Showing posts with label sapphire rapids. Show all posts
Showing posts with label sapphire rapids. Show all posts

11.10.2022

Intel Announces Xeon Max - World's First x86 Processors with HBM Memory

Intel Announces Xeon Max - World's First x86 Processors with HBM Memory

Intel Announces Xeon Max - World's First x86 Processors with HBM Memory

Intel has announced Xeon Max, a series of processors equipped with HBM2e high speed memory, and according to the company it is the world's first CPUs with HBM.
Previously these chips appeared under the codename Sapphire Rapids HBM.
The processors offer up to 56 cores with support for 112 virtual threads and have a TDP of 350W, and they are designed for high-performance server systems.Image source: IntelThe Xeon Max processors use an EMIB interface that integrates four chips with compute cores, as well as located next to the same substrate 64 GB of high-speed HBM2e memory, split into four clusters of 16 GB each.
The total bandwidth is about 1 Tbyte/s.
So there is more than 1GB of HBM2e memory per processor core.
Key features of the Xeon Max are support for PCIe 5.0 and CXL 1.1 interfaces.
HBM2e memory can be used both as additional cache and as additional RAM.
Moreover, a server with Xeon Max can be equipped with no RAM modules at all - the system will rely solely on HBM.
The company claims that Xeon Max power consumption is 68% lower than that of AMD Milan-X processors for the same performance.
The Xeon Max's support for the new AMX instructions speeds up AI-related tasks and provides an eight-fold increase in peak performance over AVX-512 instructions in INT8 and INT32 operations.
Intel says the Xeon Max is up to five times faster in some types of operations compared to the Intel Xeon 8380 or AMD EPYC 7773X (Milan-X with 3D V-Cache technology).
The presentation also compares the Xeon Max to the AMD EPYC 7763, against which the new Intel product shows up to 3.6 times better performance.
Compared to the NVIDIA A100 server computing gas pedal in the MLPerf DeepCAM test, which deals with calculations that accelerate and augment simulations on supercomputers using AI, Xeon Max is up to 1.2 times faster than the competitor.
The Xeon Max series of server processors will hit the market in January 2023.
Key competitors to Xeon Max will be AMD's new EPYC Genoa processors.
Their announcement is expected tomorrow, November 10.
According to rumors, these processors will be available as solutions equipped with HBM memory.

11.02.2022

Intel will finally release Sapphire Rapids server processors in January - their release has been delayed since 2021

Intel will finally release Sapphire Rapids server processors in January - their release has been delayed since 2021

Intel will finally release Sapphire Rapids server processors in January - their release has been delayed since 2021

Intel has set a release date for its 4th generation Xeon Scalable server processors, better known by their codename Sapphire Rapids.
The company will release chips based on the new Intel 7 manufacturing process technology, scheduled for January 10 next year.Image source: IntelIntel said it is ready to present solutions that \"meet the requirements for the product being released.\" This wording could mean that not all Sapphire Rapids processor models planned for release will be released at the same time.
This will apply only to those models that have passed all necessary tests and are ready for mass production.
Earlier it was reported that during the development of processors Sapphire Rapids since the announcement the company had to fix about 500 bugs, as well as release 12 different stepping chips - from A0 to E5.As it became known from a recent report of the analyst company TrenForce, the reason for the continued postponed release of Sapphire Rapids is a significant amount of manufacturing defects.
In particular, Intel has faced with a low yield of good chips in the mass production chipsets MCC processors Sapphire Rapids.
These chipsets are produced using the Intel 7 process technology.
The number of quality chips is only 50-60% of the total production volume.
In such circumstances, the transition to mass production of Sapphire Rapids seemed to the company economically impractical, so the manufacturer is trying to eliminate the problem.
Source image: VideoCardzThe Intel Sapphire Rapids processors are designed for Eagle Stream platform, which is announced to support PCIe 5.0 interface and DDR5 RAM.
The platform will also be used by 5th generation Xeon Scalable processors, code-named Emerald Rapids, which are also expected to launch next year.
Sapphire Rapids processors will offer up to 56 processing cores on Golden Cove architecture.

