Haswell Celerons officially launched.

[Torn Mind]

Yep :thumbsup:

They could instead have segmented it with something like 128bit cache only for Celerons/Pentiums. And 256bit on i3/i5/i7.

It's price discrimination(I think 3rd degree). Probably to induce one more upgrade cycle from the buyers who do eventually figure out they need to go bigger(as in $120) to get fuller-featured chips. I suppose they thought that even a limited implementation would reduce the amount of people who upgrade their chips just to get AVX or AVX2 support.

 

[ElFenix]

It's price discrimination(I think 3rd degree). Probably to induce one more upgrade cycle from the buyers who do eventually figure out they need to go bigger(as in $120) to get fuller-featured chips. I suppose they thought that even a limited implementation would reduce the amount of people who upgrade their chips just to get AVX or AVX2 support.

these are going into $300 desktops. AVX isn't something thought of by the purchasing manager.

 

[NTMBK]

It's price discrimination(I think 3rd degree). Probably to induce one more upgrade cycle from the buyers who do eventually figure out they need to go bigger(as in $120) to get fuller-featured chips. I suppose they thought that even a limited implementation would reduce the amount of people who upgrade their chips just to get AVX or AVX2 support.

So in order to secure a miniscule number of upgrade sales, they should retard the development of AVX software?

 

[Shivansps]

Actually, the competition are the Sempron/Athlons based on FS1b as well as A4-4000, but im not sure if A4-4000 gona have a Kaveri reemplacement.

 

[Yuriman]

Actually, the competition are the Sempron/Athlons based on FS1b as well as A4-4000, but im not sure if A4-4000 gona have a Kaveri reemplacement.

Indeed, and Kaveri is a pretty big die compared to Intel's duals.

 

[Insert_Nickname]

Indeed, and Kaveri is a pretty big die compared to Intel's duals.

Why not? It seems a good place to dump partially defective chips. AMD wont make a lot of money doing that, but if they can cover the cost of fabbing that chip instead of throwing it away...

I -think- I saw a listing for an A4-7300 somewhere. One Kaveri module + 192 shader GPU.

 

[Torn Mind]

So in order to secure a miniscule number of upgrade sales, they should retard the development of AVX software?

I'm not saying they should be doing this, but that is what they seem to be doing. It's like them disabling TSX-NI on the "K" chips in an apparent attempt to stimulate sales of non-K chips, perhaps with the assumption that many buyers won't notice and would have upgraded by the time TSX-NI has become ubiquitous.

Seems like that is the case. Apparently, AMD's A4-4000 has not even phased them implementing it despite it being the competition and that chip actually does support AVX. I don't know much about the CPU hardware itself, but since you guys said that AVX is an integral part the FPU, I think its quite clear they have some reason they're doing this at the consumer's expense.

Or maybe they think that adoption won't be wide

 

[sm625]

2MB of cache.... god these guys are so chincy they may as well just sell their fabs and use the money to buy apple stock. No one is going to buy a pc to get ripped off by intel. Why do that when you can buy an ipad and get ripped off by apple for the same price?

 

[Blue_Max]

Sheesh... what more do you expect for $40??

 

[Torn Mind]

2MB of cache.... god these guys are so chincy they may as well just sell their fabs and use the money to buy apple stock. No one is going to buy a pc to get ripped off by intel. Why do that when you can buy an ipad and get ripped off by apple for the same price?

Because 2 MB of cache isn't going to slow down that many things to begin with? Where as the Ipad's more limited environment and means of interfacing is not suitable for some things that one wants to do on a nice monitor and keyboard. The chip is still faster than most inexpensive laptops and even some expensive ones. It's Core 2 Duo performance in a at $45 price instead of $180.

 

[StrangerGuy]

So in order to secure a miniscule number of upgrade sales, they should retard the development of AVX software?

Don't you know Intel won't sleep well at night if people bought $50 chips with AVX that are already massively overpowered for normal usage on non-AVX software instead of a $200+ chip that is even more so? Poor peasant scum shouldn't deserve nice things like AVX.

One would have thought that completely missing out on mobile may have taught them a few things.

 

[VirtualLarry]

Don't you know Intel won't sleep well at night if people bought $50 chips with AVX that are already massively overpowered for normal usage on non-AVX software instead of a $200+ chip that is even more so? Poor peasant scum shouldn't deserve nice things like AVX.

One would have thought that completely missing out on mobile may have taught them a few things.

Not to mention, giving ARM a one-up on themselves, by not embracing the AVX/AVX2 ISA stack top-to-bottom on their chips. HSA will be ubiquitous before you know it (even on ARM!), and Intel's AVX will only be a footnote in history. Seriously, who would be foolish enough to code for it (HPC applications excepted, since those are primarily bespoke), given how small a percentage of Intel Core CPUs can even run the code?

 

[ninaholic37]

It's Core 2 Duo performance in a at $45 price instead of $180.

You can find Core2Duo chips on ebay for $5 actually. Why spend 900% more?

 

[crashtech]

Hyperbole much? The MT performance of the Celeron G1820 is roughly on par with a Q8300 fully loaded, but the single threaded performance of the G1820 is over 60% better, while consuming roughly half the power. The cheapest used Q8300 was about $10 less than the Celeron.

So much for the Core2 comparisons...

 

[cbn]

Not to mention, giving ARM a one-up on themselves, by not embracing the AVX/AVX2 ISA stack top-to-bottom on their chips. HSA will be ubiquitous before you know it (even on ARM!), and Intel's AVX will only be a footnote in history. Seriously, who would be foolish enough to code for it (HPC applications excepted, since those are primarily bespoke), given how small a percentage of Intel Core CPUs can even run the code?

Speaking of HSA, I do find it somewhat ironic that AMD includes AVX on the Jaguar/Steamroller CPU cores but Intel purposely omits it on their Celeron/Pentium "Core" line-up.

Hopefully Intel can find a better way to differentiate their chips in the future.

 

[Ken g6]

Not to mention, giving ARM a one-up on themselves, by not embracing the AVX/AVX2 ISA stack top-to-bottom on their chips. HSA will be ubiquitous before you know it (even on ARM!), and Intel's AVX will only be a footnote in history.

I wonder if Intel is preparing for their own HSA-type system here. Supposedly, Intel is preparing for wider AVX systems - AVX-512 or higher. I wonder if they're going to shift the AVX work over to the integrated GPU instead?

The current Intel HD graphics architecture has 8 32-bit floats per execution unit (or 256 bits, same as AVX), and a latency of 8 cycles. If you allow AVX latency to jump to 8 cycles (that's the key), then 8 HT cores could feed 1 HD graphics EU, or more likely 4 cores could feed it AVX-512 instructions. Then, if you allow AVX to extend over as many EUs as there are in the processor, you get a real equivalent to HSA without having to use OpenCL.

But what about Pentiums and Celerons? If they used the same system, they only have two cores, so they could only keep each EU half-occupied. Maybe that's what Intel's setting up here: a situation where Pentiums and Celerons are only half as fast as their other processors at vectorized work? Probably a stretch, but interesting to think about anyway.