Radeon RX 7900 XT PCB Design leaks, reveals AMD's chiplet design, VRAM configuration

[WYP]

This GPU requires three 8-pin PCIe power connectors.



Read more about AMD's Radeon RX 7900 XT PCB leaks.

 

[AlienALX]

Awesome. So this must be the big boy.

Good they are sticking to standard connectors and not forcing you to use an ass whart for power.

 

[Bagpuss]

Unless RT performance can at least match the 4xxx series, which I severely doubt, then AMD won't even enter into my consideration for my next GPU.


I think the best we can hope for is that they match the 3xxx cards, or slightly exceed them, in that department.


Its a shame, as I really don't want to give Nvidia any more of my money but I equally don't want to have to turn down the settings knowing I wouldn't have to if I'd stuck with Nvidia.

 

[Warchild]

Awesome. So this must be the big boy.

Good they are sticking to standard connectors and not forcing you to use an ass whart for power.

Nothing wrong with the new connector. I like the idea of one vs three. Its the ass-position they keep putting it in right now, that I have problems with.

Why oh why would you put the connector in the middle of the card, right where the cable etc is in full view, and most likely covering any text/design...

 

[KingNosser]

I honestly feel these cards will either break even or trounce, but just like Nvidia AMD is going to want a pretty penny for them, if they do keep pricing similar to before then Nvidia you got problems i feel as AMD have much more TSMC capacity.

I expect the DXR performance will be improved but by how much anyone can guess, saying that i'll be on my 6800xt for at least one or two more gens yet at which point 3nm or lower will be a reality and beyond that chiplet is the only way.

Do feel Nvidia are going to be in a pickle unless they sort that out sharpish.

 

[AlienALX]

This one will be a pretty penny, yes. Because of materials, and AMD's increase in profit margin since Polaris. Where Raja had the superb idea of a land grab. Sadly when he left his good ideas left with him, and AMD got greedy.

It's not this one you want to be interested in. It's the smaller, potentially much cheaper, mid range cards.

 

[KingNosser]

I'm intrested in the performance and ark changes outside of that they can both whistle for a few gens, I have GF's college bills to pay atm lol

Well AMD may have got greedy but they learnt from the best of them didn't they

If this is AMD's maxwell GPU ryzen moment, then i'd expect Nvidia to be getting the loo roll ready

 

[AngryGoldfish]

This one will be a pretty penny, yes. Because of materials, and AMD's increase in profit margin since Polaris. Where Raja had the superb idea of a land grab. Sadly when he left his good ideas left with him, and AMD got greedy.

It's not this one you want to be interested in. It's the smaller, potentially much cheaper, mid range cards.

Radeon has excelled since Raja left. Not because Raja left necessarily. But they reached higher and higher each generation post his exit. I still think they priced themselves too high with RDNA2, but in the end it turned out to be the most beneficial for them financially, which we will see a return on with higher R&D budgets.

 

[NeverBackDown]

Radeon has excelled since Raja left. Not because Raja left necessarily. But they reached higher and higher each generation post his exit. I still think they priced themselves too high with RDNA2, but in the end it turned out to be the most beneficial for them financially, which we will see a return on with higher R&D budgets.

He definitely set up their future designs. I'd say he probably laid out through RDNA2 as well as CDNA. These architectures are laid out years in advance. He's only been gone a few. His blueprints are more than likely all over current released designs.

 

[AlienALX]

Well I will politely disagree dude. They haven't excelled IMO.

They started to excel, and IMO were on track to. I know a lot of people diss Polaris but IMO it was the best thing to come out of AMD since the 5000 series. I know, people will say I am stupid and Navi has been much better than Polaris but Navi is too expensive. And the cheaper GPUs lower down the Navi stack are crap.

After Maxwell Nvidia just blatantly started taking the . Polaris was a return to sanity and normality for a lot of PC gamers. Then Raja leaves and AMD just say "Oh well, if you can't beat 'em join 'em!" and started taking the equally.

If PC gaming is to thrive you need more people on the platform. And not just because it's the in thing kids want to do.

