Tenstorrent Launches Wormhole AI Processors: 466 FP8 TFLOPS at 300W
Tenstorrent has unveiled its next-generation Wormhole processor for AI workloads that promises to offer decent performance at a low price. The company currently offers two add-on PCIe cards carrying one or two Wormhole processors as well as TT-LoudBox, and TT-QuietBox workstations aimed at software developers. The whole of today's release is aimed at developers rather than those who will deploy the Wormhole boards for their commercial workloads.
“It is always rewarding to get more of our products into developer hands. Releasing development systems with our Wormhole™ card helps developers scale up and work on multi-chip AI software.” said Jim Keller, CEO of Tenstorrent. “In addition to this launch, we are excited that the tape-out and power-on for our second generation, Blackhole, is going very well.”
Each Wormhole processor packs 72 Tensix cores (featuring five RISC-V cores supporting various data formats) with 108 MB of SRAM to deliver 262 FP8 TFLOPS at 1 GHz at 160W thermal design power. A single-chip Wormhole n150 card carries 12 GB of GDDR6 memory featuring a 288 GB/s bandwidth.
Wormhole processors offer flexible scalability to meet the varying needs of workloads. In a standard workstation setup with four Wormhole n300 cards, the processors can merge to function as a single unit, appearing as a unified, extensive network of Tensix cores to the software. This configuration allows the accelerators to either work on the same workload, be divided among four developers or run up to eight distinct AI models simultaneously. A crucial feature of this scalability is that it operates natively without the need for virtualization. In data center environments, Wormhole processors will scale both inside one machine using PCIe or outside of a single machine using Ethernet.
From performance standpoint, Tenstorrent's single-chip Wormhole n150 card (72 Tensix cores at 1 GHz, 108 MB SRAM, 12 GB GDDR6 at 288 GB/s) is capable of 262 FP8 TFLOPS at 160W, whereas the dual-chip Wormhole n300 board (128 Tensix cores at 1 GHz, 192 MB SRAM, aggregated 24 GB GDDR6 at 576 GB/s) can offer up to 466 FP8 TFLOPS at 300W (according to Tom's Hardware).
To put that 466 FP8 TFLOPS at 300W number into context, let's compare it to what AI market leader Nvidia has to offer at this thermal design power. Nvidia's A100 does not support FP8, but it does support INT8 and its peak performance is 624 TOPS (1,248 TOPS with sparsity). By contrast, Nvidia's H100 supports FP8 and its peak performance is massive 1,670 TFLOPS (3,341 TFLOPS with sparsity) at 300W, which is a big difference from Tenstorrent's Wormhole n300.
There is a big catch though. Tenstorrent's Wormhole n150 is offered for $999, whereas n300 is available for $1,399. By contrast, one Nvidia H100 card can retail for $30,000, depending on quantities. Of course, we do not know whether four or eight Wormhole processors can indeed deliver the performance of a single H300, though they will do so at 600W or 1200W TDP, respectively.
In addition to cards, Tenstorrent offers developers pre-built workstations with four n300 cards inside the less expensive Xeon-based TT-LoudBox with active cooling and a premium EPYC-powered TT-QuietBox with liquid cooling.
Sources: Tenstorrent, Tom's Hardware
AI
Posted by The Techinical Support
Kioxia Details BiCS 8 NAND at FMS 2024: 218 Layers With Superior Scaling 
Kioxia's booth at FMS 2024 was a busy one with multiple technology demonstrations keeping visitors occupied. A walk-through of the BiCS 8 manufacturing process was the first to grab my attention. Kioxia and Western Digital announced the sampling of BiCS 8 in March 2023. We had touched briefly upon its CMOS Bonded Array (CBA) scheme in our coverage of Kioxial's 2Tb QLC NAND device and coverage of Western Digital's 128 TB QLC enterprise SSD proof-of-concept demonstration. At Kioxia's booth, we got more insights.
Traditionally, fabrication of flash chips involved placement of the associate logic circuitry (CMOS process) around the periphery of the flash array. The process then moved on to putting the CMOS under the cell array, but the wafer development process was serialized with the CMOS logic getting fabricated first followed by the cell array on top. However, this has some challenges because the cell array requires a high-temperature processing step to ensure higher reliability that can be detrimental to the health of the CMOS logic. Thanks to recent advancements in wafer bonding techniques, the new CBA process allows the CMOS wafer and cell array wafer to be processed independently in parallel and then pieced together, as shown in the models above.
The BiCS 8 3D NAND incorporates 218 layers, compared to 112 layers in BiCS 5 and 162 layers in BiCS 6. The company decided to skip over BiCS 7 (or, rather, it was probably a short-lived generation meant as an internal test vehicle). The generation retains the four-plane charge trap structure of BiCS 6. In its TLC avatar, it is available as a 1 Tbit device. The QLC version is available in two capacities - 1 Tbit and 2 Tbit.
