• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

[Rumor, Tweaktown] AMD to launch next-gen Navi graphics cards at E3

Page 144 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Status
Not open for further replies.
These roadmaps take a long time to materialize and AMD would have gotten the TSMC roadmap a long time ago about ramp, availability, timeline etc. and they need to order wafers then, not now. Take Huawei for example, they overprovision and book wafers and then cancel them if they cannot move that many units.
For NVIDIA, they design with Samsung and booked wafers there (for several reasons) and when roadmaps don't materialize plans go for a toss. Moving to TSMC cannot happen overnight and there is a long tapeout and industrialization lead time.
Not sure what will happen.
But I think there is a probability of Orin's 2021 time frame being related to this.
 
Another issue is the present AMD being able to pay higher prices for wafers if fabbing CPU chiplets, than if the same wafer is being used for a GPU, not including the monster compute beasts. They can be an earlier TSMC customer than previously possible.

Does this mean that with AMD now competitive in CPU performance, we will see them having 1st dibs on the latest node wafers going forward. I found it very strange that they were able to command so much of the increased 7nm wafer supply soon coming to the market. If true, then Nvidia has some strategizing to do as this could be significant for future competitiveness.
 
Another issue is the present AMD being able to pay higher prices for wafers if fabbing CPU chiplets, than if the same wafer is being used for a GPU, not including the monster compute beasts. They can be an earlier TSMC customer than previously possible.

Does this mean that with AMD now competitive in CPU performance, we will see them having 1st dibs on the latest node wafers going forward. I found it very strange that they were able to command so much of the increased 7nm wafer supply soon coming to the market. If true, then Nvidia has some strategizing to do as this could be significant for future competitiveness.

I would not doubt that part of AMD's strategy is to tie up as much 7nm capacity as they can to make it harder for nVidia to catch up. And with AMD CPU's selling like hot cakes, it gives them the upper hand in wanting additional capacity for their GPUs.
 
I would not doubt that part of AMD's strategy is to tie up as much 7nm capacity as they can to make it harder for nVidia to catch up. And with AMD CPU's selling like hot cakes, it gives them the upper hand in wanting additional capacity for their GPUs.

Catch up?

Nvidia have stable drivers, and higher performance across the board. Of course they cost more, though people are willing to pay more for stability and performance.

When NV eventually get 7nm GPU's out, they'll completely crush the shambles that is Radeon.

AMD need to pump far more R&D into GPU's, hopefully the CPU success will enable this.
 
Catch up?

Nvidia have stable drivers, and higher performance across the board. Of course they cost more, though people are willing to pay more for stability and performance.

When NV eventually get 7nm GPU's out, they'll completely crush the shambles that is Radeon.

AMD need to pump far more R&D into GPU's, hopefully the CPU success will enable this.
By the time Nvidia will get to 7 nm on consumer cards AMD will be on the brink of releasing RDNA3 GPUs...
 
Catch up?

Nvidia have stable drivers, and higher performance across the board. Of course they cost more, though people are willing to pay more for stability and performance.

When NV eventually get 7nm GPU's out, they'll completely crush the shambles that is Radeon.

AMD need to pump far more R&D into GPU's, hopefully the CPU success will enable this.

Why do you keep insisting on the fallacy of "better NV drivers"?
Their drivers are NOT better, they have their own share of issues.
Oh, and "better performance" is only at the upper end (>$500, RTX 2070 super and above)
At almost every other price point AMD has better performance.
 
Do people still think nVidia drivers are (way) more stable than AMD's? Literally every new nVidia's architecture product was launched with bunch of issues, and took some time to be solved, just like AMD's. IMHO, it is better be early adopter who buys RX 5700XT for 400€ and have to live with buggy drivers for few months, than early adopter of RTX 2080 for 800€ and have to live with buggy drivers for few months. And it's also quite common for nVidia that new driver version breaks something that worked in previous version. Plus their Forceware is maybe even bigger bloat than Catalyst/Adrenalin (or however it's called now)

When it comes to bang for the buck, again, AMD offers better deal in every price range they cover. Which is 100-400€ at the moment, i.e. ~90% of the dGPU market volume

I don't like Navi and Turing prices, so I have recently bought RX 570 and WX2100 (3yo architecture). If I wanted to spend more money, my choice would be Vega 56. I really don't care about power consumption, temperature, noise, RGB b0ll0cks, etc. Only about price and performance. Though both cards have quite good PC, temp, noise, looks, etc
 
Last edited:
Go check the AMD reddit. The 2020 drivers have had the worst reception in years. Countless threads and posts all confirming driver issues.

