Intel Enthusiast CPU Predictions

[TidusZ]

I just DON'T get this mentality. Why is IVB not enough? What the hell are you doing?

http://www.uberent.com/pa/

ivy bridge at 4.7 ghz is not enough



Also, yes Intel has no competition and they are trying to squeeze every last cent out of everyone while offering last year's performance every year. Hopefully as long as some smart people still have hammers, knives, and conductive pencils we can still get decent computing for a reasonable price in the future, since with the new Intel its going to be more necessary than ever. Personally I do need more computing power but Intel hasnt released anything faster than a 2500K yet so obviously there is no option anyway.

 

[crashtech]

See the problem?

Vs.

OT, but I always liked the LGA771 HSF mounting system, which is direct to the tray with large through holes on the motherboard. Mounting a heavy HSF to a flexible PCB seems like a kludge to me.

 

[aigomorla]

I think you forgot how many thunderbird cores got chipped or crushed. RMA rates would go from sub 0.01% today to what 2-3% if not more?

yeah yet all the haswell prepacks like the intel NUC will feature the cpu soldered and IHS removed.

And shintai... u forget one major thing one typically needs to remove after delidding...

The arm clamp must be removed if u delid.
However if the cpu was made, the pcb plastic would need to be thicker so the arm could lock on.

The arm try could act as a shim in some cases, if the cpu tray was properly calibrated.

 

[BallaTheFeared]

yeah yet all the haswell prepacks like the intel NUC will feature the cpu soldered and IHS removed.

And shintai... u forget one major thing one typically needs to remove after delidding...

The arm clamp must be removed if u delid.
However if the cpu was made, the pcb plastic would need to be thicker so the arm could lock on.

The arm try could act as a shim in some cases, if the cpu tray was properly calibrated.

Only if you go bare die, which IDC showed wasn't really worth the risk.

 

[aigomorla]

Only if you go bare die, which IDC showed wasn't really worth the risk.

i said delidding..

And IDC showed a 20C reduction on a H#O.
If he had my LC gear.. i bet you that reduction would probably be even greater.

And if needed, i can always bust out my Storm G5 which was originally designed to be used on a bare die with high pressure injectors.

 

[crashtech]

One of the big reasons to have an IHS these days is in fact the method of retention and the amount of force required to ensure contact of over a thousand connections. The first time I bent the arm down to retain an 1155 CPU I swore something was wrong, and double checked everything to be sure. It's a lot of force, perceptibly more than with 775. Enough to crush some PCBs, I'd think.

 

[aigomorla]

One of the big reasons to have an IHS these days is in fact the method of retention and the amount of force required to ensure contact of over a thousand connections. The first time I bent the arm down to retain an 1155 CPU I swore something was wrong, and double checked everything to be sure. It's a lot of force, perceptibly more than with 775. Enough to crush some PCBs, I'd think.

no...

The force in which the sink needs to come down on a IHS is serveral magnitudes greater then a bare die.

And those pins which connect to cpu do not need a insane amount of pressure.

IHS require pressure tho, about ~ 15lbs (Skinnylabs data) of force to get a nice tight clamp on the cpu and IHS for bowing to work effectively.

This is why laptop coolers have tiny springs holding the pressure down on bare dies.

 

[crashtech]

no...

The force in which the sink needs to come down on a IHS is serveral magnitudes greater then a bare die.

And those pins which connect to cpu do not need a insane amount of pressure.

IHS require pressure tho, about ~ 15lbs (Skinnylabs data) of force to get a nice tight clamp on the cpu and IHS for bowing to work effectively.

This is why laptop coolers have tiny springs holding the pressure down on bare dies.

Several orders of magnitude? Surely this is hyperbole. Do you have any links or independent research results to back these claims?

 

[BallaTheFeared]

i said delidding..

And IDC showed a 20C reduction on a H#O.
If he had my LC gear.. i bet you that reduction would probably be even greater.

And if needed, i can always bust out my Storm G5 which was originally designed to be used on a bare die with high pressure injectors.