Kioxia also noted that while the number of layers (218) doesn't compare favorably with the latest layer counts from the competition, its lateral scaling / cell shrinkage has enabled it to be competitive in terms of bit density as well as operating speeds (3200 MT/s). For reference, the latest shipping NAND from Micron - the G9 - has 276 layers with a bit density in TLC mode of 21 Gbit/mm2, and operates at up to 3600 MT/s. However, its 232L NAND operates only up to 2400 MT/s and has a bit density of 14.6 Gbit/mm2.
It must be noted that the CBA hybrid bonding process has advantages over the current processes used by other vendors - including Micron's CMOS under array (CuA) and SK hynix's 4D PUC (periphery-under-chip) developed in the late 2010s. It is expected that other NAND vendors will also move eventually to some variant of the hybrid bonding scheme used by Kioxia.
Storage
The Endorfy Fortis 5 Dual Fan CPU Cooler Review: Towering Value Standard CPU coolers, while adequate for managing basic thermal loads, often fall short in terms of noise reduction and superior cooling efficiency. This limitation drives advanced users and system builders to seek aftermarket solutions tailored to their specific needs. The high-end aftermarket cooler market is highly competitive, with manufacturers striving to offer products with exceptional performance.
Endorfy, previously known as SilentiumPC, is a Polish manufacturer that has undergone a significant transformation to expand its presence in global markets. The brand is known for delivering high-performance cooling solutions with a strong focus on balancing efficiency and affordability. By rebranding as Endorfy, the company aims to enter premium market segments while continuing to offer reliable, high-quality cooling products.
SilentiumPC became very popular in the value/mainstream segments of the PC market with their products, the spearhead of which probably was the Fera 5 cooler that we reviewed a little over two years ago and had a remarkable value for money. Today’s review places Endorfy’s largest CPU cooler, the Fortis 5 Dual Fan, on our laboratory test bench. The Fortis 5 is the largest CPU air cooler the company currently offers and is significantly more expensive than the Fera 5, yet it still is a single-tower cooler that strives to strike a balance between value, compatibility, and performance.
Cases/Cooling/PSUs
Kioxia Details BiCS 8 NAND at FMS 2024: 218 Layers With Superior Scaling
Kioxia's booth at FMS 2024 was a busy one with multiple technology demonstrations keeping visitors occupied. A walk-through of the BiCS 8 manufacturing process was the first to grab my attention. Kioxia and Western Digital announced the sampling of BiCS 8 in March 2023. We had touched briefly upon its CMOS Bonded Array (CBA) scheme in our coverage of Kioxial's 2Tb QLC NAND device and coverage of Western Digital's 128 TB QLC enterprise SSD proof-of-concept demonstration. At Kioxia's booth, we got more insights.
Traditionally, fabrication of flash chips involved placement of the associate logic circuitry (CMOS process) around the periphery of the flash array. The process then moved on to putting the CMOS under the cell array, but the wafer development process was serialized with the CMOS logic getting fabricated first followed by the cell array on top. However, this has some challenges because the cell array requires a high-temperature processing step to ensure higher reliability that can be detrimental to the health of the CMOS logic. Thanks to recent advancements in wafer bonding techniques, the new CBA process allows the CMOS wafer and cell array wafer to be processed independently in parallel and then pieced together, as shown in the models above.
The BiCS 8 3D NAND incorporates 218 layers, compared to 112 layers in BiCS 5 and 162 layers in BiCS 6. The company decided to skip over BiCS 7 (or, rather, it was probably a short-lived generation meant as an internal test vehicle). The generation retains the four-plane charge trap structure of BiCS 6. In its TLC avatar, it is available as a 1 Tbit device. The QLC version is available in two capacities - 1 Tbit and 2 Tbit.
Kioxia also noted that while the number of layers (218) doesn't compare favorably with the latest layer counts from the competition, its lateral scaling / cell shrinkage has enabled it to be competitive in terms of bit density as well as operating speeds (3200 MT/s). For reference, the latest shipping NAND from Micron - the G9 - has 276 layers with a bit density in TLC mode of 21 Gbit/mm2, and operates at up to 3600 MT/s. However, its 232L NAND operates only up to 2400 MT/s and has a bit density of 14.6 Gbit/mm2.
It must be noted that the CBA hybrid bonding process has advantages over the current processes used by other vendors - including Micron's CMOS under array (CuA) and SK hynix's 4D PUC (periphery-under-chip) developed in the late 2010s. It is expected that other NAND vendors will also move eventually to some variant of the hybrid bonding scheme used by Kioxia.
Storage
The Endorfy Fortis 5 Dual Fan CPU Cooler Review: Towering Value Standard CPU coolers, while adequate for managing basic thermal loads, often fall short in terms of noise reduction and superior cooling efficiency. This limitation drives advanced users and system builders to seek aftermarket solutions tailored to their specific needs. The high-end aftermarket cooler market is highly competitive, with manufacturers striving to offer products with exceptional performance.
