (as far as I'm aware) for newer ITX cases but I think LP is less about gaming on ITX and more about other more useful things GPU's can do other than gaming, such as video encoding or etc.

That's as far as my understanding goes anyway.

Again, I think a SFF 9060XT would rock. 8GB of VRAM should be fine for SFF builds who probably are using it more portably anyway on smaller displays.

The problem is, there is no standardised definition for SFF. You have LP (low profile) which means half-height to fit into a slim case. We used to have ITX which meant single fan and not longer than the PCI-e connector to fit into a cube case. SFF is just whatever we randomly come up with. That's why Nvidia is using the term for any card that's less than 4326538 slots wide and 234956924 mm in length.

 

[TSiAhmat]

Gigabyte 5060 OC Low profile

Now to see how much of an upgrade it is compared to the existing 4060 LP.

that gpu looks looks very cute (in a good way)

 

[Mrgravia]

Go look it up. Reference OEM makers like HP, Dell, Lenovo and the like.

I've already explained this above and with photo's in the thread. SFF = LP = half height expansion cards.

Go look it up where? Where is this consortium of OEM who defines what exactly is an SFF PC? Wiki points to it being originally used for Shuttle form factor, which doesn't appear to meet your requirements.

From intel:

"What exactly qualifies a PC as having a small form factor?

There isn’t a universally agreed-upon definition, but there are some loose guidelines tied to case measurements. When discussing the internal layout and size of PC cases, volume is often an important consideration, and is usually measured in liters. A standard-size ATX desktop tower is usually around 40-45 liters. By comparison, a common measurement for SFF is around half of that — 25 liters or less.

However you define SFF, there’s a lot of information out there about building in this unique form factor. Here’s what you should be thinking about when planning your build."

Lenovo, HP, all say similar. Nothing I can find says anything about the length of video cards or their height as a requirement to be SFF.

 

[lexluthermiester]

Go look it up where? Where is this consortium of OEM who defines what exactly is an SFF PC? Wiki points to it being originally used for Shuttle form factor, which doesn't appear to meet your requirements.

From intel:

"What exactly qualifies a PC as having a small form factor?

There isn’t a universally agreed-upon definition, but there are some loose guidelines tied to case measurements. When discussing the internal layout and size of PC cases, volume is often an important consideration, and is usually measured in liters. A standard-size ATX desktop tower is usually around 40-45 liters. By comparison, a common measurement for SFF is around half of that — 25 liters or less.

However you define SFF, there’s a lot of information out there about building in this unique form factor. Here’s what you should be thinking about when planning your build."

Lenovo, HP, all say similar. Nothing I can find says anything about the length of video cards or their height as a requirement to be SFF.

I'm not going to continue debating this at length. SFF = low profile = half height. That's the way it's been for 25+ years and that's the way it's going to stay.

 

[Vayra86]

Ummmm head over to Youtube and look at comparison videos between amd and nvidia card there. on avg amd card is usually using 10% more vram then nvidia cards and depending on use of upscaler i seen it be even 20% more.

Show me then. I aint gonna go look for it Do we know this is due to different compression? Might just as well be a different memory management.

 

[Scattergrunt]

The problem is, there is no standardised definition for SFF. You have LP (low profile) which means half-height to fit into a slim case. We used to have ITX which meant single fan and not longer than the PCI-e connector to fit into a cube case. SFF is just whatever we randomly come up with. That's why Nvidia is using the term for any card that's less than 4326538 slots wide and 234956924 mm in length.

Ahh okay I understand. I was scratching my head for a second there but I get it. Probably best we leave it at that though (might be worth its own thread?)

 

[Mrgravia]

I'm not going to continue debating this at length. SFF = low profile = half height. That's the way it's been for 25+ years and that's the way it's going to stay.

So basically you declare the format by fiat but don’t seem willing to actually provide proof that that is what the format means?

Not our job to prove your argument for you btw.

 

[Rover4444]

For 5060 and 5070 it's true though. Heck, some 5060ti are down to MSRP these days. I could walk into microcenter right now and buy a PNY 5060TI 16 gig for $429. also have some $489 AIB ones in stock.

9070 and 9070XT show in stock, still at a markup, but they go very very quickly.

Maybe in your area. Been looking at the MS here pretty much daily and that hasn't been the case, at least for the 5070 Ti, 5060 Ti, and RDNA 4 cards. Also the 5060 hasn't released yet.

