AMD is currently selling AMD RX Vega 64 graphics cards at a loss, specifically, it would lose $ 100 for each graphics card, according to the specialized media Fudzilla.
I think I remember that the PlayStation 3, when it hit the market, was priced $ 100 below cost, as the company was clear that it would more than recoup its losses by selling games. This strategy is quite dangerous, since you depend on third parties. Within the console market it has its logic, but in the graphics card market, how is the investment recovered? Well, neither we nor Fudzilla know, since this medium has learned from industry sources that AMD is selling the [amazon_textlink asin = 'B074SR3NBX' text = 'AMD RX Vega 64 ′ template =' ProductLink 'store =' hardwa028- 21 ′ marketplace = 'ES' link_id = '8dad5146-8d77-11e7-8fda-9f17a5fcbac7 ′] to losses, specifically, € 100 of losses per graph.
Fudzilla publishes that AMD loses $ 100 per Vega 64 graphics card in terms of the retail price, this being $ 499 (modifications and price increases apart). The reasons seem to be the HBM2 memories, the cost of the encapsulation and the substrate, which would be extremely high for a sustainable price of $ 499. This is not the first time that it has been discussed, but the AMD Vega are not made to make money, they are developed to gain market share, which would improve the company's business options. Even so, we must emphasize that it is difficult to gain market share, if there is no stock.
The main problem is not so much the cost of manufacturing the GPU, higher according to Fudzilla than NVIDIA. The main problem is the HBM2 memories, which are complex to manufacture and have a very high price. Hynix should offer these HBM2 memories in October, something that could alleviate AMD's problem a bit. Everything indicates that AMD has made a mistake using these memories in its graphics cards. Equally betting on the GDDR5X, inferior in performance, it would have been a better mattress and wait for the HBM2 to be cheaper to launch a 'Premium' model.
Source: Fudzilla
Shameful, let's see if they kick their CEO Raja the Indian out at once
Were the Radeon Vega developed to gain market share? Neither you believe it, nor AMD, to gain market share from Nvidia there are only two options ...
-The first is to offer a better product than the competitor, for example the Radeon 5870 arrived months before the Geforce 480 which is also the worst Geforce in history.
-The second offer an unmatched cost performance ratio like the Radeon 4870.
Chance? They are the best Radeon in history, all the other generations lost clear market share in the time of AMD, since the ATI Radeon 9000 series is missing out there.
Why the hell don't they kick Raja kodurI out
If they had put GDDR5X in Vega, the consumption was out of control (even more so), and they would have had to put them with lower frequencies, (more still), because the HBM2 consumes much less than the GDDR5X, like 50 or 60 watts. less, so they had no other choice.
Exactly this I saw in gamer nexus where they explained that they possibly used HBM2 for the same reason you have.
AMD has an obsolete architecture, it bases everything on hardware power, a lot of memory and fast, a lot of bandwidth, a lot of computing units ... but the architecture is not good. Nvidia with much less hardware, achieves better performance.
A 1080 is half in hardware quality and performs better than a Vega 64, and this is called architecture ... nothing more.
Obsolete? What part of the planet of the apes are you from?
Amd has pioneered many technologies that are used today.
PS: sorry apes. There is everything in the Lord's vineyard.
Mining charts.
I also sell at a loss! Hahaha!!
The fudzilla sources are unreliable enough to believe your article! Are they trying to misinform ???
Here, a red basin for you to put your tears and your bile inside. https://uploads.disquscdn.com/images/b510148a2657712043dda37f6d137e0b39bd2aadf79c51af9bbb0f7a3c5b4f16.jpg
I am more and more convinced that this page is as blue as Papa Smurf. You know what I mean!
1. You can see from the cover that most of it is news about AMD. That you don't like being told the truth, it's your problem, not mine at all. That AMD sells the graphics for $ 100 less than what it costs, that it has had to pay almost 30 million dollars for the shit of the Llano APUs and that the AMD RX Vega 64 consume a stupid thing for the power they give, they are facts.
2. I care little about the opinion of four AMD fanboys who would buy anything AMD releases.
3. We have tested the Vega 64 and we have had a thousand problems, something that with the RX 500 Series and the RX 400 Series we have not had.
4. AMD has lied with the stock, since it has delayed its launch several times with the excuse of the stock and in the end there is not even Vegas 64 and less Vega 56.
5. They lied when they said they would be good for mining, because 40MH / s for Ethereum and 400W of consumption, for more than € 600, does not compensate at all for mining, profitability is on the ground.
6. Insist on the part of Raja Koduri that the problem is mining, when no one would touch a Vega with a stick for mining.
7. Throwing in a graphics card that you have delayed several times, some Beta drivers that do not work well and do not display the data in a universal software like the GPU-Z, is pathetic.
4. They did not lie, it took MINUTES to almost sell out in all stores at the recommended price and therefore, as always, they inflated the price. That's why you still see some exorbitant stock in 64 and 56. This does not mean that AMD does not have enough stock, it means that the amount of GPUs that miners eat is exaggerated.