I am really hoping that AMD go back to far more affordable cards soon. They don't have to be world beating, just affordable.

Because it's clear that it isn't that Nvidia see in the future.

 

[AngryGoldfish]

Well I will politely disagree dude. They haven't excelled IMO.

They started to excel, and IMO were on track to. I know a lot of people diss Polaris but IMO it was the best thing to come out of AMD since the 5000 series. I know, people will say I am stupid and Navi has been much better than Polaris but Navi is too expensive. And the cheaper GPUs lower down the Navi stack are crap.

After Maxwell Nvidia just blatantly started taking the . Polaris was a return to sanity and normality for a lot of PC gamers. Then Raja leaves and AMD just say "Oh well, if you can't beat 'em join 'em!" and started taking the equally.

If PC gaming is to thrive you need more people on the platform. And not just because it's the in thing kids want to do.

I am really hoping that AMD go back to far more affordable cards soon. They don't have to be world beating, just affordable.

Because it's clear that it isn't that Nvidia see in the future.

So architecturally you feel that AMD has not excelled?

Raja was an engineer. As NBD said, he likely designed the blueprints for these generations. However I can't imagine it was his decision to decide what to price them at. So we're talking about engineering here mostly. And RDNA1 and RDNA2 were engineered well. I don't see how that can be denied. You yourself happily own a 6800XT because you found one at a good price. It's a good card.

So your issue is the price. But that's not what I'm talking about. I literally said, "I still think they priced themselves too high with RDNA2." So you're not disagreeing with me; you're agreeing with me. The architecture and the designs were quite good; they just priced them too high to appeal to everyone. AMD are probably glad they did that though and followed after Nvidia because it earned them vast amounts of money.

That tells me that AMD excelled at two things: earning a lot of money and making a good product that could reach a wide audience (if it weren't for mining). Those are two of the biggest factors attributed to a successful company. The other is definitely availability and appeal to budget gamers, the biggest market by numbers. And they fell short in those areas.

But then look at Polaris. What did Polaris do for AMD? They made a good product that fit one niche (it was the biggest) and priced it affordably with great supply. But did it earn them a lot of money? I can't imagine it earned them more money than RDNA2 did. Did it reach a huge audience? No. So some of the other major factors were missing. It did allow AMD to seep into more systems (the RX 580 is AMD's most popular card according to Steam surveys) but that's only part of the business puzzle. AMD has sold more RDNA2 cards than Polaris cards, appealed to a wider user base, and offered greater efficiency. Those are huge parts of the puzzle

RDNA2 was engineered well. It appealed to a wide audience. It sold incredibly well and made AMD a lot of money. It was efficient. That to me is a success.

Did it cater to the largest gamer niche? No. AMD priced their cards too high. Was it available? No because of mining and other factors. But that doesn't change what I originally said.

 

[AlienALX]

So architecturally you feel that AMD has not excelled?

Nope ! I totally do not.

In fact, had Nvidia not had an Ampere moment AMD would have been screwed. Imagine Ampere at 500mhz faster. Which is what AMD were last round - 500mhz faster. Which just about managed to keep them about level, or a bit higher in places, but much lower in terms of RT and features on Nvidia.

I guess it depends as what you see as a great product. I see it as accessible, affordable yet offering good features and performance. Something big Navi did not achieve.

It was a good product, put up the stack because Nvidia F-ed up.

I was hearing really bad things about AMD next round. For the obvious reasons. Nvidia were not about to mess up like that again, and thus Navi III or whatever you would want to call it would trail badly. That was all I was hearing until I heard about this "Maxwell moment" they were about to have. And that was about 3 weeks ago.

If they manage to make a mid range cracker for peanuts? that will be great. That is why I will always see Polaris as great, because it was. It got more people PC gaming on a shoestring budget. These days the entry level price is like £300 lmao. A tiny bit less than a console, for worse performance.

Polaris did not compete at the high end, but it did not need to. It started the ball rolling for more products of the same ilk to bring in more market share. Once you have that market share you can begin delivering tank cards, and maybe then people will buy them.

