Ok. Two things need to happen now.
1. Provide a BOM for both AMD FuryX and Nvidia 980Ti.
2. Provide credible reasoning why said BOMs for each allows, or prohibits said GPU from being competitive at a 649.00 price point.
Otherwise you guys will go round robin for eternity.
1. Provide a BOM for both AMD FuryX and Nvidia 980Ti.
2. Provide credible reasoning why said BOMs for each allows, or prohibits said GPU from being competitive at a 649.00 price point.
Lol to the bolded....
Where is your such requirements for the post i answered to, did it include :
There was nothing of the sort, but i guess that some wild speculations are bought at face value if ever they fuel some beliefs..
So my estimation is no worse than the opposite and apparently accepted ones, i would say i m even much more on the money since i can point that the PCB is smaller, that s already a valuable argument...
Ok. Two things need to happen now.
1. Provide a BOM for both AMD FuryX and Nvidia 980Ti.
2. Provide credible reasoning why said BOMs for each allows, or prohibits said GPU from being competitive at a 649.00 price point.
Otherwise you guys will go round robin for eternity.
Both GM200 and Fiji is around the same size. So assume same cost. However 980ti is the cutdown version. So cheaper 980ti.
Fury X needs interpsoer=extra cost and made at UMC?
HBM1 cost more than GDDR5. I dont think anyone wants to argue about that.
Fury X uses expensive water cooling solution.
980Ti got bigger PCB, but PCB is pennies.
So unless someone got something magic to show for the Fury X. The BOM is higher.
Thank you Skintai for illustrating something incredibly obvious that should have been visible at first glance.
Both GM200 and Fiji is around the same size. So assume same cost. However 980ti is the cutdown version. So cheaper 980ti.
Fury X needs interpsoer=extra cost and made at UMC?
Fury X uses expensive water cooling solution.
980Ti got bigger PCB, but PCB is pennies.
So unless someone got something magic to show for the Fury X. The BOM is higher.
Lol to the bolded....
Where is your such requirements for the post i answered to, did it include :
There was nothing of the sort, but i guess that some wild speculations are bought at face value if ever they fuel some beliefs..
So my estimation is no worse than the opposite and apparently accepted ones, i would say i m even much more on the money since i can point that the PCB is smaller, that s already a valuable argument...
Bottom line, nobody knows what costs what. Deductive reasoning goes out the window. especially if you don't even know if AMD and Nvidia pays the same for a wafer of GPUs. The most likely don't. Quantity. Who knows what other outliers.
Funny thing is you sound like you think I was disagreeing with your last posts. I was not.
I wasn't agreeing with them either. Nor Enigmoid's.
Please provide edvidence that water cooler is cheaper than air coolers. Because thats certainly not what we can see in the retail channel, both for complete products and parts.
Also show me that an interposer is "3-4$ at most". An interposer is wafer based and its quite big.
And yes, when you can use chips with and without a defect, yield is higher and its cheaper to make.
In case you didnt know. AMD got nothing to do with memory production. They are simply a customer like any other. But they can only buy from SK Hynix for a low volume new product. So I would like to see how you make that one cheaper as well, against GDDR5 thats highly matured, high volume and with multiple manufactors.
And then I spare you about what happens if a 2.5D product got a defect.
Re lol, let s see what is that first glance based on :
So i take a 600mm2 chip, i disable part of it and it magically become cheaper to produce..?.
http://www.anandtech.com/show/9266/amd-hbm-deep-dive/33-4$ at most...
http://electroiq.com/blog/2012/12/lifting-the-veil-on-silicon-interposer-pricing/Of course the interposer does come with a drawback as well, and thats cost. While AMD is not talking about costs in great detail this is a technology deep dive, not an analyst meeting the fact that the interposer is essentially a very large, partially developed silicon chip means that its relatively expensive to produce, especially compared to the very low costs of PCBs and traditional substrates. Mitigating this is the fact that interposers dont need to go through the most complex and expensive phases of photolithography the actual front-end lithography so the cost is only the silicon wafer itself, along with the work required to create the metal layers, with the final interposer only being some 100 microns thick. Furthermore this doesnt require cutting-edge fabs old, fully amortized 65nm equipment works quite well which further keeps the costs down. The end result is that the interposer is still a significant cost, but it is not as bad as it initially seems. This ultimately is why HBM will first be introduced on high margin products like high-end video cards before potentially making its way down to cheaper devices like APUs.
Bottom line, nobody knows what costs what. Deductive reasoning goes out the window. especially if you don't even know if AMD and Nvidia pays the same for a wafer of GPUs. The most likely don't. Quantity. Who knows what other outliers.
Was going to post this. Seems neither knows.
In reality no. Effectively, strange as it sounds, yes.
L For instance, using purely hypothetical numbers to illustrate this point say chip A gets 90% perfect yields and of that 10% of that failed 5% can be 'feature binned' (usable with deactivated cache, core, compute unit, etc.).
Requiring a perfect chip gives only 90% yields. Requiring a functional with units disabled chip gives 95% yields. Thus selling the disabled chip is cheaper per wafer.
You can make an educated guess. Remember no one is predicting exact costs and you do NOT need exact costs to conclude: " Costs are significantly higher for one." Many things point to the BOM of the Fury X being more expensive than the 980 Ti.
That s flawed logic, it doesnt cost less, it cost the same and you dont even sold it at a same price, so its cost is actualy more...
It costs less because you can sell the defective ones. Therefore per wafer you may get 85 dies instead of 82 because the defective ones can be sold as well. Again it gets messy when SKUs and such come into play.
Saying "many things" is like saying "i think", that s a blank statement with no factual evidence, it can be an opinion but is not even an educated guess...
It is an educated guess. If the reasoning is there and it is sound then the point has some merit. I purposely hedge my statements because I will not attempt to pass off my predictions as absolute truth.
The 'many things' were clearly detailed earlier in my post and earlier in this thread.
It's hard to believe that this will be the case.
AMD must have known where the card stood against the 980ti.
AMD must have known that overclocking was the key.
It's difficult to accept the idea that the card overclocks very well, but this was restricted at release due to driver instability.
AMD could simply have said, "We also have this beta driver, so you can see what sort of overclocks Fury can achieve. It's beta, so don't use it for anything but testing. We will release a more mature version soon."
It s less expensive than a conventional cooler, there s less machining...
.
I highly doupt Intel will ever be making discrete video cards, it's all focused on processor graphics now.AMD can't complete with Nvidia no more. They haven't in a while. I heard rumor that Intel was going to join the GPU market, but I guess either that was a rumor or they decided against it.
This is not a good time to be a PC gamer not when the market could soon be a monopoly.
I highly doupt Intel will ever be making discrete video cards, it's all focused on processor graphics now.
The i740 is one of many reasons why Intel won't be making any more dGPUs.
AMD can't complete with Nvidia no more.
It costs less because you can sell the defective ones. Therefore per wafer you may get 85 dies instead of 82 because the defective ones can be sold as well. Again it gets messy when SKUs and such come into play.