Intel plans to rake you over their coals.

[kc5tja]

It looks like I'll need to work harder and more diligently on completing my Kestrel-2 design. Just one more reason to abandon the x86 platform like a hot potato. It might run at a snail's pace of 12MHz to 20MHz, but doggone it, it'll be mine, and it'll operate as intended with no surprises. This makes me frothingly livid.

Intel wants to charge $50 to unlock stuff your CPU can already do

[GARTHWILSON]

this, in addition to your recent post here which was so well put

BTW, besides the fact that 1MHz Commodore 64's have been used as servers, and the performance you pointed out with the SuperCPU addition, and some very professional-sounding MIDI music produced on C64's (with MIDI instruments, not the SID), did you see that this year (2010), someone even used a C64 for MP3 encoding and decoding?
http://www.youtube.com/watch?v=eM3yp7U5 ... re=related
mahoney.c64.org which they mention in the YouTube video has a link to http://www.livet.se/mahoney/ which in turn links to the http://www.ejeson.se/mahoney/c64-files/ ... vesson.zip file which contains the README.txt file with the explanation. This song, "Tom's Diner," sung a capella by Suzzane Vega, is apparently one of the most difficult to compress and handle by MP3, so he did it on the C64 as a demo. It sounds kind of toy-like, but I'm sure it wouldn't with an '816 and a faster clock speed.

Edit, 2/5/19: Check out the various settings that can be changed on the fly: https://www.youtube.com/watch?v=MDrqBYkco-Y

[kc5tja]

GARTHWILSON wrote:

It sounds kind of toy-like, but I'm sure it wouldn't with an '816 and a faster clock speed.

I doubt it; I suspect the problem here is the limited dynamic range of the SID chip when (ab)used like it is. He actually would have had a much better audio drive if he'd modulated the volume register instead.

I downloaded it and ran it in VICE, and it sounds pretty awful. VICE's SID emulation does accurately reproduce the audio, but it's handling of the oscillators is such that it just dwarfs the audio produced by the demo. I'm not sure why a real SID wouldn't similarly produce the ear-screeching tones reSID would, except perhaps that the real SID has a low-pass filter that takes care of the high frequencies heard in the reSID output.

If he used a proper 8-bit resistor-ladder DAC on the user-port, he would have gotten stellar audio, even at 1MHz (it would have competed well with the Amiga). At 20MHz, he would have had a few extra bits of audio resolution and actually exceeded the Amiga's quality.

That being said, I remain somewhat skeptical that it's decoding an MP3 in real-time. I do not know at what clock speed a 6502 would be adequate for real-time decoding; but, 1MHz leaves me skeptical. I note the long lag at program start-up seems to support my pre-decoding hypothesis.

Still pretty neat though!

Update: The encoder runs in Matlab; a C64 did not produce the MP3 file.

[jt_eaton]

This makes me frothingly livid.

Intel wants to charge $50 to unlock stuff your CPU can already do

------------------------------------------------------------

It looks like Intel marketing is taking things to a new level. This allows them to offer a lower entry price for their part to attract more buyers and then gives them a way to convert these buyers to a secondary sale. They will sell to power users who need ( or think they need) a boost as well as end of life sales as users see a $50 upgrade as better than having to buy a new pc. I'm guessing they will have a little sticker that you can put on your pc to show that you have the upgrade.

They have to use some serious encryption if they expect this to work and it doesn't appear that they have any way for you to enter your cpu's serial number so I expect that the upgrade program needs to phone the mothership to get the key.

John Eaton

[kc5tja]

I suspect they are using the writable control store functionality that most Intel CPUs have[1] to enable/disable these features.

That being said, I have no intention of purchasing a piece of silicon that I cannot use to its fullest. Unused circuitry still draws power that I have to pay for.

[1] - ever since the floating point division bug and the F00F bug, Intel has implemented writable control stores for their CPUs to allow a certain degree of post-silicon bug fixing capability.

[jt_eaton]

Unused circuitry still draws power that I have to pay for.

------------------------

Not any more. You isolate blocks on their own switched power island and only turn on when they are needed.

Si is now so cheap it is a commodity. Buy a coke at McD's and it only costs them about $.02 cents (less than the cup and ice). It is mostly profit , thats why you get free refills. Now intel is giving free silicon.

