Had this discussion with some posters on a different forum. Woulda been nice if AMD had been able to have a scaled up ~450mm2 Polaris coming in at 4608sp. Understandably its not just a matter of doubling everything.. even identical chips require respins on different processes. However a chip of this size would have worked well against the 1080. It just seems that AMD simply didn't have the manpower available to draft a larger chip based on Polaris.The underlying problem here wasn't the mining sales, but the fact that Polaris didn't bring any raw performance improvements to the table. It offered some new features (e.g. HEVC decoding and HDMI 2.0 output), and better perf/watt than 28nm, but didn't really game any better than Hawaii. AMD kept making 28nm Fiji for a while despite its shortcomings, because it still outperformed Polaris and was the only thing they had that could even attempt to compete with the GTX 1070.
If AMD had a "big" Polaris card,, then there would have been more sales because there would have been a meaningful upgrade path. The lack of R&D funding has hurt the GPU division greatly.
At 400W.Had this discussion with some posters on a different forum. Woulda been nice if AMD had been able to have a scaled up ~450mm2 Polaris coming in at 4608sp. Understandably its not just a matter of doubling everything.. even identical chips require respins on different processes. However a chip of this size would have worked well against the 1080. It just seems that AMD simply didn't have the manpower available to draft a larger chip based on Polaris.
At 400W.
No, they needed a new architecture. Hopefully that is Vega.
Is that what's going on. I was looking at NV GFX cards and noticed the 1070s had gone up in priceHardly... now the mining hysteria is contagious and it affected nVIDIA too. And seems that nVIDIA will get affected eventually.
Wondering if Intel will take part of that....
The 1070 is at least in stock and compares favorably in terms of outright hash rate and especially performance/watt in Ethereum vs Polaris.Is tha what's going on. I was looking at NV GFX cards and noticed the 1070s had gone up in price
The problem?The 1070 is at least in stock and compares favorably in terms of outright hash rate and especially performance/watt in Ethereum vs Polaris.
Had they simply taken Fiji performance and added what they could with the 14nm power savings (30-40%) they would have at least matched 1080. Neither here nor there now, even though this chip was feasible they simply didn't have the manpower to get it done.At 400W.
No, they needed a new architecture. Hopefully that is Vega.
Had they simply taken Fiji performance and added what they could with the 14nm power savings (30-40%) they would have at least matched 1080. Neither here nor there now, even though this chip was feasible they simply didn't have the manpower to get it done.
Well, they needed to move the architecture forward. They didn't have the manpower to both port a highend GPU to 14FF, develop Polaris and do a major architectural overhaul of GCN at the same time. Losing revenue to Nvidia for a couple of generations has consequences, sad to say.
On the other hand, a hypothetical polaris based GPU that accommodates 48CU, 384bit memory interface (something that maintained a TDP of <225W) might have been good for them to atleast have an AMD alternative at the low/mid high end instead of completely giving away the high end for one financial year (and counting) to its competitor. A 20~30% (or higher?) performing part wouldn't have been too bad imo but such is reality that at the end of the day I guess it all came down to ROI given limited resources.
Now that Vega has been confirmed to be indefinitely delayed,
On the other hand, a hypothetical polaris based GPU that accommodates 48CU, 384bit memory interface (something that maintained a TDP of <225W) might have been good for them to atleast have an AMD alternative at the low/mid high end instead of completely giving away the high end for one financial year (and counting) to its competitor. A 20~30% (or higher?) performing part wouldn't have been too bad imo but such is reality that at the end of the day I guess it all came down to ROI given limited resources.
Confirmed? Source?
Miners are going to screw this one up as bad as they did Hawaii, but still. Where did you hear that?
I said long before Computex that Vega has been indefinitely delayed.
Only it's not. Delayed yes, but not "indefinitely". There is release date for Vega FE, and there is a release window for RX Vega.
Radeon Vega FE don't count. It would be like NVIDIA saying that Volta is ready out since V100 GPU Accelerator has been released.
AMD said that it will "talk more" about Radeon RX Vega at SIGGRAPH for a very specific reason: AMD doesn't know when the product will be ready.
Notice how the actual launch date for Radeon RX Vega was not mentioned.
If Radeon RX Vega is ready in time for SIGGRAPH, AMD will launch Radeon RX Vega at the event.
If Radeon RX Vega is not ready for SIGGRAPH, AMD will only be making an announced as to when Radeon RX Vega should be available.
In addition, AMD showcased the highly-anticipated Radeon™ RX Vega graphics card, asserting that the "Vega" architecture based gaming powerhouse is expected to launch at SIGGRAPH 2017
The negative to this strategy is to always fall further behind. Dying slowly following a rearguard action. There is no way AMD can compete with Nvidia, [Intel also, for that matter] using traditional practices. They arrived at another one.Had they simply taken Fiji performance and added what they could with the 14nm power savings (30-40%) they would have at least matched 1080. Neither here nor there now, even though this chip was feasible they simply didn't have the manpower to get it done.
It appears to be increasingly obvious that they are going small die tied together with a very energy and latency efficient network. The old ways could not work well here and being R&D constrained, we get a piecemeal introduction of the tech.Well, they needed to move the architecture forward. They didn't have the manpower to both port a highend GPU to 14FF, develop Polaris and do a major architectural overhaul of GCN at the same time. Losing revenue to Nvidia for a couple of generations has consequences, sad to say.
The negative to this strategy is to always fall further behind. Dying slowly following a rearguard action. There is no way AMD can compete with Nvidia, [Intel also, for that matter] using traditional practices. They arrived at another one.
By arrive, I mean another strategy. I never meant they have finished the journey.I don't think AMD GFX has 'arrived' at another point yet. They are paving the way - but NV is a moving target with a huge treasure chest. From what I've seen so far, TRG needs to be executing much better than they are. Honestly, what I hope for at, at this point, is that Ryzen and the Ryzen based APU will take up all the production capacity at GF. Then TRG (the radeon group) can move back to TSMC. I think this would be the best play for them (doesn't seem like GF has done as well executing on large ASICs).
As far as the multi-die strategy, so long as NV is able to stay ahead of AMD with ~400mm2 GPUs, AMD will have a tough time grabbing back market share. The engineering for stitching together two dice with some fast and wide fabric will be very challenging, IMHO. Unless IF scales up really to that level, AMD will need some customer interconnect system, and even then, there will be latency issues with off-die communication of shared info.