AMD Announces RX 5000 Series Graphics processors at Computex - Demos RX 5700

And they're gonna charge $500 for it. Should have been a $349 part if they want any marketshare. Maybe they like less than 10% of the market? The entire gpu price spectrum is a complete shit show right now. We're paying top dollar for 4 year old tech numbers right now.
So where are you getting your pricing info? I don't see anything in any of the coverage mentioning pricing. There have been leaks and rumors, but until MSRP is announced by AMD, it's all speculation at this point.
 
I would speculate that it will be rx570 prices?.. i mean isn't it like the successor?
 
AMD is playing the 'ethical' card because their deck is that thin... to pretend than any corporation is ethical from the perspective of consumers is to live in a fantasy world.

Do note that AMD has lost GPU marketshare since the GPP was exposed.
Yeah people care about ethics, Companies and Corporations care about profit margins and returns on investment, the purchasing department for animation at Disney doesn't care about GPP, they care about what 5 million dollars in GPU's is going to get their work done fastest while looking the best with 24/7 driver & hardware support for their rendering farms / animation studios.
 
Yeah people care about ethics, Companies and Corporations care about profit margins and returns on investment, the purchasing department for animation at Disney doesn't care about GPP, they care about what 5 million dollars in GPU's is going to get their work done fastest while looking the best with 24/7 driver & hardware support for their rendering farms / animation studios.

Ah, so that is why Nv thought they could get away with monopolistic practices .
 
Ah, so that is why Nv thought they could get away with monopolistic practices .

They did it to get ahead of Intel; AMD is a fly that they don't even have to swat. Would you refrain from accusing Intel of monopolist practices?
 
Ah, so that is why Nv thought they could get away with monopolistic practices .
They didn't think they could get away with it, they know they could, and they have. Their profits are up and their market share is up and they have the fastest cards on the market with the widest range of support from software and hardware developers when you are at the top of the pack you get to charge for it.
 
So the memory that forced Fury to be limited to 4GB, caused Fury and Vega to be way too expensive, and in all but a handful of situations did fuck all for consumer and even prosumer applications was the right choice?

They took a bet.... that the volume would grow and the costs would drop. Its happened over the years with plenty of alternate ram types... or even adopting newer versions of DDR too soon.

There is zero question HBM is superior, its the cost to performance that was the issue. The bet volume would correct that and it didn't pan out.

Can't knock that to hard... it could have paid off it didn't. We don't want GPU companies to stop swinging. Reports point to Intel taking a few chances themselves with their upcoming parts. Might pan out might not. If you want an advantage you got to take a few chances.
 
Not sure what the issue is, the 5700 (or whatever the hell it’ll be called) is going to launch MSRP at 2070 prices while being +2-5% better and probably the same -1% in DX11.

It’s at least competitive in their price ranges if not better. Sure nVidia can release something superior within a few months or drop prices but that’s why this is a good thing.
 
They didn't think they could get away with it, they know they could, and they have. Their profits are up and their market share is up and they have the fastest cards on the market with the widest range of support from software and hardware developers when you are at the top of the pack you get to charge for it.
Pretty much. People only bring up ethics and monopolies when the brand they're weirdly attached to is losing. From the sound of some of them they also want a pony ride and probably candy.

The critical thing to understand is that capitalism is just another word for deathmatch.

If AMD were in NV's position, not only would they behave the same but would be legally obligated to by their shareholders - max profits über alles.
 
Last edited:
Pretty much. People only bring up ethics and monopolies when the brand they're weirdly attached to is losing. From the sound of some of them they also want a pony ride and probably candy.

The critical thing to understand is that capitalism is just another word for deathmatch.

If AMD were in NV's position, not only would they behave the same but would be legally obligated to by their shareholders - max profits über alles.

Maybe it is just plainly unfair :) and people feel that way.
You don't have to feel attached to it for your feelings to be right and if it was vice versa it would still be unfair , it is not beholden to a brand it is a situation which triggers it.

I choose to approach it from a different angle if you are that good as a company and you have faith in your products why would you need all of those tactics to begin with ...
 
Maybe it is just plainly unfair :) and people feel that way.
You don't have to feel attached to it for your feelings to be right and if it was vice versa it would still be unfair , it is not beholden to a brand it is a situation which triggers it.

I choose to approach it from a different angle if you are that good as a company and you have faith in your products why would you need all of those tactics to begin with ...

