AMD unleashes first ever commercial “5GHz” CPU, the FX-9590

[Vesku]

There's a lot of workarounds if heatsink pressure is a problem. A bracket around the die resting on the PCB seems like a good idea.

Yes, that's one of the things which screams that the change revolves around cost. Adding a bracket would defeat much if not all of the cost savings.

 

[Idontcare]

This is one of the worst car analogies that I have read.

And you'd know because you have a degree in engineering like I do?

Or is this just more of your usual pure BS that you routinely post?

We know Intel would not go to bankruptcy with a better design.

Intel is not a non-profit organization like AMD and GloFo.

If you have a problem with capitalism and companies that are in business to make profit year-over-year then you should just say as much and stop making silly irrelevant comments like the one you are making above.

 

[Vesku]

I feel IB-E may put the whole "debate" (really people think it's a technical issue and not cost based?) to rest. Pretty sure with their premium line they'll use solder or if adhesive and TIM they will have another solution to gap distance (bracket or what not).

Wonder if Kabini uses TIM, no expectations from enthusiasts so it would be an easy cost reduction.

 

[Idontcare]

I feel IB-E may put the whole "debate" (really people think it's a technical issue and not cost based?) to rest. Pretty sure with their premium line they'll use solder or if adhesive and TIM they will have another solution to gap distance (bracket or what not).

Wonder if Kabini uses TIM, no expectations from enthusiasts so it would be an easy cost reduction.

There is no question it is a cost-driven solution.

Name one thing about either Intel's or AMD's products that isn't?

From the nodes themselves (Finfet vs SOI) to the product design (planar vs 3D) to the packaging (MCM vs single-die vs TSV) to the thermal solution (air-cooled, water cooled, passive cooled)...it is all about designing a product to match the technical issues of reality with the cost-sensitive priorities of a business that exists as a business for the express purpose of making money for its shareholders.

Technical issues drive the cost-sensitive solutions. You don't get one without the other.

 

[Vesku]

Yet some people persist in the "they had to use TIM because of technical reasons". P.T. Barnum at work I suppose.

 

[Abwx]

And you'd know because you have a degree in engineering like I do?

Or is this just more of your usual pure BS that you routinely post?

I wouldnt expect someone with a degree to always
use car analogies , that is classical physics mechanics ,
that is gravitation ruled phenomenons , when discussing
matters that are either electromagnetic force ruled or at
the extreme when the electromgnetical phenomenon produce
measurable macroscopic mechanic effects.

No pun intended but for me car analogies are indeed Bs....

 

[Imouto]

@IDC

Man, you're using car analogies to punish me, right? That comparison is over the top here. You are somehow saying that the gap between the IHS and the die is, again, somehow like the mechanics inside a car engine. A static part compared to moving pieces. Just lol.

I'm not trying to cut Intel earnings here, I'm just saying that having a temps problem for 2 damn years is unheard of in a powerhouse like Intel and much more when yourself identified the problem. The bracket I mentioned earlier or shrinking the IHS to make it sturdier to heat deformations are something that came to mind just thinking about it for a little while. Intel giving the worst/average/best possible solution to whatever the problem they have is something that neither you or I know. But us peasants can't understand the change.

And the HSF bundled with Intel (or AMD for that matter) is just a way to raise the price of the product. You should ask yourself why the stock cooler hasn't changed a single bit for years. Luckily I've been able to buy CPUs without stock HSF for years saving in something that most of the ppl I know store forever and have a huge collection already.

Yes, that's one of the things which screams that the change revolves around cost. Adding a bracket would defeat much if not all of the cost savings.

1 cent worth of rubber would make wonders and could even contain the TIM if Intel hates solder that much.

 

[sushiwarrior]

Imouto, I don't know your background so I have no idea if you are aware of what goes into the engineering trade-offs that a company might make in the pursuit of developing a product that performs consistently.

Consistency is key ....

Can you explain why they never had any issues previously? Like, this problem didn't exist before. It evidently wasn't a problem for what, 5+ years? But then Intel starts making consumer CPU's locked... And then ruins temperatures for overclockers... It's not like they aren't capable of doing it, and it's not like it's a problem for a company which manufactures the most complex devices on the planet to fix a tiny gap. There's more at work here than cost savings, it's about upselling.

 

[Vesku]

1 cent worth of rubber would make wonders and could even contain the TIM if Intel hates solder that much.

1 cent? It's a new step in packaging, your joke estimate isn't even amusing. If they were magicians like that the adhesive wouldn't have tolerance issues and the OC issue would have never occurred.