As we all know (and has been statistically proven if you check Steam out) the bread and butter for any company NEVER comes from god tier products. Never ever ever. So OK, Nvidia made 60% on the 3090Ti. Big bananas, not. They also made 60% on the 3060 and would have sold them in bucket loads. In fact, in a year or so I will have a fiver the 3060 will be the most used card on Steam. Because for ages and ages it was the 1060, and the 580 and 570.

That is how AMD need to compete. Not by being an asshole equal to the size of Nvidia.

 

[AngryGoldfish]

Nope ! I totally do not.

In fact, had Nvidia not had an Ampere moment AMD would have been screwed. Imagine Ampere at 500mhz faster. Which is what AMD were last round - 500mhz faster. Which just about managed to keep them about level, or a bit higher in places, but much lower in terms of RT and features on Nvidia.

I guess it depends as what you see as a great product. I see it as accessible, affordable yet offering good features and performance. Something big Navi did not achieve.

It was a good product, put up the stack because Nvidia F-ed up.

I was hearing really bad things about AMD next round. For the obvious reasons. Nvidia were not about to mess up like that again, and thus Navi III or whatever you would want to call it would trail badly. That was all I was hearing until I heard about this "Maxwell moment" they were about to have. And that was about 3 weeks ago.

If they manage to make a mid range cracker for peanuts? that will be great. That is why I will always see Polaris as great, because it was. It got more people PC gaming on a shoestring budget. These days the entry level price is like £300 lmao. A tiny bit less than a console, for worse performance.

Polaris did not compete at the high end, but it did not need to. It started the ball rolling for more products of the same ilk to bring in more market share. Once you have that market share you can begin delivering tank cards, and maybe then people will buy them.

As we all know (and has been statistically proven if you check Steam out) the bread and butter for any company NEVER comes from god tier products. Never ever ever. So OK, Nvidia made 60% on the 3090Ti. Big bananas, not. They also made 60% on the 3060 and would have sold them in bucket loads. In fact, in a year or so I will have a fiver the 3060 will be the most used card on Steam. Because for ages and ages it was the 1060, and the 580 and 570.

That is how AMD need to compete. Not by being an asshole equal to the size of Nvidia.

Is it really reasonable to say that RDNA2 was not a good architecture because Ampere wasn't blisteringly fast? Was Zen a poor architecture because Intel had nothing to counter it? Comparisons only make sense when comparing reality against reality. Even then it doesn't make sense. Was Maxwell a poor architecture because Pascal was better? The reality is, Ampere was not 500Mhz faster. Nvidia chose to use Samsung's inferior node for reasons that only they really know. We can all stipulate, but ultimately we're on a fly on the wall of Jensen's meeting room. Whatever reason is irrelevant. We all used to say it was Intel's greed that made them stop innovating until Zen came along. But actually, a large reason why they stopped innovating was because 10nm, which was supposed to come out in 2017, was delayed by years.

I agree that, as a consumer, I would prefer it if AMD focused on the low to high-end of the market with affordable, accessible, efficient no-thrills GPU's that don't chase after the bleeding edge aimlessly. But I'm just an average consumer. I'm a consumer trying to see architectural design from other perspectives. RDNA2 is a very solid architecture in my opinion that was priced too high for me. However, if it weren't for the mining boom and all the rest of it, it would have served a lot of gamers well. It's serving you well.

 

[AlienALX]

Is it really reasonable to say that RDNA2 was not a good architecture because Ampere wasn't blisteringly fast? Was Zen a poor architecture because Intel had nothing to counter it? Comparisons only make sense when comparing reality against reality. Even then it doesn't make sense. Was Maxwell a poor architecture because Pascal was better? The reality is, Ampere was not 500Mhz faster. Nvidia chose to use Samsung's inferior node for reasons that only they really know. We can all stipulate, but ultimately we're on a fly on the wall of Jensen's meeting room. Whatever reason is irrelevant. We all used to say it was Intel's greed that made them stop innovating until Zen came along. But actually, a large reason why they stopped innovating was because 10nm, which was supposed to come out in 2017, was delayed by years.