The more control of a market a company has the more that company will abuse that control and try to lock down even more of the market. When a company has real competition they can't pull those tactics because people have viable options. However, when they have no real competition then they're able to abuse their power and do whatever they want. It has nothing to do with "faith" or any of that crap and everything to do with ensuring other companies can never compete.
 
The more control of a market a company has the more that company will abuse that control and try to lock down even more of the market. When a company has real competition they can't pull those tactics because people have viable options. However, when they have no real competition then they're able to abuse their power and do whatever they want. It has nothing to do with "faith" or any of that crap and everything to do with ensuring other companies can never compete.

Yeah and that triggers things as splitting up monopolies and the whole song and dance starts again. Rather futile.
 
Yeah and that triggers things as splitting up monopolies and the whole song and dance starts again. Rather futile.

Considering that iGPUs are generally considered part of the "graphics card" market Nvidia would likely avoid any monopoly abuse charges.
 
  • Like
Reactions: Auer
like this
Pretty much. People only bring up ethics and monopolies when the brand they're weirdly attached to is losing. From the sound of some of them they also want a pony ride and probably candy.

The critical thing to understand is that capitalism is just another word for deathmatch.

If AMD were in NV's position, not only would they behave the same but would be legally obligated to by their shareholders - max profits über alles.

You must realize that you are wrong. The fact is, Nvidia is doing monopoly practices and they are not even needed. Heck, if Kyle did not let us know, we would be ignorantly unaware that GPP even occurred. But hey, guess competition is not really needed since you clearly are biased in the most obvious way.

I do not recall AMD ever doing what Intel did to them in the past nor what Nvidia is doing even now. We finally have competition going an the way you want it, we need to kill that competition before it takes root. Oh well, thankfully, you are just an arm chair stock analyst and an arm chair engineer...........
 
Considering that iGPUs are generally considered part of the "graphics card" market Nvidia would likely avoid any monopoly abuse charges.

Not quite, since there are two distinct gpu markets, discrete and iGPU's. The nice thing is that despite Nvidia and Intel's best attempts, AMD death has been greatly exaggerated. :)
 
Was about expected and not very interesting except to see what consoles get, as that can help push game design.

Arcturus/next gen free from GCN is what I have really been waiting for. Navi is just the side show and the last card from Raja too I think?
 
Not quite, since there are two distinct gpu markets, discrete and iGPU's. The nice thing is that despite Nvidia and Intel's best attempts, AMD death has been greatly exaggerated. :)

I was going to make some kind of snarky comment about how tech illiterate the courts can be and how they might not realize this but then I saw new of US courts ruling that Qualcomm has an illegal monopoly on mobile modems. So...Yeah, maybe Nvidia would have been hit with antitrust lawsuits if things had gone like they planned.
 
Was about expected and not very interesting except to see what consoles get, as that can help push game design.

Arcturus/next gen free from GCN is what I have really been waiting for. Navi is just the side show and the last card from Raja too I think?

arcturus is not next gen and never was. It was just internet making things up. Arcturus was a code name for a particular card not an architecture.
 
And even if the top card is $499 MSRP, you won't be able to get one for MSRP.
Yeah no, if the 2080TI was sub $600 you would never see one, production would be backed up and you would wait 6 months after order for delivery and you would be paying some scalper an additional $600 for the privilege.
 
So the memory that forced Fury to be limited to 4GB, caused Fury and Vega to be way too expensive, and in all but a handful of situations did fuck all for consumer and even prosumer applications was the right choice?

Annddd a lot of us at [H] saw the whole debackle coming. One thing you missed that’s even more important, it fucked their mfg capability.

So I am happy they dropped it.... hopefully the top card doesn’t use it. The last thing we need is more heat on the die. You basically have to custom cool Radeon VII to control the thermals. There’s also a problem with mixed heights between the gpu die and hbm stacks.
 
They took a bet.... that the volume would grow and the costs would drop. Its happened over the years with plenty of alternate ram types... or even adopting newer versions of DDR too soon.

There is zero question HBM is superior, its the cost to performance that was the issue. The bet volume would correct that and it didn't pan out.

Can't knock that to hard... it could have paid off it didn't. We don't want GPU companies to stop swinging. Reports point to Intel taking a few chances themselves with their upcoming parts. Might pan out might not. If you want an advantage you got to take a few chances.

This is one that I was hoping would take over. Had yields not been so difficult it could have brought sizes down, gotten memory bandwidth out of the way of performance, and simplified cooling (assuming kinks worked out).
 