He also showed direct die mount made little difference over delidding, replacing the tim, cutting off the glue, and placing the IHS back on.

Which I think is the best thing to do, once delidded, reliddit.

 

[crashtech]

I have wondered whether relidding with a flexible adhesive such as clear silicone in place of the stock glue would hold the IHS in place while locking down, yet be conformable enough to provide good die to IHS contact when the HSF is installed.

 

[Sheep221]

The IHSes and sockets along with retention brackets get soon abolished from mainstream desktops, as Intel plans to start sell CPUs exclusively in BGA package.

The reason why socket A CPUs were easy to break is because there was no support under the PCB as the socket was just around the perimeter of the CPU rather than its entire area.

The next thing is, I'm not sure why Intel transferred to LGA system, but I guess it had to do something with heat dissipation and energy transfer, AMD still uses regular PGA chips and not sure if BD and PD CPUs would perform better given the situation they would be LGA based. Someone can explain more.

 

[crashtech]

The IHSes and sockets along with retention brackets get soon abolished from mainstream desktops, as Intel plans to start sell CPUs exclusively in BGA package.

I don't believe this. Link please.

The reason why socket A CPUs were easy to break is because there was no support under the PCB as the socket was just around the perimeter of the CPU rather than its entire area.

The next thing is, I'm not sure why Intel transferred to LGA system, but I guess it had to do something with heat dissipation and energy transfer, AMD still uses regular PGA chips and not sure if BD and PD CPUs would perform better given the situation they would be LGA based. Someone can explain more.

PGAs make CPUs more fragile. BGA transfers that fragility to the mobo, which, in spite of being larger, is most often the cheaper part of the two.

 

[Sheep221]

I don't believe this. Link please.


PGAs make CPUs more fragile. BGA transfers that fragility to the mobo, which, in spite of being larger, is most often the cheaper part of the two.

You didn't hear about that yet?
Links are here, here, here and here

 

[piasabird]

So Intel is basially purposely crippling their mainstream processors just to get enthusiasts to buy more expensive processors. This could cause a custom shop to spring up that specializes in delidding Intel Processors and then repackaging them with a better thermal paste. I wonder what Intel would think of that?

Instead of cleaning the package maybe just take a new package from whoever makes it for Intel. That could speed up the process by not having to clean the lid.

Makes you wonder if the CPU Cooler companies are in cahoots with Intel?

 

[gun5l1ng3r]

You didn't hear about that yet?
Links are here, here, here and here

Those articles are quite old by now. I thought I remember hearing that Intel said this rumor was false...

 

[crashtech]

You didn't hear about that yet?
Links are here, here, here and here

Sorry, perhaps I should have stipulated recent articles with the same definite language you used, and not "reports suggest," "rumor has it," and "may be."

Aside from that, there will be socketed server parts as far as the eye can see. So even IF Intel kills socketed mainstream, there will be socketed server and no doubt server-derived socketed enthusiast parts.

 

[ShintaiDK]

yeah yet all the haswell prepacks like the intel NUC will feature the cpu soldered and IHS removed.

And shintai... u forget one major thing one typically needs to remove after delidding...

The arm clamp must be removed if u delid.
However if the cpu was made, the pcb plastic would need to be thicker so the arm could lock on.

The arm try could act as a shim in some cases, if the cpu tray was properly calibrated.

Mobile chips without IHS tend to be installed by OEMs and use small coolers. Not average joe with or without an oversized cooler.

 

[ShintaiDK]

You didn't hear about that yet?
Links are here, here, here and here

I think you missed the latest desktop roadmap with Skylake using LGA.

But that doesnt sell many clicks.

 

[aigomorla]

Several orders of magnitude? Surely this is hyperbole. Do you have any links or independent research results to back these claims?

LOL...

i said it was a test done by Vapor / Eric from skinnie labs.
http://skinneelabs.com/

Its somewhere on that site from a test done by Vapor on XS.

Mobile chips without IHS tend to be installed by OEMs and use small coolers. Not average joe with or without an oversized cooler.

see thats the thing.. if ur using the stock heat sink, its most definitely capable.
And it wouldnt be oversized, and without the IHS it would be more efficient.