Intel faces high defect rate in production of Sapphire Rapids server chips

Intel faces high defect rate in production of Sapphire Rapids server chips

Intel faces high defect rate in production of Sapphire Rapids server chips

Release of new Intel Sapphire Rapids server processors on the new Intel 7 process technology has been postponed several times already.
According to the last plan, the manufacturer was going to start mass production in the fourth quarter of this year.
However, in a recent report TrendForce analysts said that the company will move the start of mass production of Sapphire Rapids in the first quarter of 2023.Image source: IntelThe reason for the transfer of experts call the significant amount of manufacturing defects.
In particular, Intel has faced with a low yield of good chips in the mass production of chipsets MCC processors Sapphire Rapids.
These chipsets are produced using the Intel 7 process technology.
It is reported that the number of quality MCC chiplets is only 50-60% of the total production volume.
In such circumstances, the transition to mass production of Sapphire Rapids does not look economically viable, so the manufacturer will try to eliminate the problem.
Delayed mass production of server processors Sapphire Rapids will force the OEMs of server equipment and cloud computing providers to delay the launch of new projects based on them, experts say.
AMD could take advantage of the situation.
As pointed out by TrendForce, its share of the server market in 2023 could grow from 15% to 22%.
Given the fact that AMD supplies the server chips, which have a large number of cores, many companies may choose them as a basis for their solutions.
The TrendForce report also points out that due to rising costs of operating the server equipment, amid rising global energy prices, more companies choose to use single-processor server platforms.
At the same time, Intel and its company Altera had problems with the supply of entry-level FPGA, which are used in motherboards with multiple sockets, as a result of the shortage led to a reduction in the supply of dual-processor systems.
As an alternative, manufacturers are looking toward CPLDs from Lattice, but it too is currently in short supply.This is why cloud server holders and server OEMs have increased their interest in AMD's solutions, which have proven to be easier and simultaneously cheaper to use.
According to TrendForce, by the fourth quarter of 2023, AMD will increase the share of its processors in the global market to 25%.

6.08.2022

Intel Sapphire Rapids server processors will not come out as planned due to \"side problems\"

Intel Sapphire Rapids server processors will not come out as planned due to \"side problems\"

Intel Sapphire Rapids server processors will not come out as planned due to \

The transition to mass production of 4th generation Xeon Scalable server processors (Sapphire Rapids) will occur later than expected.
This was announced at the BofA Securities Global Technology Conference by Sandra Rivera, head of the server division.
In fact, it is the second postponement of the release of these chips.
Previously, the company had intended to release the processors in the first quarter of this year.Image source: IntelInternet has previously reported that during the first quarter, it shipped samples of 4th generation Intel Xeon Scalable codename Sapphire Rapids to some customers.
The chips are currently being tested by the company's partners.
According to Rivera, Intel needs more time to develop new platforms for Sapphire Rapids and validate them.
Therefore, mass shipments of the processors, although they will start this year, but later than planned.
\"We need more time to develop and test new platforms, so the mass production release of Sapphire Rapids is expected later than originally expected.
However, we are still seeing strong demand for these chips.
I would also like to add that our Intel 7 processor, on which Sapphire Rapids are built, feels great.
I would like to point out that the same process is used in our Alder Lake consumer chips.
We've ramped up production to 15 million units.
We already mentioned in our Q1 report that it's the fastest ramp-up in consumer processor production in the last ten years.
The process technology is doing very well, and we're projecting a good chip volume picture, but there are some side issues that need to be addressed with our customers,\" commented Sandra Rivera.
It is very important for Intel to make sure that all components of the Sapphire Rapids platform pass the necessary testing, as the successors to the Emerald Rapids codename processors will also be compatible with the platform.
These chips are expected in the second half of 2023 and will also use the Intel 7 (10nm) process technology.

6.01.2022

Intel Sapphire Rapids-SP server processors with HBM2e memory will be up to three times faster than their Ice Lake-SP predecessors

Intel Sapphire Rapids-SP server processors with HBM2e memory will be up to three times faster than their Ice Lake-SP predecessors

Intel Sapphire Rapids-SP server processors with HBM2e memory will be up to three times faster than their Ice Lake-SP predecessors