Pot, meet kettle.

You two are already well acquainted. If you think I lied anywhere here then I apologize.

How exactly do you define a "SFF" pc then?

If it has SFF in the name. If a "SFF" card can't fit into most "SFF" PCs because it's not half height then it's not an SFF card.

 

[Scattergrunt]

So basically you declare the format by fiat but don’t seem willing to actually provide proof that that is what the format means?

Not our job to prove your argument for you btw.

Lets not spill this over further in this thread, it is getting off topic.

 

[Vayra86]

Go look it up. Reference OEM makers like HP, Dell, Lenovo and the like.

I've already explained this above and with photo's in the thread. SFF = LP = half height expansion cards.

Nonsense. AFAIK the Lian li A3 is an sff category case and I use a 3 slot gpu in it. I think LP is a very precise term; Low Profile so clearly a flattened version of whatever part we're looking at. SFF... anything smaller than a tower might apply. Its certainly not just a NUC for example, and also not exclusively ITX cases.

The problem is, there is no standardised definition for SFF. You have LP (low profile) which means half-height to fit into a slim case. We used to have ITX which meant single fan and not longer than the PCI-e connector to fit into a cube case. SFF is just whatever we randomly come up with. That's why Nvidia is using the term for any card that's less than 4326538 slots wide and 234956924 mm in length.

Exactly this. The definition of SFF simply suits the latest fashion in parts... if there is money to be made that gen doing so. Nvidia simply redefined it along the lines of their selling GPUs and hopes the industry follows along with cases and other components.

 

[Raffles]

Yeah, today. When RT becomes a requirement, that alone increases memory usage. Games will also become more demanding naturally as time goes by. How many people buy new graphic card every year or every 2 years? I'm enthusiast and last one I bought was 5 years ago, because I couldn't justify how little they progressed and how much they cost. And I still dumped almost 1000€ into it. People who upgrade every 5 years and pay 300-400€ for graphic cards have that timeframe dramatically more stretched than I had with a high end card...

Well, your post didn't clarify that, so I could only garner that you meant today

We don't know how long that's going to take. My guess is such a long time, that the 3070 or newer cards with 8gb VRAM, the VRAM will be the least of their concern. By the time RT uniformly becomes a requirement, games will probably be demanding enough that a 3070 or 5060ti with a hypothetical 32gb of VRAM would be running stuff at 1080p30 at best. The compute and memory bandwidth will become a bottleneck before the VRAM.

To clarify, I don't think people should be rushing to buy an 8gb card today, the price difference is too close to 12gb/16gb cards, they're not cheap. And the extra VRAM can be useful for the few games that really struggle on 8gb cards, like Rift Apart.

But at the same time, people completely exaggerate how viable 8gb cards are, and almost act like they're an iGPU. "1080p low settings" is simply detached from reality.

TechPowerUp's own findings across a variety of games match my experience. 1440p60 is what you can expect today for a 3070 class card, even the min FPS is 60. And for example the 4060ti's 16gb VRAM is not enough for it to catch the 3070.

https://tpucdn.com/review/asus-geforce-rtx-5060-ti-tuf-16-gb/images/average-fps-2560-1440.png

 

[mkppo]

8GB is becoming more and more of a bottleneck over the past couple of years. Given the trajectory, expect the performance gap to widen considerably, regardless of missing textures in a year or two. We'll see where the defenders run to at that time.

 

[Mrgravia]

Exactly this. The definition of SFF simply suits the latest fashion in parts... if there is money to be made that gen doing so. Nvidia simply redefined it along the lines of their selling GPUs and hopes the industry follows along with cases and other components.

The market for sff is, if I remember correctly, a niche of a niche. They are going to focus on what makes them the most money, or at least what the majority of the DIY market is building.

Which sucks for the folks who are in the market for certain parts. Quite a few gorgeous cases that require gpu of very specific sizes or less.

For example, i'm taking a very strong look at the Mechanic Master C24, which is <10L, doesn't require any PCI-e risers, and takes up to a 135mm air cooler. Problem is it only takes up to 245mm long GPU.

That...really narrows down choices in terms of current and future announced GPU.

 

[DaemonForce]

During the first mining boom I'd say cards were regularly under $300 for AIB models and the 480 wasn't really affected until the 580 was released. They were also in stock, something these big companies are suddenly incapable of doing today. Miners really wanted the base 570s honestly, 4GB of VRAM wasn't a problem yet.

