From WikiChip
Difference between revisions of "nervana/nnp/nnp-t 1300"
(4 intermediate revisions by the same user not shown) | |||
Line 1: | Line 1: | ||
{{nervana title|NNP-T 1300}} | {{nervana title|NNP-T 1300}} | ||
− | {{chip}} | + | {{chip |
− | '''NNP-T 1300''' is a [[neural processor]] designed by [[Intel Nervana]] and introduced in late 2019. Fabricated on TSMC [[16 nm process]] based on the {{nervana|Spring Crest|l=arch}} microarchitecture, the NNP-T 1300 has 22 {{nervana|Spring Crest#Tensor Processing Cluster (TPC)|TPCs|l=arch}} along with 55 MiB of scratchpad memory and operates at up to 950 MHz. This chip comes in a PCIe 4.0 [[accelerator card]] form factor and incorporates 32 GiB of [[HBM2]] memory. This NPU exposes 16 {{nervana|Spring Crest#InterChip Link (ICL)|inter-chip links|l=arch}} for scale-out capabilities. | + | |name=NNP-T 1300 |
+ | |image=spring crest package (front).png | ||
+ | |back image=spring crest package (back).png | ||
+ | |caption=NPU with 4 HBM2 stacks | ||
+ | |designer=Intel | ||
+ | |manufacturer=TSMC | ||
+ | |model number=NNP-T 1300 | ||
+ | |market=Server | ||
+ | |first announced=November 12, 2019 | ||
+ | |first launched=November 12, 2019 | ||
+ | |family=NNP | ||
+ | |series=NNP-T | ||
+ | |frequency=950 MHz | ||
+ | |microarch=Spring Crest | ||
+ | |process=16 nm | ||
+ | |transistors=27,000,000,000 | ||
+ | |technology=CMOS | ||
+ | |die area=680 mm² | ||
+ | |core count=22 | ||
+ | |max memory=32 GiB | ||
+ | |smp interconnect=InterChip Link | ||
+ | |smp interconnect links=16 | ||
+ | |smp interconnect rate=28 GT/s | ||
+ | |power=150 W | ||
+ | |tdp=300 W | ||
+ | |package name 1=intel,fcbga_3325 | ||
+ | }} | ||
+ | '''NNP-T 1300''' is a [[training]] [[neural processor]] designed by [[Intel Nervana]] and introduced in late 2019. Fabricated on TSMC [[16 nm process]] based on the {{nervana|Spring Crest|l=arch}} microarchitecture, the NNP-T 1300 has 22 {{nervana|Spring Crest#Tensor Processing Cluster (TPC)|TPCs|l=arch}} along with 55 MiB of scratchpad memory and operates at up to 950 MHz. This chip comes in a PCIe 4.0 [[accelerator card]] form factor and incorporates 32 GiB of [[HBM2]] memory. This NPU exposes 16 {{nervana|Spring Crest#InterChip Link (ICL)|inter-chip links|l=arch}} for scale-out capabilities. | ||
+ | |||
+ | == Peak Performance == | ||
+ | The NNP-T 1300 has a peak performance of [[peak flops (half-precision)::93.39 TFLOPS]] ([[bfloat16]]). | ||
+ | |||
+ | == Cache == | ||
+ | {{main|nervana/microarchitectures/spring_crest#Memory_Hierarchy|l1=Spring Crest § Cache}} | ||
+ | * 55 MiB of tightly-coupled scratchpad memory | ||
+ | ** 22 x 2.5 MiB/core | ||
+ | |||
+ | == Memory controller == | ||
+ | {{memory controller | ||
+ | |type=HBM2-2400 | ||
+ | |ecc=Yes | ||
+ | |max mem=32 GiB | ||
+ | |controllers=4 | ||
+ | |channels=32 | ||
+ | |max bandwidth=1.2288 TB/s | ||
+ | }} | ||
+ | |||
+ | == Interconnect Topology == | ||
+ | The NNP-T 1300 comes in a dual-slot standard PCIe 4.0 card which enables support for only the ring topology. Chips are interconnected using the 16 available {{nervana|Spring Crest#InterChip Link (ICL)|inter-chip links|l=arch}}. | ||
+ | |||
+ | == Die == | ||
+ | {{main|nervana/microarchitectures/spring_crest#Die|l1=Spring Crest § Die}} | ||
+ | * 27,000,000,000 transistors | ||
+ | * 680 mm² die size | ||
+ | |||
+ | |||
+ | :[[File:spring crest floorplan.png|400px|link=nervana/microarchitectures/spring_crest#Die]] | ||
+ | |||
+ | == Product Brief == | ||