Pascal was a shrunk Maxwell. That is why it was so good. Because it did away with all of the fat and grease and got down to what mattered - raster performance. And high clock speeds, and efficiency ETC. Turing was that with some of their new tat bolted to it. That is why it was nowhere near as impressive as Pascal, because it was Pascal. With every so slightly higher clocks. A refresh, if you will, with some stuff bolted to it.

Ampere was supposed to be released long before it did. And it was not going to be Samsung it was going to be TSMC. Somewhere Nvidia dragged their knuckles and must have gone to Samsung a while after.

The part you miss out in the above is that AMD's development budget is a mere fraction of what Nvidia invest. It is far less in both scope, size and etc than what Nvidia spend. We covered this a while back when I was still angry that AMD had the audacity to charge what Nvidia charge. Let me give you a comparison, even though they are usually all bunk.

You walk into a store and you see a pair of genuine leather shoes. They are £100. Right next to them you see faux leather shoes, they also cost £100. What would you buy? I mean OK look if you are a vegan or one of those sorts you would maybe pay £100 for the pleather ones.

The problem is, as I stated numerous times when hunting down a new GPU, AMD's GPUs are not worth the same prices as Nvidia. Because even though you may not want the features, or they do absolutely nothing for you, they are still there. Even though you may never play a ray traced game that hardware is still on the Nvidia GPU. Even though you may never use deep learning, DLSS or RTX voice? they are all still there. So my argument, nae anger, was that AMD were charging exactly the same prices as Nvidia for less of those features (IE practically none of them).

Zen? they paid up for. Trust me on that. It's a superb design, and superb technology, and Keller would have made bank for the design. In fact it's so good Intel have now borrowed their glue pot. Imitation and all that. Its problem has been clock speed, but that seems to have been fixed now. Had Intel got their finger out of their ass? AMD would have had to sell for less. I wouldn't say it was all luck though, because as I say Zen is a terrific feat of engineering.

RDNA 2, IMO, isn't. So therefore, given it costs less to make, cost less in R&D and so on? it should cost less. To some degree they should pass those savings onto customers. But they didn't. Even their FE card was mostly plastic. Nvidia's was machined from billets. Again, that saving should have been passed onto customers but it wasn't.

Now with Ada? well as you will know full well I knew exactly what it would cost. All you had to do was go back to Turing. TSMC, +Nvidia = expensive.

I agree that, as a consumer, I would prefer it if AMD focused on the low to high-end of the market with affordable, accessible, efficient no-thrills GPU's that don't chase after the bleeding edge aimlessly. But I'm just an average consumer. I'm a consumer trying to see architectural design from other perspectives. RDNA2 is a very solid architecture in my opinion that was priced too high for me. However, if it weren't for the mining boom and all the rest of it, it would have served a lot of gamers well. It's serving you well.

If you ever asked me what I considered to be the all time greats? they would all be cards that every one could access. That offered sublime value for money and etc. Again, think back through my previous postings and opinions, namely that of Polaris. In fact, in the post before this one I said how impressive I thought that was.

Maybe with these prices the way they are more people can begin to understand why I think that. I also think the mid range belters on Maxwell (namely the 970* and 980) were about some of the best gaming cards ever released.

*OK so there was the missing .5gb of VRAM, but who cares? it made hardly any difference then and it was a belting value for money.

If AMD want to impress me they simply have to go against the grain and stop emulating Nvidia. Because Nvidia's nastiest most displeasing trait is their blatant f*****g greed. And their way of making cards for data centres and spending money hand over fist on crap gamers will never use or want, but making sure they charge them for it any way.

You know? it's kinda like when companies like BT increase their charges to put in new phone lines and internet for other people. Why the heck am I paying for that? because their company is pure profit. They want me to pay extra so someone else can have the internet.

As it happens right now AMD have respectfully (if you could call it that) dropped their prices. And they are now priced where I consider they should be. So fair play for that. Bit late to turn a overpriced card into a great, but we are getting somewhere.