Lol no it was the same dance. In stock for maybe 15min bursts and then a week of no stock.
The week before I picked up the card:

It was work to get any decent 8GB card. It was a very different time.

How exactly do you define a "SFF" pc then? Volume of the case is the most common, so whats the cutoff in your opinion?

Anything that fits in the small of a desk or the size of a TV's set top box.
Size of an Xbox if it could have 1-2 HHHL cards installed. LOW PROFILE.
I think it's fair.

 

[Chrispy_]

$249 is arguably okay for an 8GB GPU. I'm a realist and low-end GPUs have rarely been burdened by the cost of surplus VRAM. A $249 lower-end GPU is not a long-term investment, nor is it a higher-end part that should expect to run new games with all the settings turned up, but I don't feel it's worthless, it's just definitely not worth $300 or more.

I suspect AMD will try and position it against the 5060 at $300, so I'm expecting a disappointing $279 MSRP and our brethren in the US will no doubt get scalped and tariffed to $400 or more, at which point it's another dumb purchase when 12GB and 16GB options exist for not much more.

I'm not going to continue debating this at length. SFF = low profile = half height. That's the way it's been for 25+ years and that's the way it's going to stay.

Ah yes, stubbornly stuck in the past as always. The official SFF-SIG is as defunct and obsolete as your argument.

The first page of results for "SFF PC" in Google is dominated (90% or so) by small mITX cases with SFX PSUs and compatibility with full-height, two-slot GPUs in varying lengths, typically 10.5" or so. There are a tiny handful of corporate SFFs from Dell/HP/Lenovo but their definition of SFF has moved to NUC/Thin clients using wholly proprietary form factors and lacking any expansion slots whatsoever.

The mATX/FlexATX SFF you're referring to as those corporate systems do still exist, and are still sold under the label of "SFF" alongside full-ATX sized workstations, but that's not representative of the entire SFF market, and judging from the limited number of offerings of that size compared to the fully-integrated NUC/Thin Clients, they are a dying breed of wholly proprietary parts and even if you can theoretically shoehorn a <75W slot-powered graphics card into one of those systems, there's a good chance that neither the power supply nor the case ventilation are adequate for such a thing.

You're always very keen in discussions to point out the relevance of context in your arguments, and in the context of a thread about the 9060XT 8GB - a consumer gaming card for the AIB/DIY market - there's no valid argument for trying to shoehorn an old definition of SFF limited to proprietary corporate PCs that are designed for a limited selection of upgrade parts from that same OEM, sometimes even firmware-locked to that limited selection of upgrade parts.

Low-profile is SFF, in that aspect you are correct.
SFF is not exclusively low-profile, and hasn't been for a long time, even in the corporate SFF world, so it's time to stop beating that particular drum.

 

[Red Hood]

IF the price sucks. (GDDR6 chips are pretty cheap, so it should be relatively cheaper no matter what) RT is probably not gonna be good anyway if I had to figure.

I think the bigger problem is not performance. There is a performance differences between the two, with more edge cases than vs their 40 series counterparts, but at the end of the day, its still just edge cases, the real problem is it's price. I keep hammering this point because I think there's many dishonest people misrepresenting alot of the criticism of the 8GB 5060Ti as "wow look at these idiots who wont lower their settings"

8GB for almost ~$450 USD on average (street price, not MSRP) is a joke.

I hope AMD can capitalize on this well and give a compelling price point for 8GB of VRAM.

$200 is ok for an 8GB card but as you mentioned $400+ is a scam

From what I've heard the 9060XT has good mainstream performance but 8GB VRAM would hinder that in games that use more than 8GB and at settings that the 16GB version can run fine

 

[Chrispy_]

$200 is ok for an 8GB card but as you mentioned $400+ is a scam

I'd like to be wrong, but AMD haven't officially launched a $200 GPU for over 3 years, and that was a 4GB 6500XT at the end of the pandemic+ETH GPU crisis. The median selling price of an RX6600 did quickly stabilise at around $200 once GPU production returned to normal in 2023 and 2024, but AMD were charging $279 for what is presumably a much slower GPU than the 9600XT (RX 7600) back in 2023 and they're unlikely to ask for much less than the 5060 8GB this time around. Based on the price and performance increases of the 9070XT over the 7800XT, for example, it's going to be 20% more expensive and 30% faster that that the RX7600.

