Wednesday, May 11th 2022

Tachyum Delivers the Highest AI and HPC Performance with the Launch of the World's First Universal Processor

Tachyum today launched the world's first universal processor, Prodigy, which unifies the functionality of a CPU, GPU and TPU in a single processor, creating a homogeneous architecture, while delivering massive performance improvements at a cost many times less than competing products.

After the company undertook its mission to conquer the processor performance plateau in nanometer-class chips and the systems they power, Tachyum has succeeded by launching its first commercial product. The Prodigy Cloud/AI/HPC supercomputer processor chip offers 4x the performance of the fastest Xeon, has 3x more raw performance than NVIDIA's H100 on HPC and has 6x more raw performance on AI training and inference workloads, and up to 10x performance at the same power. Prodigy is poised to overcome the challenges of increasing data center power consumption, low server utilization and stalled performance scaling.
Among the highlights of the newly launched Prodigy processor are:
  • 128 high-performance unified 64-bit cores running up to 5.7 GHz
  • 16 DDR5 memory controllers
  • 64 PCIe 5.0 lanes
  • Multiprocessor support for 4-socket and 2-socket platforms
  • Rack solutions for both air-cooled and liquid-cooled data centers
  • SPECrate 2017 Integer performance of around 4x Intel 8380 and around 3x AMD 7763HPC
  • Double-Precision Floating-Point performance is 3x NVIDIA H100
  • AI FP8 performance is 6x NVIDIA H100
Unlike other CPU and GPU solutions, Tachyum's Prodigy was designed to handle matrix and vector processing from the ground up, rather than as an afterthought. Among Prodigy's vector and matrix features are support for a range of data types (FP64, FP32, TF32, BF16, Int8, FP8 and TAI); 2x1024-bit vector units per core; AI sparsity and super-sparsity support; and no penalty for misaligned vector loads or stores when crossing cache lines. This built-in support offers high performance for AI training and inference workloads, increases performance and reduces memory utilization.

Prodigy is significantly better than the best-performing processors currently available in hyperscale, HPC and AI markets. Prodigy delivers up to 3x the performance of the highest-performing x86 processors for cloud workloads, up to 3x compared to the highest-performing GPUs for HPC, and up to 6x for AI applications. By increasing performance while using less power, Prodigy solves the problem of sustainable data center growth by offering unparalleled carbon footprint reduction. This is especially important as the universality of AI continues to gain traction: Prodigy will enable unprecedented data center TCO savings as part of this new-world market.

"We have long believed in our ability to overcome Moore's Law to transform hyperscale data centers into true universal computing centers. With the launch of Prodigy, we have begun the revolution," said Dr. Radoslav Danilak, founder and CEO of Tachyum. "Prodigy's ability to enable human brain-scale AI while simultaneously reducing data center power consumption and lowering the TCO of hyperscale data centers and supercomputer systems, is a breakthrough for a projected $100 billion industry. By launching Prodigy, we are advancing not only the state of technology but making the world a greener place as well."

Sampling for Prodigy will begin later this year with volume production taking place in the first half of 2023. Tachyum's Prodigy family includes eight products, ranging from the 128-core HPC/AI at the high end to a 32-core lowest-power version, to address a wide range of markets, including cloud, supercomputing, Big AI and edge. Early engagement customers wishing to see how Prodigy will benefit them can order an evaluation platform by contacting Tachyum at https://www.tachyum.com/contact/
Add your own comment

20 Comments on Tachyum Delivers the Highest AI and HPC Performance with the Launch of the World's First Universal Processor

#1
zlobby
Ah! The wunder processor!

Honestly, after so many years, delays and missed targets, the whole thing looks like a presentation from Gavin Belson.

I'd love to see some real-world reviews or tests, though. Who knows, maybe it really sparkles and outputs unicorns?
Posted on Reply
#2
Steevo
So a big FPGA built on a small node with some risc cores. I’m guessing it’s built by AI by some very competent people, but their claims for performance come with no reference to any workloads and may represent hand tuned software with very little branching or page faults rewritten in code native to their ISA.

If we had all the answers to the problems presented and they were pre programmed into an array it would run anything at Uber speed. Looking at their site the memory spec is probably a big key to the overall performance due to the lack of cache that is required for Out of Order operations with a great branch predictor and speculative preprocessing. 128 MB L2 & L3 for 128 cores is very good for in order pipelines that are shallow.