Endorfy, previously known as SilentiumPC, is a Polish manufacturer that has undergone a significant transformation to expand its presence in global markets. The brand is known for delivering high-performance cooling solutions with a strong focus on balancing efficiency and affordability. By rebranding as Endorfy, the company aims to enter premium market segments while continuing to offer reliable, high-quality cooling products.
SilentiumPC became very popular in the value/mainstream segments of the PC market with their products, the spearhead of which probably was the Fera 5 cooler that we reviewed a little over two years ago and had a remarkable value for money. Today’s review places Endorfy’s largest CPU cooler, the Fortis 5 Dual Fan, on our laboratory test bench. The Fortis 5 is the largest CPU air cooler the company currently offers and is significantly more expensive than the Fera 5, yet it still is a single-tower cooler that strives to strike a balance between value, compatibility, and performance.
Cases/Cooling/PSUs
Kioxia Details BiCS 8 NAND at FMS 2024: 218 Layers With Superior Scaling
Kioxia's booth at FMS 2024 was a busy one with multiple technology demonstrations keeping visitors occupied. A walk-through of the BiCS 8 manufacturing process was the first to grab my attention. Kioxia and Western Digital announced the sampling of BiCS 8 in March 2023. We had touched briefly upon its CMOS Bonded Array (CBA) scheme in our coverage of Kioxial's 2Tb QLC NAND device and coverage of Western Digital's 128 TB QLC enterprise SSD proof-of-concept demonstration. At Kioxia's booth, we got more insights.
Traditionally, fabrication of flash chips involved placement of the associate logic circuitry (CMOS process) around the periphery of the flash array. The process then moved on to putting the CMOS under the cell array, but the wafer development process was serialized with the CMOS logic getting fabricated first followed by the cell array on top. However, this has some challenges because the cell array requires a high-temperature processing step to ensure higher reliability that can be detrimental to the health of the CMOS logic. Thanks to recent advancements in wafer bonding techniques, the new CBA process allows the CMOS wafer and cell array wafer to be processed independently in parallel and then pieced together, as shown in the models above.
The BiCS 8 3D NAND incorporates 218 layers, compared to 112 layers in BiCS 5 and 162 layers in BiCS 6. The company decided to skip over BiCS 7 (or, rather, it was probably a short-lived generation meant as an internal test vehicle). The generation retains the four-plane charge trap structure of BiCS 6. In its TLC avatar, it is available as a 1 Tbit device. The QLC version is available in two capacities - 1 Tbit and 2 Tbit.
Kioxia also noted that while the number of layers (218) doesn't compare favorably with the latest layer counts from the competition, its lateral scaling / cell shrinkage has enabled it to be competitive in terms of bit density as well as operating speeds (3200 MT/s). For reference, the latest shipping NAND from Micron - the G9 - has 276 layers with a bit density in TLC mode of 21 Gbit/mm2, and operates at up to 3600 MT/s. However, its 232L NAND operates only up to 2400 MT/s and has a bit density of 14.6 Gbit/mm2.
It must be noted that the CBA hybrid bonding process has advantages over the current processes used by other vendors - including Micron's CMOS under array (CuA) and SK hynix's 4D PUC (periphery-under-chip) developed in the late 2010s. It is expected that other NAND vendors will also move eventually to some variant of the hybrid bonding scheme used by Kioxia.
Storage
ASMedia Preps USB4 v2 Controller and PHY 
The USB Implementers Forum (USB-IF) introduced USB4 version 2.0 in fall 2022, and it expects systems and devices with the tech to emerge later this year and into next year. These upcoming products will largely rely on Intel's Barlow Ridge controller, a full-featured Thunderbolt 5 controller that goes above and beond the baseline USB4 v2 spec. And though extremely capable, Intel's Thunderbolt controllers are also quite expensive, and Barlow Ridge isn't expected to be any different. Fortunately, for system and device vendors that just need a basic USB4 v2 solution, ASMedia is also working on its own USB4 v2 controller.
At Computex 2024, ASMedia demonstrated a prototype of its upcoming USB4 v2 physical interface (PHY), which will support USB4 v2's new Gen 4 (160Gbps) data rates and the associated PAM-3 signal encoding. The prototype was implemented using an FPGA, as the company yet has to tape out the completed controller.
Ultimately, the purpose of showing off a FPGA-based PHY at Computex was to allow ASMedia to demonstrate their current PHY design. With the shift to PAM-3 encoding for USB4 v2, ASMedia (and the rest of the USB ecosystem) must develop significantly more complex controllers – and there's no part of that more critical than a solid and reliable PHY design.
As part of their demonstration, ASMedia had a classic eye diagram display. The eye diagram demoed has a clear opening in the center, which is indicative of good signal integrity, as the larger the eye opening, the less distortion and noise in the signal. The horizontal width of the eye opening represents the time window in which the signal can be sampled correctly, so the relatively narrow horizontal spread of the eye opening suggests that there is minimal jitter, meaning the signal transitions are consistent and predictable. Finally, the vertical height of the eye opening indicates the signal amplitude and the rather tall eye opening suggests a higher signal-to-noise ratio (SNR), meaning that the signal is strong compared to any noise present.
