ThinkSystem SD650 schematic showing connection of Apache Pass / Optane DC Persistent Memory modules (Source: Tom’s Hardware) Source: Tom’s Material
Apache Pass / Optane DC persistent memory module (Source: Tom’s Hardware) Source: Tom’s Material
Apache Pass / Optane DC Persistent Memory module prototype (Source: Legit Reviews) Source: legitimate reviews)
SD650 server for which Lenovo has announced compatibility with Optane DC Persistent Memory based on 3D XPoint (Source: Tom’s Hardware) Source: Tom’s Material
The place of Optane DC Persistent Memory in the hierarchy of memory and storage in computers (respectively servers) according to Intel Source: Intel
3D XPoint chips on the wafer Source: Intel
Intel Optane DC persistent memory module Source: Intel
Intel Optane DC persistent memory module Source: Intel
Performance of Optane DC Persistent Memory and DRAM according to Intel Optimization Handbook (Source: InstLatX46, Twitter) Source: InstLatX46
More and more clouds are gathering over the once very promising 3D XPoint (Intel Optane) memories. Although significantly better than NAND, they are probably too expensive for practical use…
When Intel and Micron introduced 3D XPoint memory, it was a big event: the first next-gen replacement for NAND Flash memory, which isn’t exactly an elegant technology, simply because of the lack of transcriptions. But Intel uses a lot less memory than 3D XPoint than we’d like, and Micron hasn’t even abandoned them, which doesn’t paint their future very well. We seem to know why. They are very expensive and Intel has heavy losses on them.
Next-gen replacement NAND is likely too expensive to manufacture
Intel never gave specific numbers on the cost of 3D XPoint memory versus DRAM and NAND, but the fact that Micron for some reason never launched Optane (it had the QuantX brand in his portfolio) drove the market higher. It is now probably confirmed that at present these memories probably have production costs that are too high to be viable. This despite the fact that Intel used them in more expensive servers and products and that the Optane SSD for PC had relatively high prices.
Hint: Optane SSD ends up in a PC, Intel no longer releases 3D XPoint memory drives to consumers
Accounting documents, the so-called 10-Q SEC report, show that Optane/3D XPoint technology is loss-making for Intel. Intel had previously included Optan production in the Non-Volatile Memory (NSG) Solutions Group, so profitability or loss was masked by the fact that sales and profits were mixed with those of NAND production. , which was larger in volume and therefore overall the result dominated .
However, Intel is now selling the NAND division to Hynix, while retaining production of the Optane/3D XPoint. As a result, sales and resulting profits or losses now begin to be included in the results of the Data Center Group’s server division. This means they will again be drowned and hidden, but due to this decision, Optane’s results for the first 9 months of last year have been recorded separately in this 10-Q SEC report. For the first time, we can see how Optane is doing economically, or last year.
From the first to the third quarter of 2020, revenue generated by Optane memory / 3D XPoint / products would have been $298 million. However, the associated operating “profit” was minus 473 million, so the loss created was half the sales themselves. In the third quarter alone, sales were $86 million and a loss of $116 million. We don’t know if this means that Intel is selling every 3D XPoint memory-based product well below production cost, or if it’s just because of high fixed development costs (and the product unit itself). even then sells at a profit in relation to its production costs, only the volume of sales is not high enough to amortize the fixed costs). Theoretically, it could probably be some accounting factors like showing past development costs.
This year, sales appear to be slightly lower, at $271 million from the first to the third quarter, including $188 million in the third quarter. What are the operating gains or losses, but this time we don’t know.
Will Intel eventually cancel Intel?
If the 3D XPoint memories are thus lost for a long time, it may mean that Intel will sooner or later leave them and cancel this project. It is of course possible that several years of losses will eventually return if the technology manages to break through on a large scale until production costs are reduced to a profitable level. But the question is whether this Intel can happen.
NAND memories could get cheaper at a faster rate and most likely DRAM, which is the biggest danger for Optane. If it increased its capacity and reduced its prices more quickly, 3D XPoint would lose the main niche it had in the market – the category of so-called storage class memory (Intel Optane DC Persistent Memory Modules), an alternative to the Slightly cheaper but slower RAM. and requires special treatment.
More: Intel Officially Introduces Optane DC Persistent Memory Modules, A 3D XPoint Based RAM Replacement
However, if such storage class memory could be replaced in the future by ordinary DDR5 or DDR6 memory at the same price and with the same consumption, it would probably lose its meaning, because only irreversibility will generally not be a sizeable currency. Unfortunately, it is entirely possible for Optane that DRAM as well as NAND will develop faster due to their greater presence in the market. Therefore, more money is spent on their development, and it is produced by several companies, among which there is competition, which positively stimulates the speed of development.
Intel does not yet indicate that it intends to leave the 3D XPoint/Optane memory project, so it is not yet known if the concerns are really in place. However, the unsuccessful end of this technology (at least in the mass market) is now evident in the realm of possible scenarios.
Intel Optane Memory H20 Source: Intel, via ServeTheHome
Gallery: Intel Optane Memory H20
Source: Blocks and Files, Intel