AMD's strategy of undercutting Nvidia by 10% didn't work in the past, if they are undercutting Nvidia by 20% this generation (using the 9070XT@$600 MSRP) then it stands to reason that the 9060 8GB will be $240 at the bare minimum using the most optimistic, tariff-free pricing and ignoring two years of inflation and two years of rising manufacturing costs.

 

[lexluthermiester]

Ah yes, stubbornly stuck in the past as always.

The official SFF-SIG is as defunct and obsolete as your argument.

And yet every manufacturer is still using and adhering to it.. Imagine that..

But as I said;

I'm not going to continue debating this at length.

So let it go.

so it's time to stop beating that particular drum.

It sure is. The subject was over and yet you brought it up again..

 

[Red Hood]

I'd like to be wrong, but AMD haven't officially launched a $200 GPU for over 3 years, and that was a 4GB 6500XT at the end of the pandemic+ETH GPU crisis. The median selling price of an RX6600 did quickly stabilise at around $200 once GPU production returned to normal in 2023 and 2024, but AMD were charging $279 for what is presumably a much slower GPU than the 9600XT (RX 7600) back in 2023 and they're unlikely to ask for much less than the 5060 8GB this time around. Based on the price and performance increases of the 9070XT over the 7800XT, for example, it's going to be 20% more expensive and 30% faster that that the RX7600.

AMD's strategy of undercutting Nvidia by 10% didn't work in the past, if they are undercutting Nvidia by 20% this generation (using the 9070XT@$600 MSRP) then it stands to reason that the 9060 8GB will be $240 at the bare minimum using the most optimistic, tariff-free pricing and ignoring two years of inflation and two years of rising manufacturing costs.

There was a rumour about a 9050 8GB which would make sense for an entry level card but N44 yields are high so may not be enough leftovers

The 9060 XT 8GB will probably be like £300 which is an ok price for a mainstream card but 8GB is not (192 bit bus and 12GB would have made more sense imo)

I heard the 16GB version has good mainstream performance and with these high prices, a lot of people are looking for cards they can use for several years as upgrading every gen isn't feasible and worth it for most people

 

[ARF]

I understand why people want more VRAM in the midrange, but I don't understand the public cry for the release of this card. No one is pointing a gun at your head to make you buy one.

Hardware manufacturers change course when (projected or real) sales numbers force them to do so, not when people cry on a forum. As long as 8 GB cards sell well, no one cares who says what here. If you want AMD and Nvidia to offer more, you'll have to stop buying their shit first.

1. If AMD wants to improve over the old RX 6600 / RX 6600 XT / RX 6650 XT / RX 7600 / RX 7600 XT rebrand thing, they should have given more VRAM. For such a relatively faster GPU, at least 10 or 12 GB VRAM is a must.
2. This is clearly a poor design and engineering decision which means AMD will not get more sales, it will get less sales.

As someone has already rightfully said - cheap or DOA.

 

[Mrgravia]

Ah yes, stubbornly stuck in the past as always. The official SFF-SIG is as defunct and obsolete as your argument.

Indeed, group is defunct as of 2020, and seemed more focused on what most would consider now embedded, ultra compact systems then modern small form factor. A look around the web doesn’t show any major manufacturers of gpu, cases etc sticking to it. Market moved on, outside of ultra portable mini pc, which rarely accept gpu or other add on cards.

 

[AusWolf]

1. If AMD wants to improve over the old RX 6600 / RX 6600 XT / RX 6650 XT / RX 7600 / RX 7600 XT rebrand thing, they should have given more VRAM. For such a relatively faster GPU, at least 10 or 12 GB VRAM is a must.

Then don't buy it (I don't mean to disagree with you, just saying).

2. This is clearly a poor design and engineering decision which means AMD will not get more sales, it will get less sales.

Good. Maybe then, AMD (and Nvidia) will learn.

 

[Vayra86]

Indeed, group is defunct as of 2020, and seemed more focused on what most would consider now embedded, ultra compact systems then modern small form factor. A look around the web doesn’t show any major manufacturers of gpu, cases etc sticking to it. Market moved on, outside of ultra portable mini pc, which rarely accept gpu or other add on cards.

The only constant in time is change...
Also in a world where people use terms such as '2K' liberally to denote any resolution within 40% of 2000 horizontal pixels at this point... good luck getting a true definition to last more than a year lol.