ASMedia itself is one of the major suppliers for discrete USB controllers, so the availability of ASMedia's USB4 v2 chip is crucial for adoption of the standard in general. While Intel will spearhead the industry with their Barlow Ridge Thunderbolt 5/USB4 v2 controller, ASMedia's controller is poised to end up in a far larger range of devices. So the importance of the company's USB4 v2 PHY demo is hard to overstate.
Demos aside, ASMedia is hoping to tape the chip out soon. If all goes well, the company expects their first USB4 v2 controllers to hit the market some time in the second half of 2025.
Peripherals
Silicon Motion Demonstrates Flexible Data Placement on MonTitan Gen 5 Enterprise SSD Platform 
At FMS 2024, the technological requirements from the storage and memory subsystem took center stage. Both SSD and controller vendors had various demonstrations touting their suitability for different stages of the AI data pipeline - ingestion, preparation, training, checkpointing, and inference. Vendors like Solidigm have different types of SSDs optimized for different stages of the pipeline. At the same time, controller vendors have taken advantage of one of the features introduced recently in the NVM Express standard - Flexible Data Placement (FDP).
FDP involves the host providing information / hints about the areas where the controller could place the incoming write data in order to reduce the write amplification. These hints are generated based on specific block sizes advertised by the device. The feature is completely backwards-compatible, with non-FDP hosts working just as before with FDP-enabled SSDs, and vice-versa.
Silicon Motion's MonTitan Gen 5 Enterprise SSD Platform was announced back in 2022. Since then, Silicon Motion has been touting the flexibility of the platform, allowing its customers to incorporate their own features as part of the customization process. This approach is common in the enterprise space, as we have seen with Marvell's Bravera SC5 SSD controller in the DapuStor SSDs and Microchip's Flashtec controllers in the Longsys FORESEE enterprise SSDs.
At FMS 2024, the company was demonstrating the advantages of flexible data placement by allowing a single QLC SSD based on their MonTitan platform to take part in different stages of the AI data pipeline while maintaining the required quality of service (minimum bandwidth) for each process. The company even has a trademarked name (PerformaShape) for the firmware feature in the controller that allows the isolation of different concurrent SSD accesses (from different stages in the AI data pipeline) to guarantee this QoS. Silicon Motion claims that this scheme will enable its customers to get the maximum write performance possible from QLC SSDs without negatively impacting the performance of other types of accesses.
Silicon Motion and Phison have market leadership in the client SSD controller market with similar approaches. However, their enterprise SSD controller marketing couldn't be more different. While Phison has gone in for a turnkey solution with their Gen 5 SSD platform (to the extent of not adopting the white label route for this generation, and instead opting to get the SSDs qualified with different cloud service providers themselves), Silicon Motion is opting for a different approach. The flexibility and customization possibilities can make platforms like the MonTitan appeal to flash array vendors.
Storage
![](https://twitter.com/share?url=https://compbuddey.blogspot.com/2024/07/tenstorrent-launches-wormhole-ai_25.html&text=Tenstorrent Launches Wormhole AI Processors: 466 FP8 TFLOPS at 300W <p align="center"><a href="https://www.anandtech.com/show/21482/tenstorrent-launches-wormhole-ai-processors-466-fp8-tflops-at-300w"><img src="https://images.anandtech.com/doci/21482/tesntorrent-wormhole-678_575px.jpg" alt="" /></a></p><p><p>Tenstorrent has unveiled its next-generation Wormhole processor for AI workloads that promises to offer decent performance at a low price. The company currently offers two add-on PCIe cards carrying one or two Wormhole processors as well as TT-LoudBox, and TT-QuietBox workstations aimed at software developers. The whole of today's release is aimed at developers rather than those who will deploy the Wormhole boards for their commercial workloads.</p>
<p>“<em>It is always rewarding to get more of our products into developer hands. Releasing development systems with our Wormhole™ card helps developers scale up and work on multi-chip AI software.</em>” said Jim Keller, CEO of Tenstorrent. “<em>In addition to this launch, we are excited that the tape-out and power-on for our second generation, Blackhole, is going very well.</em>”</p>
<p>Each Wormhole processor packs 72 Tensix cores (featuring five RISC-V cores supporting various data formats) with 108 MB of SRAM to deliver 262 FP8 TFLOPS at 1 GHz at 160W thermal design power. A single-chip Wormhole n150 card carries 12 GB of GDDR6 memory featuring a 288 GB/s bandwidth.</p>