I hope to see some independent benchmarks validate their claims.
Posted on Reply
#3
DeathtoGnomes
Define Universal. Not what TPU means there, but its not TechPowerUp. :rockout:We've already had CPU/GPU chips for years.
AleksandarK3x compared to the highest-performing GPUs for HPC, and up to 6x for AI applications.
I've got gas thats 3x more potent than the highest performing antacid for home use....

I'm with @Steevo on this, bring on the reviews.
Posted on Reply
#4
lexluthermiester
zlobbyI'd love to see some real-world reviews or tests, though.
SteevoI hope to see some independent benchmarks validate their claims.
DeathtoGnomesI'm with @Steevo on this, bring on the reviews.
Agreed! The talk is big, let's see the the action!
DeathtoGnomesI've got gas thats 3x more potent than the highest performing antacid for home use....
Thank You for THAT mental image.:slap:
Posted on Reply
#5
thegnome
5.7 Ghz? That must be some special cores then.
Posted on Reply
#6
Errant1
I'm just curious if this would translate to a gaming rig. Processors are one thing but there is the whole aspect of motherboard, I/O, memory, etc. Where is that infrastructure to support it, proprietary universal?
Posted on Reply
#7
lexluthermiester
Errant1I'm just curious if this would translate to a gaming rig.
No. Perhaps you did not read the title of the article...
AleksandarK

Tachyum Delivers the Highest AI and HPC Performance

...thus.
Posted on Reply
#8
Errant1
lexluthermiesterNo. Perhaps you did not read the title of the article...

...thus.
Ment the technology, as to infer. Should have iterated the point.
Posted on Reply
#9
lexluthermiester
Errant1Ment the technology, as to infer. Should have iterated the point.
Fair enough. However, the technology on offer here is very unlikely to have any gaming focus. Perhaps it could be reconfigured but it's unlikely. If you visit their web page you will see what their focus is.
www.tachyum.com/

I generally stop taking a company seriously when they display meaningless nothing awards front and center.
Posted on Reply
#10
Steevo
Errant1I'm just curious if this would translate to a gaming rig. Processors are one thing but there is the whole aspect of motherboard, I/O, memory, etc. Where is that infrastructure to support it, proprietary universal?
it’s a complete server from what I see, you buy the complete hardware package just like a business would buy a whole cray or Iseries server.

Gaming it would probably be like my Phenom2 in performance, too little cache to load cores fully and suffer from page faults due to heavily branched threads.
Posted on Reply
#11
Denver
I don't know why... this all sounds unrealistic... "too good to be real".
Posted on Reply
#12
zlobby
DenverI don't know why...
For a reason! And a good one, indeed!
Posted on Reply
#13
Frank_100
"Prodigy's ability to enable human brain-scale AI"


da da da Dunt

da da da Dunt
Posted on Reply
#14
AnarchoPrimitiv
I'll give it a week before we hear about Intel trying to buy this company
Posted on Reply
#15
Denver
zlobbyFor a reason! And a good one, indeed!
Yeah, A silly expression, not to say that "this stinks, and there's no way it will have 3x more performance than an Epyc".
Posted on Reply
#16
zlobby
Frank_100"Prodigy's ability to enable human brain-scale AI"


da da da Dunt

da da da Dunt
Given that some healthy people's mental capabilities can be reproduced by a handful of discrete transistors, thier claim may actually hold. :D
Posted on Reply
#17
R-T-B
lexluthermiesterI generally stop taking a company seriously when they display meaningless nothing awards front and center.
So, every car company ever then?

I'm with you, just pointing out the obvious.
Posted on Reply
#18
lexluthermiester
R-T-BSo, every car company ever then?
Fair point.

Of course, with car companies there are always third party reviewers and consumer reports, but I digress...
Posted on Reply
#19
yanbin_c
16ch ddr with 7200 ??

Sounds unreal ...

there is no 7200 RDIMM available yet.

latest ES is 6400 according to JEDEC meeting notes.
Posted on Reply
#20
lexluthermiester
yanbin_cthere is no 7200 RDIMM available yet.
It's called planning for the future.
Posted on Reply
Add your own comment
May 28th, 2022 03:46 EDT change timezone

New Forum Posts

Popular Reviews

Controversial News Posts