  • Like
Reactions: ChadD
like this
It offers a 1.25x IPC uplift. Which is beating Vega by a LOT. Between architecture and process it has 1.5x perf/watt uplift, which is again beating Vega by a LOT.

We'll see. They only showed one game, an AMD sponsored game, against a 2070. The Navi seemed "smoother". However again, a 2070 is still within reach of a Vega64 depending on the title.

So only time will tell. I really hope they can stretch Navi's legs. I'm a fan of competition and certainly no fan of Nvidia.
 
We'll see. They only showed one game, an AMD sponsored game, against a 2070. The Navi seemed "smoother". However again, a 2070 is still within reach of a Vega64 depending on the title.

There is a difference between architecture and a specific implementation. The architecture looks like it has had a very significant improvement.

But a specific chip is NOT an architecture. Unlike CPUs where single thread performance is very important. GPUs are all about having multiple units of execution, and single unit performance is not a significant factor, so we don't have the same kind of core vs core battle to claim one architecture better.

RX 5700 winning or losing against RTX 2070 isn't a win or loss for either architecture, but just for that specific chip, more or less execution units on either side will tip the scales either way.

Navi looks like a big architecture improvement. RX 5700 will find it's place in the market.

Hopefully AMD included enough Navi execution units to be consistently ahead of the 2070. If they didn't that still doesn't make Navi a poor architecture, but a positioning issue vs the competition which will have to be addressed with lower pricing.
 
There is a difference between architecture and a specific implementation. The architecture looks like it has had a very significant improvement.

But a specific chip is NOT an architecture. Unlike CPUs where single thread performance is very important. GPUs are all about having multiple units of execution, and single unit performance is not a significant factor, so we don't have the same kind of core vs core battle to claim one architecture better.

RX 5700 winning or losing against RTX 2070 isn't a win or loss for either architecture, but just for that specific chip, more or less execution units on either side will tip the scales either way.

Navi looks like a big architecture improvement. RX 5700 will find it's place in the market.

Hopefully AMD included enough Navi execution units to be consistently ahead of the 2070. If they didn't that still doesn't make Navi a poor architecture, but a positioning issue vs the competition which will have to be addressed with lower pricing.

That said, if a huge die can only compete with a small die (on similar node), then we can all agree that the architecture is pretty inefficient
 
That said, if a huge die can only compete with a small die (on similar node), then we can all agree that the architecture is pretty inefficient

Agreed. You need to be in the same ballpark.

I did some back of the envelope calculations based on estimates of the 5700 die size shown, compared with details know about Radeon VII.

I estimate about 10.9 Billion Transistors in for the 5700 die, which is about the same as in RTX 2070. Which sounds great until you factor in that RTX is "wastes" a lot of transistors on dedicated RT HW, though we don't know if 5700 contains any dedicated AI or RT HW (though it doesn't appear so).

AMD has the potential to be more competitive than in the recent past, with NVidia chips paying the "RT Tax", and AMD not. At least until RT becomes prevalent enough that AMD must also include RT HW.
 
Agreed. You need to be in the same ballpark.

I did some back of the envelope calculations based on estimates of the 5700 die size shown, compared with details know about Radeon VII.

I estimate about 10.9 Billion Transistors in for the 5700 die, which is about the same as in RTX 2070. Which sounds great until you factor in that RTX is "wastes" a lot of transistors on dedicated RT HW, though we don't know if 5700 contains any dedicated AI or RT HW (though it doesn't appear so).

AMD has the potential to be more competitive than in the recent past, with NVidia chips paying the "RT Tax", and AMD not. At least until RT becomes prevalent enough that AMD must also include RT HW.
https://www.engadget.com/2019/05/27/amd-radeon-rx-5000-ray-tracing-navi/

I have a feeling, though, that any dedicated hardware support will be limited this gen. They should be working on a driver by now, i would think.
 
AMD has the potential to be more competitive than in the recent past, with NVidia chips paying the "RT Tax"

Remembering that Nvidia is on an older, cheaper process, so cost per die may be similar (we'll never know). If Nvidia wanted to sling heat directly at Navi, they could just upscale the 1660 to 1080 performance on the new node. Might call it a GTX1670 or some such, but generally speaking, they're sitting pretty good with the 2070.
 