Your cost to buy a CPU pays many things. Besides the cost to make a die and package and test it you also have to pay for warranty costs. Each chip they sell puts a certain amount of money into a pool that is used to pay for replacing parts that fail during the warranty period. Well guess what? If Intel doesn't charge for that feature then they don't have to replace the chip if it doesn't work. Thats where they save their money by paying less into the pool.

Part of the $50 upgrade price goes into this pool. You are not buying Si, you are buying an insurance policy.

My guess is that at least half or more of the price goes to Best Buy.

John Eaton

[kc5tja]

Quote:

Not any more. You isolate blocks on their own switched power island and only turn on when they are needed.

You still have leakage current. And, as others on this forum have pointed out in other conversations, the smaller the feature size, the larger the current that leaks.

Quote:

Now intel is giving free silicon.

Obviously not, because you have to pay to use it. If it's so free, why can't I just use it off-the-shelf?

Quote:

If Intel doesn't charge for that feature then they don't have to replace the chip if it doesn't work. Thats where they save their money by paying less into the pool.

This is a fallacy; it looks nice in print, but it doesn't accurately reflect reality. Intel has no control what-so-ever over who buys what chips. As far as they're concerned, chip distribution resembles random distribution. If they're distributing chips with faulty L3 cache implementations, and using this program as an excuse to increase their effective yields, then any fraction of customers who crash their machines because of faulty L3 cache implementations will result in Intel paying more in warranty expenses as a result.

The question is, will warranty replacements exceed the money spent properly binning the chips at the factory, pre-sale? I argue the answer is no, because then Intel will need to pay not only for a replaced chip, but potentially also legal fees, as I predict a number of such chips will appear in business systems.

Intel will feel financially motivated to reduce warranty replacements at all costs anyway (it's how all successful businesses work), so that $50 is nothing more than an ownership tax. Plain and simple.

Quote:

You are not buying Si, you are buying an insurance policy.

Then I must re-assert my position: I am not spending my money on insurance. I am spending my money on silicon. Capitalism and its benefits are predicated on the principle of (responsible) private ownership. The idea of purchasing a computer so that I effectively pay taxes on it, payable to some third party (whether it's Intel or the government), is so revolting that it makes me nauseous.

Quote:

My guess is that at least half or more of the price goes to Best Buy.

When I was provisioning engineer at CariNet, our prices for CPUs weren't substantially less than what you could purchase at other retailers. Sure, we got bulk discounts, and wholesaler discounts, but we were still in the ballpark. Most of the expense goes to Intel directly. Best Buy and its ilk make nary anything on raw CPU sales. This is why they encourage the sale of pre-packaged, value-added systems instead (or charge labor for building a computer from components). Even that offers very little margin.

[BigEd]

Just a thought: there's a psychological difference between a discount and a premium. If you got $1 off the 6502 if you didn't use the BCD arithmetic, you might think it a nice discount. But if the base price were $1 lower, and you were charged an extra $1 to use BCD, you might be offended.

[kc5tja]

Another misinterpretation of my view.

If WDC offered a variant of the 6502 sans BCD mode for $1 less, then that's a final deal. That's an intrinsic aspect of the silicon you purchase.

If WDC, however, offered me the option to use BCD mode via a convenient software download for a premium, then that's a very different matter. In this case, I have the functioning hardware. The hardware has been tested, but without that extra premium, I'm forbidden from using it.

A closer analogy would be a refrigerator with a bundled freezer, like most sold today. But, without paying Frigidaire that extra $50, you can't use the freezer. Only with the software patch will the freezer get cold enough to make ice cubes.

[BigEd]

Exactly: the difference is psychological. You very much want to be the definitive owner of all the possible physical capability that you've got in your hand - and I'm the same, most maker types will be. If there's a technical limitation, we might be motivated to get around it.

But, in your example, the parallel would be where most people are happy to get the fridge for the better price - they don't need the freezer. (I'd be happy to pay less for a car without the use of automatic transmission, cruise control, A/C and electric windows(*))

Question is, is the product on offer a CPU, no strings attached, or a CPU with a user license. (And do you realise which is the case at the time of buying.) Which is to say, the real problem is if there's some deception going on, or a suspicion of it.

The question for the seller is: will there be a backlash? Will it be big enough to put them off?