Navi 5700 cards still having issues since launch, with black screens, downclocking and stuttering.

I'm a long term AMD GPU owner, currently rocking a Radeon 7, but even I am not big enough of a fanboy to admit AMD's drivers have gone way down hill since Navi launch.

More evidence you folk can't deny >

I suppose you'll next resort to accusing Steve of being an NV shill. Sign, not sure why I'm bothering to reply to trolls, as will fall on deaf ears.

Trolling and name calling is not allowed. -Shmee
 
Last edited by a moderator:
Catch up?

Nvidia have stable drivers, and higher performance across the board. Of course they cost more, though people are willing to pay more for stability and performance.

When NV eventually get 7nm GPU's out, they'll completely crush the shambles that is Radeon.

AMD need to pump far more R&D into GPU's, hopefully the CPU success will enable this.
You're totally missing the sub-discussion.

If AMD can out compete Nvidia for wafers on a leading edge node, then they will have to play catchup, unless you're asserting that Nvidia will always have an architectural advantage going forward to compensate for this node deficit. A big assumption.

Right now, Apple, because of their superior ASP can outbid anyone outside of exotics (Xilink, etc) for leading edge wafers. AMD, due to their success in server CPUs ASP, have now risen to 2nd tier. They now have the ability to leverage this for their lower margin products like GPUs & APUs.
 
Catch up?

Nvidia have stable drivers, and higher performance across the board. Of course they cost more, though people are willing to pay more for stability and performance.

When NV eventually get 7nm GPU's out, they'll completely crush the shambles that is Radeon.

AMD need to pump far more R&D into GPU's, hopefully the CPU success will enable this.

You are not understanding the conversation. The 'Catch Up' is in reference to what node nVidia and AMD are on. AMD gets priority over nVidia because AMD has all of their CPU's and new GPU's on 7nm. So by the time nVidia gets to 7nm, AMD will most likely be on to 5nm or one of the more advanced 7nm processes.
 
By the time Nvidia will get to 7 nm on consumer cards AMD will be on the brink of releasing RDNA3 GPUs...

You are not understanding the conversation. The 'Catch Up' is in reference to what node nVidia and AMD are on. AMD gets priority over nVidia because AMD has all of their CPU's and new GPU's on 7nm. So by the time nVidia gets to 7nm, AMD will most likely be on to 5nm or one of the more advanced 7nm processes.

You're totally missing the sub-discussion.

Right now, Apple, because of their superior ASP can outbid anyone outside of exotics (Xilink, etc) for leading edge wafers. AMD, due to their success in server CPUs ASP, have now risen to 2nd tier. They now have the ability to leverage this for their lower margin products like GPUs & APUs.

The underlying assumption being that AMD will be able to constantly outbid Nvidia for leading edge node access. Given that AMD is cash burning for 2019 YTD while Nvidia has ~$3B FCF for the same year, AMD doesn't have the cash flow necessary to sustain such a lead.
 
The underlying assumption being that AMD will be able to constantly outbid Nvidia for leading edge node access. Given that AMD is cash burning for 2019 YTD while Nvidia has ~$3B FCF for the same year, their respective financial wherewithal suggests that this is an unrealistic assumption.
Not at all.

Whatever price you pay for a node HAS to take into account the prices of the final product using that node. That's why GPUs are never the 1st product to use a node. That falls to higher margin dies. Mobile, especially Apple, is the main example of this. The argument is that now AMD has such a product, namely the server die chiplet. In this case AMD will have a greater ability to pay more for wafers and they will then leverage their purchasing strength to pass this advantage to the lower margin dies, APUs & consumer GPUs.

I hope you realize that AMD is probably making as much, if not more margin on a server chiplet as Apple makes on an Iphone SOC. Outside of the compute GPU dies, which in wafer terms is not the majority of their demand, Nvidia cannot pay as much.

Nvidia's main product, volume wise, is their consumer GPUs, which is a price sensitive product, so they do not have the same ability to pay whatever for the wafers. This is playing out right now in 7nm and my thinking is that this might be the new normal. This has nothing to do with cash reserves, but cash flow and margins on a product.

The reduction in Fab companies capable of leading edge node production has brought us here. TSMC, Intel and an apparently struggling Samsung. No where else to go.

I think this is a very big deal for the future and worthy of more thinking.
 
Guys, since this thread gas gone way off topic, and many people have decided to ignore my warning, this thread is now closed for review and infractions will be handed out as necessary.
 
Status
Not open for further replies.
Back
Top