 

[Abwx]

IB and HW have higher thermal density than SB ,
hence dilatation of the die and of the heat absorbing
surface at the contacting area is consequently higher ,
while the temperature gradient is steeper , increasing
the mechanical stress of the cover and heatsink pressures.
As a consequence there must be an elastic gap between
the die and the copper cover of the SKU to ensure
adequate reliability by maintening a constant pressure
with varying temperatures.

 

[notty22]

Can you explain why they never had any issues previously? Like, this problem didn't exist before. It evidently wasn't a problem for what, 5+ years? But then Intel starts making consumer CPU's locked... And then ruins temperatures for overclockers... It's not like they aren't capable of doing it, and it's not like it's a problem for a company which manufactures the most complex devices on the planet to fix a tiny gap. There's more at work here than cost savings, it's about upselling.

You do know that Intel went to paste with the switch from 32nm to 22nm?

The car analogy was used as example for people to understand.

Experience in a number of production situations. (soda/beer cans/newspapers/electronics -die bonding, assembly to solder). A area of production, especially near the end like the installation of a heat spreader(on working dies) has little room for failures. Is that so hard to grasp?

We know at the switch from 32nm to 22nm Intel went to paste. It's possible that solder caused flex/expansion issues on the new die. Creating waste. There is no heat problem, with normal use, paste or solder.

The paranoia witch hunting is amusing. Sad at the same time.

GO BRUINS- Stanley Cup!

 

[Imouto]

1 cent? It's a new step in packaging, your joke estimate isn't even amusing. If they were magicians like that the adhesive wouldn't have tolerance issues and the OC issue would have never occurred.

I said that before but that's like bundling an useless HSF with every single CPU. If it isn't 1 cent (which was an exaggeration) it's pretty close. Charge the end user 10 times for the added cost and everyone will be happy.

 

[AnandThenMan]

The engine in your car would be silly expensive if engineered to the tolerances of the gap between the IHS and CPU.

And that tolerance is?

 

[Enigmoid]

@IDC

Man, you're using car analogies to punish me, right? That comparison is over the top here. You are somehow saying that the gap between the IHS and the die is, again, somehow like the mechanics inside a car engine. A static part compared to moving pieces. Just lol.

He's using the comparison because its easy to understand for those without a engineering or physics background.

Basically the point is that quality control vs cost is not linear. To double the tolerance between parts (say how well part A fits in the car to part B--ex I bought a cheap table recently and I had to assemble it. I found out that the parts were not within spec and had to bend/modify certain parts to get the screws in a couple sections in. Obviously better manufacturing quality is needed for that table but given its cost....It was a pain in the *** but at its price point was worth it IMO) might cost engineering wise more than double the cost.

For the exact tolerances needed for industrial applications (say aircraft engines) costs are significantly higher than for consumer applications where a 'good enough' standard can be used. If your car engine was subjected to the same testing, validation, checkups, maintenance, etc, as an aircraft engine you probably couldn't afford to own it. But then, being a car engine it doesn't need the tolerance and performance levels of an aircraft engine and since doing so would result in an noncompetitive product it isn't done (even though the engine would perform a little better and get better mileage, etc if it got the same treatment).

Same for intel's CPUs. They don't want the absolute best performing part. They want the most competitive part (20% gain for twice the manufacturing costs will never be taken over 10% gain for 10% increase in manufacturing costs in the consumer market--ie if intel threw away all its chips that don't meet the mobile power/voltage/performance requirements and used only those high binned chips for its desktop line as well as mobile).

 

[BallaTheFeared]

My Haswell chip does not seem to suffer from too much gap.


*Just sayin'

 

[Vesku]

My Haswell chip does not seem to suffer from too much gap.


*Just sayin'

Now all you need is to own a hotter second Haswell chip and you will have learned the word "variance".

 

[blackened23]

With regard to the gap/IHS issue, I think the key point here is that intel does not arbitrarily change things, there are significant engineering-related discussions with regard to trade-offs that must be made to ensure quality across a wide variety of products. I've seen some suggest a conspiracy at intel to intentionally give consumers CPUs that may overclock worse, but I'm certain this isn't the case. Aside from this, a lot of users *are* getting excellent overclocks despite the "issue", I do believe that sandy bridge may have given some users a case of rose-tinted glasses. Chips that overclock that well with minimal trade-offs come once in a blue moon. Additionally, it's not like this is an AMD FX situation - intel still increased the IPC, therefore if you *do* get a slightly lower overclock in comparison to a sandy bridge, the IPC difference will more than offset that. I've seen benchmarks of a 4.3ghz 4770k exceeding that of a 5ghz 2600k.