+ | * [[:File:16433-1 NNP-announce NNP-T brief v4.3.pdf|Intel NNP-T Product Brief]] |
Latest revision as of 10:51, 1 February 2020
Edit Values | |
NNP-T 1300 | |
![]() | |
NPU with 4 HBM2 stacks | |
General Info | |
Designer | Intel |
Manufacturer | TSMC |
Model Number | NNP-T 1300 |
Market | Server |
Introduction | November 12, 2019 (announced) November 12, 2019 (launched) |
Shop | Amazon |
General Specs | |
Family | NNP |
Series | NNP-T |
Frequency | 950 MHz |
Microarchitecture | |
Microarchitecture | Spring Crest |
Process | 16 nm |
Transistors | 27,000,000,000 |
Technology | CMOS |
Die | 680 mm² |
Cores | 22 |
Max Memory | 32 GiB |
Multiprocessing | |
Interconnect | InterChip Link |
Interconnect Links | 16 |
Interconnect Rate | 28 GT/s |
Electrical | |
Power dissipation | 150 W |
TDP | 300 W |
Packaging | |
Package | FCBGA-3325 (FCBGA) |
Dimension | 60 mm × 60 mm |
Contacts | 3325 |
![]() |
NNP-T 1300 is a training neural processor designed by Intel Nervana and introduced in late 2019. Fabricated on TSMC 16 nm process based on the Spring Crest microarchitecture, the NNP-T 1300 has 22 TPCs along with 55 MiB of scratchpad memory and operates at up to 950 MHz. This chip comes in a PCIe 4.0 accelerator card form factor and incorporates 32 GiB of HBM2 memory. This NPU exposes 16 inter-chip links for scale-out capabilities.
Contents
Peak Performance[edit]
The NNP-T 1300 has a peak performance of 93.39 TFLOPS93,390,000,000,000 FLOPS
93,390,000,000 KFLOPS
93,390,000 MFLOPS
93,390 GFLOPS
0.0934 PFLOPS
(bfloat16).
93,390,000,000 KFLOPS
93,390,000 MFLOPS
93,390 GFLOPS
0.0934 PFLOPS
Cache[edit]
- Main article: Spring Crest § Cache
- 55 MiB of tightly-coupled scratchpad memory
- 22 x 2.5 MiB/core
Memory controller[edit]
![]() |
Integrated Memory Controller
|
|||||||||||
|
Interconnect Topology[edit]
The NNP-T 1300 comes in a dual-slot standard PCIe 4.0 card which enables support for only the ring topology. Chips are interconnected using the 16 available inter-chip links.
Die[edit]
- Main article: Spring Crest § Die
- 27,000,000,000 transistors
- 680 mm² die size
Product Brief[edit]
Facts about "NNP-T 1300 - Intel Nervana"
back image | ![]() |
base frequency | 950 MHz (0.95 GHz, 950,000 kHz) + |
core count | 22 + |
designer | Intel + |
die area | 680 mm² (1.054 in², 6.8 cm², 680,000,000 µm²) + |
family | NNP + |
first announced | November 12, 2019 + |
first launched | November 12, 2019 + |
full page name | nervana/nnp/nnp-t 1300 + |
has ecc memory support | true + |
instance of | microprocessor + |
ldate | November 12, 2019 + |
main image | ![]() |
main image caption | NPU with 4 HBM2 stacks + |
manufacturer | TSMC + |
market segment | Server + |
max memory | 32,768 MiB (33,554,432 KiB, 34,359,738,368 B, 32 GiB, 0.0313 TiB) + |
max memory bandwidth | 1,144.409 GiB/s (1,171,875 MiB/s, 1,228.8 GB/s, 1,228,800 MB/s, 1.118 TiB/s, 1.229 TB/s) + |
max memory channels | 32 + |
microarchitecture | Spring Crest + |
model number | NNP-T 1300 + |
name | NNP-T 1300 + |
package | FCBGA-3325 + |
peak flops (half-precision) | 93,390,000,000,000 FLOPS (93,390,000,000 KFLOPS, 93,390,000 MFLOPS, 93,390 GFLOPS, 93.39 TFLOPS, 0.0934 PFLOPS, 9.339e-5 EFLOPS, 9.339e-8 ZFLOPS) + |
power dissipation | 150 W (150,000 mW, 0.201 hp, 0.15 kW) + |
process | 16 nm (0.016 μm, 1.6e-5 mm) + |
series | NNP-T + |
smp interconnect | InterChip Link + |
smp interconnect links | 16 + |
smp interconnect rate | 28 GT/s + |
supported memory type | HBM2-2400 + |
tdp | 300 W (300,000 mW, 0.402 hp, 0.3 kW) + |
technology | CMOS + |
transistor count | 27,000,000,000 + |