Sorry, perhaps I should have stipulated recent articles with the same definite language you used, and not "reports suggest," "rumor has it," and "may be."

Aside from that, there will be socketed server parts as far as the eye can see. So even IF Intel kills socketed mainstream, there will be socketed server and no doubt server-derived socketed enthusiast parts.

This i agree with.
Until the days the OEM market becomes not profitable.

Even then for pure supply issues, socketed items win overall.
Less excess is created by compartmentalization vs complete package form.

However i dont see mainstream getting all those benifits anymore.
Intel is basically telling the ground level user, get ready for MAC type packages, because Apple does such a good job were gonna copy them.

Leaving us enthusiast out with the enterprise people, like how now import race cars are controlled by the big boys and use specialized racing parts.

 

[ShintaiDK]

see thats the thing.. if ur using the stock heat sink, its most definitely capable.
And it wouldnt be oversized, and without the IHS it would be more efficient.

But you are again leaving it to average Joe, rather than some trained/skilled worker or robot. Both AMD and Intel decided from experience that average Joe wasnt fit for the task.

 

[ElFenix]

Sockets aren't going anywhere. Computer companies want to be able to offer a selection of processors on their computer lines and it makes way more sense to be able stock one board for the product line and then integrate the cpu rather than stock a bunch of soldered on cpu/board combos. With sockets you can keep some overlap in product lines, makes demand forecasting simpler, and you don't have a bunch of boards to ship back when intel does its mid-cycle clockspeed bump. In short, intel will continue to offer socketed processors because that's what its major customers want.

 

[crashtech]

LOL...

i said it was a test done by Vapor / Eric from skinnie labs.
http://skinneelabs.com/

Its somewhere on that site from a test done by Vapor on XS.

I'll check it out. This may seem nitpicky to some, but when you say several orders of magnitude less than 15 lbs, if several = three, that indicates you believe all the BGA pins can be held in reliable contact with less than 7 grams of pressure.

 

[aigomorla]

I'll check it out. This may seem nitpicky to some, but when you say several orders of magnitude less than 15 lbs, if several = three, that indicates you believe all the BGA pins can be held in reliable contact with less than 7 grams of pressure.

actually u need less then 1lb of pressure on the naked die.

When u have TIM / Surface contact your only aiming to seal microbubbles / tiny tiny airgaps which are blemishes on the surface of the IHS.
Everyone knows metal -> heat source contact direct is always the best.
Adding TIM only gets rid of the air gaps.
Adding pressure causes the air gaps to push out and get more surface area on the sink + IHS.
IHS is far from being a perfect surface.

The Naked Die on the other had, has much smaller gaps which tim needs to fill.
Which also in turn leads to a better contact on 2 mirror like objects.

Hence less pressure is required to fill those micro air gaps.

But you are again leaving it to average Joe, rather than some trained/skilled worker or robot. Both AMD and Intel decided from experience that average Joe wasnt fit for the task.

I can count many many times where i always pull off the stock heat sinks on everything... cake off that old TIM with a razor blade, and retim the board before install.



I dont think the average Joe cracks naked Dies.
Its the average Joe Jr. who has no concept of proper research, patience to be working on a PC, that ends up cranking the heat sink down, or using a improper heat sink, who cracks dies.

 

[Ajay]

wouldn't something like this (used in many GPUs) be cheaper than covering the entire thing?

That's exactly what I used, but with a flat rectangular bottomed Thermalright cooler. Worked perfectly. Always used a shim after I dropped a HS on a bare die and my nads tried to ascend back to their pre-pubescent resting place.

 

[crashtech]

actually u need less then 1lb of pressure

yes, but your initial claim can be interpreted as <7 grams. One pound is >450 grams. So we still disagree by well over one order of magnitude.

I'm sure there are published specs that could lay this to rest pretty quickly. I'm just too lazy to wade through them.

And I'm talking about the pressure required to ensure electrical contact only right now.