<p>Wormhole processors offer flexible scalability to meet the varying needs of workloads. In a standard workstation setup with four Wormhole n300 cards, the processors can merge to function as a single unit, appearing as a unified, extensive network of Tensix cores to the software. This configuration allows the accelerators to either work on the same workload, be divided among four developers or run up to eight distinct AI models simultaneously. A crucial feature of this scalability is that it operates natively without the need for virtualization. In data center environments, Wormhole processors will scale both inside one machine using PCIe or outside of a single machine using Ethernet. </p>
<p>From performance standpoint, Tenstorrent's single-chip Wormhole n150 card (72 Tensix cores at 1 GHz, 108 MB SRAM, 12 GB GDDR6 at 288 GB/s) is capable of 262 FP8 TFLOPS at 160W, whereas the dual-chip Wormhole n300 board (128 Tensix cores at 1 GHz, 192 MB SRAM, aggregated 24 GB GDDR6 at 576 GB/s) can offer up to 466 FP8 TFLOPS at 300W (according to <a href="https://www.tomshardware.com/pc-components/gpus/tenstorrents-risc-v-based-wormhole-ai-accelerators-are-available-for-pre-order-today-pre-built-workstations-start-at-dollar12000">Tom's Hardware</a>).</p>
<p style="text-align: center;"><a href="https://www.anandtech.com/show/21482/tenstorrent-launches-wormhole-ai-processors-466-fp8-tflops-at-300w"><img alt="" src="https://images.anandtech.com/doci/21482/tesntorrent-wormhole_575px.jpg" /></a></p>
<p>To put that 466 FP8 TFLOPS at 300W number into context, let's compare it to what AI market leader Nvidia has to offer at this thermal design power. Nvidia's A100 does not support FP8, but it does support INT8 and its peak performance is 624 TOPS (1,248 TOPS with sparsity). By contrast, Nvidia's H100 supports FP8 and its peak performance is massive 1,670 TFLOPS (3,341 TFLOPS with sparsity) at 300W, which is a big difference from Tenstorrent's Wormhole n300. </p>
<p>There is a big catch though. Tenstorrent's Wormhole n150 is offered for $999, whereas n300 is available for $1,399. By contrast, one Nvidia H100 card can retail for $30,000, depending on quantities. Of course, we do not know whether four or eight Wormhole processors can indeed deliver the performance of a single H300, though they will do so at 600W or 1200W TDP, respectively.</p>
<p>In addition to cards, Tenstorrent offers developers pre-built workstations with four n300 cards inside the less expensive Xeon-based TT-LoudBox with active cooling and a premium EPYC-powered TT-QuietBox with liquid cooling.</p>
<p>Sources: <a href="https://tenstorrent.com/vision/tenstorrent-launches-next-generation-wormhole-based-developer-kits-and-workstations">Tenstorrent</a>, <a href="https://www.tomshardware.com/pc-components/gpus/tenstorrents-risc-v-based-wormhole-ai-accelerators-are-available-for-pre-order-today-pre-built-workstations-start-at-dollar12000">Tom's Hardware</a></p>
</p> AI){kind=link}
![](https://www.pinterest.com/pin/create/button/?url=https://compbuddey.blogspot.com/2024/07/tenstorrent-launches-wormhole-ai_25.html&media=https://lh3.googleusercontent.com/blogger_img_proxy/AEn0k_ukJHgcUIYfb79AL60V_M7xMHt4kvuLTMq_MSLWGXCoMY2MEXC-TtsV37TVVEGrUNJRMTMMWdaAeey0o4NmOnqBRzAcZQZAOkqWDb45aDhRR8jgRVfzlcnEOd3PlTHleIit9CxHY_u4iLQ72Xgw&description=Tenstorrent Launches Wormhole AI Processors: 466 FP8 TFLOPS at 300W <p align="center"><a href="https://www.anandtech.com/show/21482/tenstorrent-launches-wormhole-ai-processors-466-fp8-tflops-at-300w"><img src="https://images.anandtech.com/doci/21482/tesntorrent-wormhole-678_575px.jpg" alt="" /></a></p><p><p>Tenstorrent has unveiled its next-generation Wormhole processor for AI workloads that promises to offer decent performance at a low price. The company currently offers two add-on PCIe cards carrying one or two Wormhole processors as well as TT-LoudBox, and TT-QuietBox workstations aimed at software developers. The whole of today's release is aimed at developers rather than those who will deploy the Wormhole boards for their commercial workloads.</p>
<p>“<em>It is always rewarding to get more of our products into developer hands. Releasing development systems with our Wormhole™ card helps developers scale up and work on multi-chip AI software.</em>” said Jim Keller, CEO of Tenstorrent. “<em>In addition to this launch, we are excited that the tape-out and power-on for our second generation, Blackhole, is going very well.</em>”</p>
<p>Each Wormhole processor packs 72 Tensix cores (featuring five RISC-V cores supporting various data formats) with 108 MB of SRAM to deliver 262 FP8 TFLOPS at 1 GHz at 160W thermal design power. A single-chip Wormhole n150 card carries 12 GB of GDDR6 memory featuring a 288 GB/s bandwidth.</p>
