News Intel Discontinues Bitcoin-Mining Blockscale Chips, No Future Gens Announced

[Admin]

Intel released a bulletin that announces it is ending production of all of its known bitcoin-mining Blockscale ASICs.

Intel Discontinues Bitcoin-Mining Blockscale Chips, No Future Gens Announced : Read more

 

[gg83]

Lmao! This makes me think they never wanted to do it in the first place and it was just to appease the stupid investors.

 

[teodoreh]

"Came in late, leaving too early." should be their next motto.
Solid State disks
mobile processors
Mining chips
Next stop: Graphics cards

 

[TerryLaze]

"Came in late, leaving too early." should be their next motto.
Solid State disks
mobile processors
Mining chips
Next stop: Graphics cards

As a great philosopher once said:
“You miss 100% of the shots you don't take"
Companies have to try new things if they want to get bigger and they need to know when to cut losses if they don't want to get smaller.

 

[kjfatl]

I suspect they will continue with graphics cards. They need a tile to drop into their processors that need graphics equivalent to today's mid-range graphics cards This won't make them a lot of money but will kill 1/2 of the PC graphics card market. For those doing AI and similar things Intel is more likely to fab parts for Nvidia and AMD.

 

[JamesJones44]

I suspect they will continue with graphics cards. They need a tile to drop into their processors that need graphics equivalent to today's mid-range graphics cards This won't make them a lot of money but will kill 1/2 of the PC graphics card market. For those doing AI and similar things Intel is more likely to fab parts for Nvidia and AMD.

This one is a fairly simple equation in my opinion. GPU demand is growing and growing quickly, CPUs, mining equipment, etc. are fairly flat and the trends looks like they will be this way for at least a few years. It would be crazy to cut the GPU division in this environment, but it's hard to put anything past Intel.

 

[2Be_or_Not2Be]

This one is a fairly simple equation in my opinion. GPU demand is growing and growing quickly, CPUs, mining equipment, etc. are fairly flat and the trends looks like they will be this way for at least a few years. It would be crazy to cut the GPU division in this environment, but it's hard to put anything past Intel.

I don't know about GPU demand growing quickly. I think you had a spike before that was linked to crypto (multiple times), but I think that's leveling off now and even dropping. Intel doesn't seem to be a competitor w/AMD & Nvidia above the integrated graphics waterline, and I don't know if they have the willpower to keep trying to make their dGPUs better both in hardware & software, much less sales, long-term. I could see them getting out of the dGPU market again.

 

[sivaseemakurthi]

Year 2026:
"Intel to exit CPU business to prioritize IDM2.0"

 

[JamesJones44]

I don't know about GPU demand growing quickly. I think you had a spike before that was linked to crypto (multiple times), but I think that's leveling off now and even dropping. Intel doesn't seem to be a competitor w/AMD & Nvidia above the integrated graphics waterline, and I don't know if they have the willpower to keep trying to make their dGPUs better both in hardware & software, much less sales, long-term. I could see them getting out of the dGPU market again.

In datacenter it is going gang busters. For example Musk just ordered 10k A100's at 10k a pop for his "TruthAI". Add in all of the companies racing to get their version of LLM running and Nvidia is poised to capture yet another upswing in GPU sales. Only this time it's on very high margin enterprise customers.

On the consumer side I agree, things are largely level, but with the AI explosion, datacenter is going to be where the sales stay hot for a few years IMO.

 

[bit_user]

For those doing AI and similar things Intel is more likely to fab parts for Nvidia and AMD.

Intel bought Habana Labs, several years ago. They continue to crank out new products, though I have no idea how market uptake is going:

https://habana.ai/products/
​

In datacenter it is going gang busters. For example Musk just ordered 10k A100's at 10k a pop for his "TruthAI". Add in all of the companies racing to get their version of LLM running and Nvidia is poised to capture yet another upswing in GPU sales. Only this time it's on very high margin enterprise customers.

Just like bitcoin mining eventually moved past GPUs, AI will do the same. If you compare the power-efficiency of Nvidia's H100 to Cerebras' WSE2, they're not even in the same league. AI accelerators favor a dataflow architecture and gobs of on-die SRAM. Conversely, hardware like fp64 and even most of the fp32 capability that the H100 features are largely a waste of die space, for AI.

 

[Kamen Rider Blade]

Nothing of value was lost.

 

[edzieba]

"Came in late, leaving too early." should be their next motto.
Solid State disks

Have people forgotten the early days of consumer SSDs so soon? It was a case of either using an Intel X25 or X18 and having good random read performance, or a slightly cheaper drive with a JMicron controller with single-digit IOPS and write latencies hitting 1 full SECOND.

mobile processors

Pentium-M and the Centrino shipset umbrella (mandating WiFi as a standard feature) upended the laptop industry. Laptops were beating desktops outright on CPU performance at that time.

Plus, there's the X86 architecture itself. Imagining Intel as some always-late slow moving juggernaut that never innovates makes for nice fantasy, but does not match reality.

 

[JamesJones44]

Just like bitcoin mining eventually moved past GPUs, AI will do the same. If you compare the power-efficiency of Nvidia's H100 to Cerebras' WSE2, they're not even in the same league. AI accelerators favor a dataflow architecture and gobs of on-die SRAM. Conversely, hardware like fp64 and even most of the fp32 capability that the H100 features are largely a waste of die space, for AI.

That's why in the next paragraph under that quote I put "datacenter is going to be where the sales stay hot for a few years" .

Eventually they will migrate to chips targeted specifically for AI, who wins that round is up in the air, but it's going to take a few years to get to something generic that non Fortune 50 can use/afford without making their own chips.