Cheers
Ed

(*) In fact I do! Granted, none of my cars ever had those features, but I still don't want to pay for them.

[jt_eaton]

When I was provisioning engineer at CariNet, our prices for CPUs weren't substantially less than what you could purchase at other retailers.
-------------------------------

You not buying hardware, you are buying a card that costs maybe a $1 to make and put on the shelf. If sold then Intel picks up the support cost for the upgrade program and the warranty cost if the chip doesn't work. All the rest is profit.

Retailers make very little moving hardware and survive on selling cables and protection plans. This is the type of product that you would expect the sales guy to push once you decide to buy. Those always mean big profit for the store and that is why they push them so hard. Intel is probably giving most of the profit to Best Buy. It is a plum to show the Best Buy execs why they are better off selling the I brand instead of the A brand.

This has nothing to do with whats best for the customer. It's all about marketing.

John Eaton

[GARTHWILSON]

Quote:

you are buying a card that costs maybe a $1 to make and put on the shelf.

If you know of a way to make a card for a dollar, I'd like to know about it. I really don't think you can even make the bare board for that, regardless of volume, and that's before putting any components on it, packaging it, advertising, shipping, etc.. Our company's volumes are not nearly up where those of cell pones and PCs are, but I have countless times been involved in the process of laying out, quoting, and getting them made.

[jt_eaton]

GARTHWILSON wrote:

Quote:

you are buying a card that costs maybe a $1 to make and put on the shelf.

If you know of a way to make a card for a dollar, I'd like to know about it. I really don't think you can even make the bare board for that, regardless of volume, and that's before putting any components on it, packaging it, advertising, shipping, etc.. Our company's volumes are not nearly up where those of cell pones and PCs are, but I have countless times been involved in the process of laying out, quoting, and getting them made.

-------------------------

The "Card" is like a gift card. They are that cheap. You buy the card at best buy and it is activated at the cash register. You run a program on your win7 pc and type is in secret code from the back. The program checks with Intel to see if it is valid and if so then you get the upgrade. It will probably send your cpu's serial number and will then only work in the future on that cpu.

Let me recast this discussion in another light. It is common to now build chips with repairable rams. You oversize the ram array and if testing shows some errors you then have spares that you can overlay when your chip powers up.

Suppose I sorted out the chips with zero errors and had a switch that I could set that would use all the array and give you more ram? This part would have more value and I could sell it as a different sku for more money. Is that wrong?

Now suppose I created a another sku for zero error parts but with the switch turned off and the original size ram. It is identical to the original part and I sell it at the same price as the original. Is that wrong?

Intel is selling a program that lets you buy the cheaper part and throw the switch in the field. You need to compare price of the upgrade vs simply buying the full part in the first place but if you are not certain that you will need it then it might be a better deal. If you are certain then you are probably better off buying the full part.

It's like the mini-bar in your hotel room. Yes you did pay for the room but pull something from that fridge and you will pay some more.

John Eaton

[ElEctric_EyE]

This sounds very similar to a certain video card company disabling their chips for their cheaper boards. There is no speed "binning" of IC's. Their masterpiece video IC is 100% pass or fail (prob just a killer FPGA). If the IC gets put into a cheap video board, than certain features are disabled, i.e. cooling measures (fans, extravagant heatsinks) are done away with to lessen costs further for the low end board.

Still, that being said, Intel is nickel and diming their customers even further from their "locked multipliers" and I wouldn't buy one of their newest i7 cores. (I still wouldn't buy an AMD either). Something smells rotten. Maybe they're trying to lock a CPU# with an IP address.

Sidenote: You guys know Intel purchased MCAfee antivirus recently?
Rumors of them putting anti virus software technology on-die in future processors I've heard...

[BigDumbDinosaur]

kc5tja wrote:

It looks like I'll need to work harder and more diligently on completing my Kestrel-2 design. Just one more reason to abandon the x86 platform like a hot potato. It might run at a snail's pace of 12MHz to 20MHz, but doggone it, it'll be mine, and it'll operate as intended with no surprises. This makes me frothingly livid.

Intel wants to charge $50 to unlock stuff your CPU can already do

So don't buy Intel's stuff. Around here, we haven't built anything powered by an Intel MPU since 2003, which is when dual AMD Athlon hardware became available. All of our servers and workstations are AMD powered, many with four and eight way processing..