<p>Wormhole processors offer flexible scalability to meet the varying needs of workloads. In a standard workstation setup with four Wormhole n300 cards, the processors can merge to function as a single unit, appearing as a unified, extensive network of Tensix cores to the software. This configuration allows the accelerators to either work on the same workload, be divided among four developers or run up to eight distinct AI models simultaneously. A crucial feature of this scalability is that it operates natively without the need for virtualization. In data center environments, Wormhole processors will scale both inside one machine using PCIe or outside of a single machine using Ethernet. </p>
<p>From performance standpoint, Tenstorrent's single-chip Wormhole n150 card (72 Tensix cores at 1 GHz, 108 MB SRAM, 12 GB GDDR6 at 288 GB/s) is capable of 262 FP8 TFLOPS at 160W, whereas the dual-chip Wormhole n300 board (128 Tensix cores at 1 GHz, 192 MB SRAM, aggregated 24 GB GDDR6 at 576 GB/s) can offer up to 466 FP8 TFLOPS at 300W (according to <a href="https://www.tomshardware.com/pc-components/gpus/tenstorrents-risc-v-based-wormhole-ai-accelerators-are-available-for-pre-order-today-pre-built-workstations-start-at-dollar12000">Tom's Hardware</a>).</p>
<p style="text-align: center;"><a href="https://www.anandtech.com/show/21482/tenstorrent-launches-wormhole-ai-processors-466-fp8-tflops-at-300w"><img alt="" src="https://images.anandtech.com/doci/21482/tesntorrent-wormhole_575px.jpg" /></a></p>
<p>To put that 466 FP8 TFLOPS at 300W number into context, let's compare it to what AI market leader Nvidia has to offer at this thermal design power. Nvidia's A100 does not support FP8, but it does support INT8 and its peak performance is 624 TOPS (1,248 TOPS with sparsity). By contrast, Nvidia's H100 supports FP8 and its peak performance is massive 1,670 TFLOPS (3,341 TFLOPS with sparsity) at 300W, which is a big difference from Tenstorrent's Wormhole n300. </p>
<p>There is a big catch though. Tenstorrent's Wormhole n150 is offered for $999, whereas n300 is available for $1,399. By contrast, one Nvidia H100 card can retail for $30,000, depending on quantities. Of course, we do not know whether four or eight Wormhole processors can indeed deliver the performance of a single H300, though they will do so at 600W or 1200W TDP, respectively.</p>
<p>In addition to cards, Tenstorrent offers developers pre-built workstations with four n300 cards inside the less expensive Xeon-based TT-LoudBox with active cooling and a premium EPYC-powered TT-QuietBox with liquid cooling.</p>
<p>Sources: <a href="https://tenstorrent.com/vision/tenstorrent-launches-next-generation-wormhole-based-developer-kits-and-workstations">Tenstorrent</a>, <a href="https://www.tomshardware.com/pc-components/gpus/tenstorrents-risc-v-based-wormhole-ai-accelerators-are-available-for-pre-order-today-pre-built-workstations-start-at-dollar12000">Tom's Hardware</a></p>
</p> AI){kind=link}
![](https://web.whatsapp.com/send?text=Tenstorrent Launches Wormhole AI Processors: 466 FP8 TFLOPS at 300W <p align="center"><a href="https://www.anandtech.com/show/21482/tenstorrent-launches-wormhole-ai-processors-466-fp8-tflops-at-300w"><img src="https://images.anandtech.com/doci/21482/tesntorrent-wormhole-678_575px.jpg" alt="" /></a></p><p><p>Tenstorrent has unveiled its next-generation Wormhole processor for AI workloads that promises to offer decent performance at a low price. The company currently offers two add-on PCIe cards carrying one or two Wormhole processors as well as TT-LoudBox, and TT-QuietBox workstations aimed at software developers. The whole of today's release is aimed at developers rather than those who will deploy the Wormhole boards for their commercial workloads.</p>
<p>“<em>It is always rewarding to get more of our products into developer hands. Releasing development systems with our Wormhole™ card helps developers scale up and work on multi-chip AI software.</em>” said Jim Keller, CEO of Tenstorrent. “<em>In addition to this launch, we are excited that the tape-out and power-on for our second generation, Blackhole, is going very well.</em>”</p>
<p>Each Wormhole processor packs 72 Tensix cores (featuring five RISC-V cores supporting various data formats) with 108 MB of SRAM to deliver 262 FP8 TFLOPS at 1 GHz at 160W thermal design power. A single-chip Wormhole n150 card carries 12 GB of GDDR6 memory featuring a 288 GB/s bandwidth.</p>
<p>Wormhole processors offer flexible scalability to meet the varying needs of workloads. In a standard workstation setup with four Wormhole n300 cards, the processors can merge to function as a single unit, appearing as a unified, extensive network of Tensix cores to the software. This configuration allows the accelerators to either work on the same workload, be divided among four developers or run up to eight distinct AI models simultaneously. A crucial feature of this scalability is that it operates natively without the need for virtualization. In data center environments, Wormhole processors will scale both inside one machine using PCIe or outside of a single machine using Ethernet. </p>