Anyway, while we'll never know the specifics, you can be sure that the engineers at intel had a lot of discussion, analysis, and cost-benefit studies to determine the best solution for the greater user base. And that greater user base doesn't necessarily want maximum overclocks, and intel doesn't necessarily cater to that. They could, but what would the opportunity cost be? Would it be at the expense of quality? Again, we'll never know for sure but i'm fairly certain there is no conspiracy.

 

[Imouto]

Now all you need is to own a hotter second Haswell chip and you will have learned the word "variance".

The endless story about "How awesome my 470s are" just got a spinoff. Now we will have to read about his golden Haswell for 2 years. Ain't that cool.

 

[Idontcare]

Can you explain why they never had any issues previously? Like, this problem didn't exist before. It evidently wasn't a problem for what, 5+ years? But then Intel starts making consumer CPU's locked... And then ruins temperatures for overclockers... It's not like they aren't capable of doing it, and it's not like it's a problem for a company which manufactures the most complex devices on the planet to fix a tiny gap. There's more at work here than cost savings, it's about upselling.

Engineering channel stress for a 22nm finfet is nothing like that of a 32nm or larger planar CMOS xtor.

I get that people really truly want to believe there is some conspiracy afoot here, but just remember the only difference between you and the dudes who decried witchcraft at the Salem witch trials is a 21st century education that tells you witches don't exist...and if you are lacking one in the physics and engineering of process technology then you should be wary of your own prejudices before jumping to the conclusion that there must be witches under the IHS of a 22nm CPU.

 

[bradley]

Hector Ruiz earned a B.S., M.S. and Ph.D. in electrical engineering. Maybe we should ask his opinion on Haswell and whether it's a worthy upgrade. Personally, I think Intel pulled a Windows 8 on us with their subpar effort.

 

[SOFTengCOMPelec]

some conspiracy afoot here

We must be able to work out which member is the conspirator.
It must be someone who is continually defending Intels use of paste rather than Solder, TIM.
Intel
Don't (seem to)
Care = I-dont-care = Idontcare
maybe there is a member of that name ? (...joke)

 

[Ayah]

Engineering channel stress for a 22nm finfet is nothing like that of a 32nm or larger planar CMOS xtor.

I get that people really truly want to believe there is some conspiracy afoot here, but just remember the only difference between you and the dudes who decried witchcraft at the Salem witch trials is a 21st century education that tells you witches don't exist...and if you are lacking one in the physics and engineering of process technology then you should be wary of your own prejudices before jumping to the conclusion that there must be witches under the IHS of a 22nm CPU.

They're also forgetting that not only is it a finfet-type but intel also used multiple sources and drains with their trigate which adds an extra mechanical change.

 

[AnandThenMan]

He's using the comparison because its easy to understand for those without a engineering or physics background.

Why would anyone with an engineering background (or even someone with common sense) compare tolerances of a static, never moving part to components that must maintain a spec even after trillions of mechanical cycles?

And keep in mind we have no idea what kind of spec/tolerance we are talking about, just a claim of one.

 

[Lepton87]

IDC you act like Intel's engineers were infallible. They are a bunch of a very smart dudes but nobody is unerring. I'm sure they could deal with that thermal problem if they wanted to. 22nm xtors don't necessarily need to be treated like eggs, otherwise we wouldn't have bare die cooled notebook chips, nothing changed there.

 

[Idontcare]

Why would anyone with an engineering background (or even someone with common sense) compare tolerances of a static, never moving part to components that must maintain a spec even after trillions of mechanical cycles?

And keep in mind we have no idea what kind of spec/tolerance we are talking about, just a claim of one.

Ponder the reality of how the IHS just so happens to find itself mated to the CPU's PCB prior to its arrival at your nearest Newegg warehouse.

There is a rather mechanical process involved in the creation of the finalized product, and the mechanical process entails engineering tolerances for clearances, pressures, etc.

If you think nothing in your CPU moves then you are being silly.

Direct-current (DC) flow is the movement of electrons. Degredation in your CPU involves the relocation of atoms (diffusion), physical movement and relocation of copper atoms, hydrogen, dopants, etc.

Not to mention any change in temperature involves movement of the materials that are changing temperature (hence the phrase coefficient of thermal expansion )