Intel at the International Supercomputing Conference has shared fresh details about its upcoming 4th generation Xeon Scalable server processors, also known as Sapphire Rapids.
Processors of this series will be available in two versions: standard and with HBM2e memory.
There were fresh details about the latter.Image source: IntelThe standard variant of Sapphire Rapids processors has a chip design of four XCC crystals.
The area of each chiplet is about 400 mm2.
The crystals will be interconnected by an EMIB bus.
These processors will have 10 EMIB interconnects and the total standard Sapphire Rapids-SP processor package area will be 4,446mm2.
Sapphire Rapids-SP HBM variants will be able to offer up to 64GB of HBM2e high bandwidth memory in the form of four 16GB stacks each.
Given the increased number of chips on the substrate of these processor models including HBM2e memory, these processor versions will have 14 EMIB interconnect points.
The total packaging area of Sapphire Rapids-SP HBM will be 28% greater than the standard variant and will be 5,700 mm2.
Intel says the EMIB channel will offer twice the bandwidth density and four times the power efficiency compared to standard packaging designs.The company also shared performance data for Sapphire Rapids-SP with HBM2e memory.
According to the manufacturer, the next-generation Xeon processors deliver twice the metrology performance of Xeon Ice Lake-SP chips, and they are more than three times faster in the YASK benchmark, have twice the performance in openFOAM (a platform for numerical simulation of continuum mechanics problems), and are more than three times the performance in CloverLeaf.
In addition, the company claims that Sapphire Rapids-SP with HBM2e memory will offer twice the performance in fluid dynamics calculations (Ansys Fluent) and twice the performance in ParSeNet.
Recall that Sapphire Rapids-SP processors will get up to 56 cores with support for up to 112 virtual threads.
They will work with 8-channel DDR5-4800 memory, support up to 80 PCIe 5.0 lanes, and their maximum L3 cache capacity is 105 Mbytes.

7.24.2021

Intel admitted that it is forced to reduce server processor prices due to competition with AMD

Intel admitted that it is forced to reduce server processor prices due to competition with AMD

Intel admitted that it is forced to reduce server processor prices due to competition with AMD

The discussion of the situation in the server segment at Intel's quarterly reporting event boiled down to management's attempts to warn investors against jumping to conclusions, as revenue and operating profit in this line of business fell in the second quarter. Regarding the delayed announcement of Sapphire Rapids, it was said that it was not caused by technology issues.

In general, during the whole event Intel was trying in every way to demonstrate that it's all right with 10nm technology. Ice Lake-SP server processors are quickly gaining popularity with their target customers. In the second half of the year Intel expects to increase server revenue by double-digit percent sequentially. Company executives had to explain that the 7 % decrease in the average selling price of server processors is mainly due to increased competition. In other words, AMD EPYC processors continued to gain popularity, and Intel had to counter this trend with a price reduction for its own products. In the second half of the year, Intel expects to keep not only the prices, but also the market share. With the lack of production capacity on the whole market, Intel has an advantage in the form of the opportunity to produce processors in its own facilities. The company promises to make every effort to regain lost market positions. Intel believes that the server market has passed its bottom in the first quarter, and by the end of the year it will show continuous growth. On the reasons for delaying the announcement of Sapphire Rapids until the next six months, the head of Intel said rather vaguely, but from his words we can conclude that the company's customers needed more time to validate all the changes that the new family brings. DDR5 support and PCI Express 5.0 were mentioned as those new features that take extra time to adapt. The artificial intelligence market, according to Gelsinger, will continue to grow by an average of 20 % annually, and therefore Intel could regain the leading position in all its segments, not excluding the server segment, by 2024. The performance of the Sapphire Rapids samples they have at their disposal, the company's customers are very satisfied, as Intel's head noted.

12.22.2020

Future Intel and AMD server processors have been sequenced

Future Intel and AMD server processors have been sequenced

Future Intel and AMD server processors have been sequenced

The first half of next year should see the arrival of Intel's highly anticipated 10nm Ice Lake-SP server processors and their potential competitors & ; AMD EPYC processors from the Milan family, which combine 7nm production technology with Zen 3 architecture. The first AMD 5nm processors in this segment will not appear before the second quarter of 2022.

Image source: AMD

The sequence of release of new Intel and AMD server processors was reported yesterday by DigiTimes website, citing information from Taiwanese server-class cooling system manufacturers. It is underlined that mass deliveries of Intel Ice Lake-SP processors will not start before the second quarter. Intel has recently repeatedly reported that it expects to roll out shipments in the middle of the first quarter, but from a logistical standpoint this discrepancy is understandable. As you may recall, Intel expects to introduce 10nm Sapphire Rapids processors late next year, but Taiwanese sources are convinced it won't do so until the first quarter of 2022. A year later, Intel's first 7nm server processors will be unveiled under the Granite Rapids designation. Finally, in 2024 there will be Diamond Rapids processors, which will be manufactured using an improved version of the 7nm technology. On the further plans of AMD is said to be rather modest, mentioning company's intentions to offer in the second quarter of 2022 EPYC processors of Genoa family with Zen 4 architecture, which the company is currently developing. TSMC will produce these 5nm processors, which gives an indication of the approximate timing of the other Genoa products. It is believed that 5nm technology will be used not only by NVIDIA, but also by Intel to manufacture performance components.