<p>From performance standpoint, Tenstorrent's single-chip Wormhole n150 card (72 Tensix cores at 1 GHz, 108 MB SRAM, 12 GB GDDR6 at 288 GB/s) is capable of 262 FP8 TFLOPS at 160W, whereas the dual-chip Wormhole n300 board (128 Tensix cores at 1 GHz, 192 MB SRAM, aggregated 24 GB GDDR6 at 576 GB/s) can offer up to 466 FP8 TFLOPS at 300W (according to <a href="https://www.tomshardware.com/pc-components/gpus/tenstorrents-risc-v-based-wormhole-ai-accelerators-are-available-for-pre-order-today-pre-built-workstations-start-at-dollar12000">Tom's Hardware</a>).</p>
<p style="text-align: center;"><a href="https://www.anandtech.com/show/21482/tenstorrent-launches-wormhole-ai-processors-466-fp8-tflops-at-300w"><img alt="" src="https://images.anandtech.com/doci/21482/tesntorrent-wormhole_575px.jpg" /></a></p>
<p>To put that 466 FP8 TFLOPS at 300W number into context, let's compare it to what AI market leader Nvidia has to offer at this thermal design power. Nvidia's A100 does not support FP8, but it does support INT8 and its peak performance is 624 TOPS (1,248 TOPS with sparsity). By contrast, Nvidia's H100 supports FP8 and its peak performance is massive 1,670 TFLOPS (3,341 TFLOPS with sparsity) at 300W, which is a big difference from Tenstorrent's Wormhole n300. </p>
<p>There is a big catch though. Tenstorrent's Wormhole n150 is offered for $999, whereas n300 is available for $1,399. By contrast, one Nvidia H100 card can retail for $30,000, depending on quantities. Of course, we do not know whether four or eight Wormhole processors can indeed deliver the performance of a single H300, though they will do so at 600W or 1200W TDP, respectively.</p>
<p>In addition to cards, Tenstorrent offers developers pre-built workstations with four n300 cards inside the less expensive Xeon-based TT-LoudBox with active cooling and a premium EPYC-powered TT-QuietBox with liquid cooling.</p>
<p>Sources: <a href="https://tenstorrent.com/vision/tenstorrent-launches-next-generation-wormhole-based-developer-kits-and-workstations">Tenstorrent</a>, <a href="https://www.tomshardware.com/pc-components/gpus/tenstorrents-risc-v-based-wormhole-ai-accelerators-are-available-for-pre-order-today-pre-built-workstations-start-at-dollar12000">Tom's Hardware</a></p>
</p> AI | https://compbuddey.blogspot.com/2024/07/tenstorrent-launches-wormhole-ai_25.html){kind=link}
![](mailto:?subject=Tenstorrent Launches Wormhole AI Processors: 466 FP8 TFLOPS at 300W <p align="center"><a href="https://www.anandtech.com/show/21482/tenstorrent-launches-wormhole-ai-processors-466-fp8-tflops-at-300w"><img src="https://images.anandtech.com/doci/21482/tesntorrent-wormhole-678_575px.jpg" alt="" /></a></p><p><p>Tenstorrent has unveiled its next-generation Wormhole processor for AI workloads that promises to offer decent performance at a low price. The company currently offers two add-on PCIe cards carrying one or two Wormhole processors as well as TT-LoudBox, and TT-QuietBox workstations aimed at software developers. The whole of today's release is aimed at developers rather than those who will deploy the Wormhole boards for their commercial workloads.</p>
<p>“<em>It is always rewarding to get more of our products into developer hands. Releasing development systems with our Wormhole™ card helps developers scale up and work on multi-chip AI software.</em>” said Jim Keller, CEO of Tenstorrent. “<em>In addition to this launch, we are excited that the tape-out and power-on for our second generation, Blackhole, is going very well.</em>”</p>
<p>Each Wormhole processor packs 72 Tensix cores (featuring five RISC-V cores supporting various data formats) with 108 MB of SRAM to deliver 262 FP8 TFLOPS at 1 GHz at 160W thermal design power. A single-chip Wormhole n150 card carries 12 GB of GDDR6 memory featuring a 288 GB/s bandwidth.</p>
<p>Wormhole processors offer flexible scalability to meet the varying needs of workloads. In a standard workstation setup with four Wormhole n300 cards, the processors can merge to function as a single unit, appearing as a unified, extensive network of Tensix cores to the software. This configuration allows the accelerators to either work on the same workload, be divided among four developers or run up to eight distinct AI models simultaneously. A crucial feature of this scalability is that it operates natively without the need for virtualization. In data center environments, Wormhole processors will scale both inside one machine using PCIe or outside of a single machine using Ethernet. </p>