Remembering that Nvidia is on an older, cheaper process, so cost per die may be similar (we'll never know). If Nvidia wanted to sling heat directly at Navi, they could just upscale the 1660 to 1080 performance on the new node. Might call it a GTX1670 or some such, but generally speaking, they're sitting pretty good with the 2070.

Might be pretty close. Which is why I chose to concentrate on transistor count. Historically, the new process is more expensive per mm2, but less expensive/transistor, so with similar transistor count AMD should be paying less/die (even if only a little less). But it is probably only a little less. But at least it isn't more and this time AMD is not relying on more expensive HBM.

So AMD is definitely in a better position than they were with Vega having both a more expensive die, and more expensive memory technology.
 
Might be pretty close. Which is why I chose to concentrate on transistor count. Historically, the new process is more expensive per mm2, but less expensive/transistor, so with similar transistor count AMD should be paying less/die (even if only a little less). But it is probably only a little less. But at least it isn't more and this time AMD is not relying on more expensive HBM.

So AMD is definitely in a better position than they were with Vega having both a more expensive die, and more expensive memory technology.

7nm is insanely expensive

AMD-Wafer-Price.png
 
You figured out the area that is reduced versus the rise in cost ?

Here is the most direct comparison I can find:

Radeon VII has 5.6% more transistors than Radeon RX Vega 56/64

die size only decreases by 31.89%
 
Last edited:
7nm is insanely expensive

View attachment 164057


Do they have cost/transistor?

Here is an example of the whole picture:
main-qimg-4bed7def49e8bb0a8e9a596c79a093e3.png


Wafer costs just keep going up, but so does density, so in the end cost per transistor keeps going down. Though it is flat lining.

It should still be a little cheaper to build equal transitor counts on 7nm than 14nm.
 
Last edited:
  • Like
Reactions: B770
like this
I don't care what the excuses are, we're dealing in what is. For selfish reasons I would've preferred more emphasis on GPU development because CPUs have blown past the point of "good enough" years ago. Whoopdy do, 15% faster than what was already overkill, with 4 more cores than was already overkill -- let's shoot our guns in the air and fire up our video transcoders just to have something to do.

"Wait for Navi, wait for Navi" is the cry we've been hearing for eons. And its a mousefart, right on cue.

Wait until Intel rolls into town swinging it's dGPU dick around. AMD will end up just a pure CPU company at that point. Can't blame them though and I understand the business reasons.
I wouldn't blame AMD for shit GPU launches either. When you offer superior products with better performance at a lower cost and still get outsold 6 to 1 by Nvidia (RX 570 Vs GTX 1050/ti) what do you expect.
 
This is really a thing?

Saying that a company with a superior and cheaper product is not to blame for it's own poor sales..?

Seriously. No wait, don't answer..
 
I wouldn't blame AMD for shit GPU launches either. When you offer superior products with better performance at a lower cost and still get outsold 6 to 1 by Nvidia (RX 570 Vs GTX 1050/ti) what do you expect.

The RX 570 being cheaper than 1050 Ti is only a recent post crypto-bubble, inventory clearing situation.

The 1050 Ti launched at $140, the 570 at $170. Also the 1050 kept the same name throughout, 570 is rebranded/refreshed 470 splitting the numbers further.
 
  • Like
Reactions: Auer
like this
arcturus is not next gen and never was. It was just internet making things up. Arcturus was a code name for a particular card not an architecture.
Looks like you are right looking again, Arcturus if it's 2020 it won't be anything new. If RNDA is the 'departure from GCN' then we will see, it appears as AMD has split the GPU divisions properly in this manner, the die for 5700 is pretty small, weird being back at that number range again too.

P.s. seen at least two shops with your name in EU before, you are a popular guy. Missing the V64 though.. should have pulled it and taken with.

And imagine what happens when AMD rebrand numbers go back to AMD early staff after ATi reign. Still in ATi...

Radeon HD 5700
The codename for the 5700 GPU was Juniper and it was exactly half of Cypress. Half the shader engines, half the memory controllers, half the ROPs, half the TMUs, half everything. The 5750 had one shader engine disabled (of 10), so had 720 stream processors, while the 5770 had all ten enabled. Additionally, the 5750 ran at 700 MHz and a lower voltage, while the 5770 used more power, but ran at 850 MHz. Both cards were normally found with 1 GB of GDDR5 memory, but 512 MB variants did exist, performance suffering somewhat.
Imagine selling them as 5700 on fleabay....
 
Back
Top