You have videos on YT of people with dozens of vega 64 to mine ... I imagine that since they usually have no problems flashing them, they will have undervolted them and changed the clocks to throw more memo than core, which is what eats up watts and thus raise the hashrate somewhat. But I already mentioned before the start of everything, that when people said that they would have 70-100Mhs they were simply dreaming, and I gave them 40 xD in the best of cases
5. They have not lied, since the miners have aired them, both 64 and 56 (these to a greater extent). The rumors that were on the internet of its power of hashing have lied, which is far from what they say. And a GPU with 2 8-pin PEG connectors, you who know hard, you should know that it has no possibility of reaching 400W of consumption or 100% ... [75W of the motherboard + 150w * 2 of the PEG connectors] based on those famous tables that came out the day the ban was lifted? Because precisely from that premise it could be assumed that the consumption was not from the GPU, but from the entire system xD.
6. That said, mining is a problem and it has been shown both in the output of 64 and in the output of 56. Because of this fact, I know that I will have to be aware of the minute the assembler versions come out to be able to get a 56 to an acceptable price.
7. This is not new and both houses sin from it… And you know it. GPU-z is free software that, like all software that reads hardware data, needs to be able to access and interpret that information. This is a process in which both parties intervene and I am at least tired of seeing that problem over the years until the official release of the hard in question and even months later ... And that some BETA drivers do not work well ... Anyway , I don't think I have to indicate anything about it xDD
Damn ... Sometimes I am afraid of how good I am at predicting the future xD
http://wccftech.com/amd-rx-vega-64-pushed-43-5mhs-130w-mining-ethereum-eclipsing-polaris-efficiency-factor-2x/
You should also say that in a later post they deny it and updated the entry you put saying that it is 43MH / s for 130W, nothing at all. I don't know if you've noticed, but the AMD drivers randomly falsify certain data.
http://wccftech.com/amds-rx-vega-64-etherium-mining-248-watts-of-power-draw/
I already mentioned that news to you ... You have to look more closely! xD
In any case, at no time did I say X consumption, I only spoke of lower consumption and 40Mhs, ergo I was completely right ...
News smoke, the same was said about the fury and many were sold, today listed as the 290X mainly for mining, I sold my Matrix 290X 2 weeks ago for € 180. and AMD in greenery textures is infinitely better than Nvidia. They have to hide in bullshit now as the temperatures of the glas founders are the same, they throw shit on the consumption, it is an HBM2 has everything in 1 and logically it needs energy, the day they let Nvidia manufacture HBM we will see the same consumption
They said the same about the fury and many were sold, today listed as the 290X mainly for mining, I sold my Matrix 290X 2 weeks ago for € 180. and AMD in greenery textures is infinitely better than Nvidia. They have to hide in bullshit now as the temperatures of the glas founders are the same, they throw shit on the consumption, it is an HBM2 has everything in 1 and logically it needs energy, the day they let Nvidia manufacture HBM we will see the same consumption
HBM2s are supposed to consume considerably less than GDDR5Xs, the world is still wrong. The problem with these graphics is that they are priced higher than several GTX 1080s, they consume twice or more than the aforementioned NVIDIA graphics and the stock problem, when they said several times that they were delaying them to have stock, well, that's it. I do not know the problem of stock graphics if it is from AMD or who manufactures the HBM2, which are a headache, since many of these memories have to be thrown away because they are defective. I'm surprised that someone who knows so much about hardware misrepresents in this way.
PS: AMD and NVIDIA do not make HBM memory, they will be made by SK Hynix, Samsung or GlobalFoundries.
@BAURK Global Foundries I think is the one that manufactures Ryzen and SkHynix or Samsung HBM2. It is unforgivable that AMD does not have enough graphics to supply the market. But let's take into account the development that has been going on for at least 3 years since the sale of its factories and the development of Zen and everything in GPUs, counting those delays until they polished new chips. The CPUs are not arriving at a bad time, but the graphics are arriving late and the 64 could be better, surpassing the 1080/Ti because the 56 is more or less on par with the 1070.
Zen is fine, let's not deny it, the bet on modularity for Ryzen has gone well, but the problem with Threadripper and EPYC will be Infinity Fabric. Vega does not go well, because they need an architecture change. I don't think anyone has noticed that the Vega has almost twice the Stream Processors of NVIDIA Pascal. That means much more consumption, but not more performance. Where is the fault?
Infinity Fabric has performance completely intrinsic to the speed and latencies of the memory that includes the board. I think that the fact of having put the faster memory controller, very fast RAM like you have the DDR4 and QuadChannel, perfectly makes up for the lack that could be in that regard.
In any case, if it were to be a problem, it would only be in tasks that involve all the cores or at least so many as to involve separate groups and in which there is information exchange between said cores with varied and non-constant workloads. And as I say, it is solved with the above mentioned xD
PS: CUDA cores and Stream Processors are not comparable. They never have been xD