<p>From performance standpoint, Tenstorrent's single-chip Wormhole n150 card (72 Tensix cores at 1 GHz, 108 MB SRAM, 12 GB GDDR6 at 288 GB/s) is capable of 262 FP8 TFLOPS at 160W, whereas the dual-chip Wormhole n300 board (128 Tensix cores at 1 GHz, 192 MB SRAM, aggregated 24 GB GDDR6 at 576 GB/s) can offer up to 466 FP8 TFLOPS at 300W (according to <a href="https://www.tomshardware.com/pc-components/gpus/tenstorrents-risc-v-based-wormhole-ai-accelerators-are-available-for-pre-order-today-pre-built-workstations-start-at-dollar12000">Tom's Hardware</a>).</p>
<p style="text-align: center;"><a href="https://www.anandtech.com/show/21482/tenstorrent-launches-wormhole-ai-processors-466-fp8-tflops-at-300w"><img alt="" src="https://images.anandtech.com/doci/21482/tesntorrent-wormhole_575px.jpg" /></a></p>
<p>To put that 466 FP8 TFLOPS at 300W number into context, let's compare it to what AI market leader Nvidia has to offer at this thermal design power. Nvidia's A100 does not support FP8, but it does support INT8 and its peak performance is 624 TOPS (1,248 TOPS with sparsity). By contrast, Nvidia's H100 supports FP8 and its peak performance is massive 1,670 TFLOPS (3,341 TFLOPS with sparsity) at 300W, which is a big difference from Tenstorrent's Wormhole n300. </p>
<p>There is a big catch though. Tenstorrent's Wormhole n150 is offered for $999, whereas n300 is available for $1,399. By contrast, one Nvidia H100 card can retail for $30,000, depending on quantities. Of course, we do not know whether four or eight Wormhole processors can indeed deliver the performance of a single H300, though they will do so at 600W or 1200W TDP, respectively.</p>
<p>In addition to cards, Tenstorrent offers developers pre-built workstations with four n300 cards inside the less expensive Xeon-based TT-LoudBox with active cooling and a premium EPYC-powered TT-QuietBox with liquid cooling.</p>
<p>Sources: <a href="https://tenstorrent.com/vision/tenstorrent-launches-next-generation-wormhole-based-developer-kits-and-workstations">Tenstorrent</a>, <a href="https://www.tomshardware.com/pc-components/gpus/tenstorrents-risc-v-based-wormhole-ai-accelerators-are-available-for-pre-order-today-pre-built-workstations-start-at-dollar12000">Tom's Hardware</a></p>
</p> AI&body=https://compbuddey.blogspot.com/2024/07/tenstorrent-launches-wormhole-ai_25.html){kind=link}
Lorem Ipsum is simply dummy text of the printing and typesetting industry. Lorem Ipsum has been the industry's.
Kioxia Details BiCS 8 NAND at FMS 2024: 218 Layers With Superior Scaling
Kioxia's booth at FMS 2024 was a busy one with multiple technology demonstrations keeping visitors occupied. A walk-through of the BiCS 8 manufacturing process was the first to grab my attention. Kioxia and Western Digital announced the sampling of BiCS 8 in March 2023. We had touched briefly upon its CMOS Bonded Array (CBA) scheme in our coverage of Kioxial's 2Tb QLC NAND device and coverage of Western Digital's 128 TB QLC enterprise SSD proof-of-concept demonstration. At Kioxia's booth, we got more insights.
Traditionally, fabrication of flash chips involved placement of the associate logic circuitry (CMOS process) around the periphery of the flash array. The process then moved on to putting the CMOS under the cell array, but the wafer development process was serialized with the CMOS logic getting fabricated first followed by the cell array on top. However, this has some challenges because the cell array requires a high-temperature processing step to ensure higher reliability that can be detrimental to the health of the CMOS logic. Thanks to recent advancements in wafer bonding techniques, the new CBA process allows the CMOS wafer and cell array wafer to be processed independently in parallel and then pieced together, as shown in the models above.
The BiCS 8 3D NAND incorporates 218 layers, compared to 112 layers in BiCS 5 and 162 layers in BiCS 6. The company decided to skip over BiCS 7 (or, rather, it was probably a short-lived generation meant as an internal test vehicle). The generation retains the four-plane charge trap structure of BiCS 6. In its TLC avatar, it is available as a 1 Tbit device. The QLC version is available in two capacities - 1 Tbit and 2 Tbit.
Kioxia also noted that while the number of layers (218) doesn't compare favorably with the latest layer counts from the competition, its lateral scaling / cell shrinkage has enabled it to be competitive in terms of bit density as well as operating speeds (3200 MT/s). For reference, the latest shipping NAND from Micron - the G9 - has 276 layers with a bit density in TLC mode of 21 Gbit/mm2, and operates at up to 3600 MT/s. However, its 232L NAND operates only up to 2400 MT/s and has a bit density of 14.6 Gbit/mm2.
It must be noted that the CBA hybrid bonding process has advantages over the current processes used by other vendors - including Micron's CMOS under array (CuA) and SK hynix's 4D PUC (periphery-under-chip) developed in the late 2010s. It is expected that other NAND vendors will also move eventually to some variant of the hybrid bonding scheme used by Kioxia.
Storage
0 Comments