Intel Ivy Bridge (22nm) News & Rumours Thread

[FlanK3r]

well 32nm to 22nm, i believe improvements should be atleast 20% better..

joke? CPUs are not GPUs...Maybe in some situation can be 20% better, but not average. Clocks to clocks maybe 5%. Others will be about higher CPU stock clocks, implementation some instruction=specific improvement, not much software support AVX now, new DX11 HD graphics etc.

[Ket]

I think people are getting a little turned around with that 20% figure. The rumor I've heard is that the IGP in Ivy will be 20% better than the IGP in Sandy. Not that the CPU part in Ivy will be 20% better than the CPU part in Sandy.

[Johnny87au]

Well know one knew that Sandybridge would be able to clock that high on AIR, i remember everyone was mentioning DMI clocking and crying because no blck clocking.. its a whole new die shrink+igp, I dont see why they'd bring out a new architecture with just a new IGP/IMC doesnt make sense..

[Ket]

Die shrink to 32nm alone was enough to start the super-high OC bells ringing. Perhaps its just me, but I do feel intel are missing a beat with only the 2500 and 2600 Sands being offered in "K" flavour. Whats wrong with a 2400k for say around £160?

[FlanK3r]

Z68 chipset and think in Q2 we will see next SB models

[Alex-Ro]

Die shrink to 32nm alone was enough to start the super-high OC bells ringing. Perhaps its just me, but I do feel intel are missing a beat with only the 2500 and 2600 Sands being offered in "K" flavour. Whats wrong with a 2400k for say around £160?

Who would then buy 2500K anymore since 2400K would mean same thing,since multi is unlocked? It would be nice to see a dual-core with Hyper-Threading unlocked,cheap and very good for benches

[Johnny87au]

joke? CPUs are not GPUs...Maybe in some situation can be 20% better, but not average. Clocks to clocks maybe 5%. Others will be about higher CPU stock clocks, implementation some instruction=specific improvement, not much software support AVX now, new DX11 HD graphics etc.

20% all round including core 4 core, IGP/IMC/AVX

[Ket]

Who would then buy 2500K anymore since 2400K would mean same thing,since multi is unlocked?

Its called speed binning

[Alex-Ro]

Yea right Ket =))

[Ket]

Speed binning is very accurate now. Hell, if intel couldn't be arsed to speed bin these theoretical 2400k chips they could simply bring out a 2400k with a max multi of 45x. Its not rocket science, intel test every CPU they make. Weaker cores would make a 2400k, stronger cores 2500k, and the strongest cores 2600k.

[zalbard]

Speed binning is very accurate now. Hell, if intel couldn't be arsed to speed bin these theoretical 2400k chips they could simply bring out a 2400k with a max multi of 45x. Its not rocket science, intel test every CPU they make. Weaker cores would make a 2400k, stronger cores 2500k, and the strongest cores 2600k.

Screw 2400K in this case, then. I personally think even 58x limit is quite ridiculous, if it's an unlocked part, unlock it properly...
And it's not like they make many chips that don't reach 2500K's clocks. So it wouldn't improve yields one bit.

[Olivon]

Screw 2400K in this case, then. I personally think even 58x limit is quite ridiculous, if it's an unlocked part, unlock it properly...

5300 MHz really starts to heat in my case, past 54X it's overkill for daily use.
The problem is more about Bclk and lottery chip IMHO.

[glen]

Do we think IvyBridge CPU's will have graphics that compete at all with modern gaming GPUs? That would be about the only reason I could see waiting to buy a Z68 board.

[informal]

Do we think IvyBridge CPU's will have graphics that compete at all with modern gaming GPUs? That would be about the only reason I could see waiting to buy a Z68 board.

According to leaked info IB's GPU should be 60% faster than HD3000 in SB. That's still a lot slower than 5550,a level of performance that will be in Fusion A series (Llano).If anything IB will be a step in right direction,with DX11 support and still solid perf. jump over HD3000.But it won't be close to modern "gaming" GPUs.

[Ket]

Screw 2400K in this case, then. I personally think even 58x limit is quite ridiculous, if it's an unlocked part, unlock it properly...
And it's not like they make many chips that don't reach 2500K's clocks. So it wouldn't improve yields one bit.

The only people that would say that are those who get SB chips by the trayful from sponsors to blow up at will with all kinds of crazy voltages. For others, 63x ("limit" on my chip) you don't have a hope in hell of reaching. A 2400K actually makes a lot of scence. Not many people want to drop £200-£300 on a CPU, but most you will find are fairly open to spending around the £160 mark because its not psycologically as big.

[LiquidReactor]

Come on Intel...give me 8 physical cores and 5 Ghz!

I need to run Crysis at 200 FPS

[zalbard]

The only people that would say that are those who get SB chips by the trayful from sponsors to blow up at will with all kinds of crazy voltages.

There are other ways of cooling but air and water, you know.
A chip that maxes out on air frankly sucks for Xtreme crowd.
And not everyone who is using sub-ambient cooling has sponsors... That's a huge generalisation and a mistake.

For others, 63x ("limit" on my chip) you don't have a hope in hell of reaching.

It's the first time I hear about 63x. Source?
The last I heard was 57x, although people are trying to crawl over that.

[Alex-Ro]

Limit is 60x,ket is obviously having an martian-manufactured SB.

[Olivon]

There are other ways of cooling but air and water, you know.
A chip that maxes out on air frankly sucks for Xtreme crowd.
And not everyone who is using sub-ambient cooling has sponsors... That's a huge generalisation and a mistake.

It's the first time I hear about 63x. Source?
The last I heard was 57x, although people are trying to crawl over that.

http://valid.canardpc.com/show_oc.php?id=1639917

[flippin_waffles]

Do we think IvyBridge CPU's will have graphics that compete at all with modern gaming GPUs? That would be about the only reason I could see waiting to buy a Z68 board.

If inetl keeps rushing out products with terrible testing like sb chipsets, it'll just be more fail....

[sin0822]

the BLCk thing sucks for RAM.

[clayton]

Come on Intel...give me 8 physical cores and 5 Ghz!

I need to run Crysis at 200 FPS

That would require more GPUs rather than more CPU cores.

[Ket]

There are other ways of cooling but air and water, you know.
A chip that maxes out on air frankly sucks for Xtreme crowd.
And not everyone who is using sub-ambient cooling has sponsors... That's a huge generalisation and a mistake.

It's the first time I hear about 63x. Source?
The last I heard was 57x, although people are trying to crawl over that.

1. Not really. Apart from those with sponsors the only others would be people with quite deep pockets.

2. Source is my UEFI. Since updating from 1.43A I don't have 63x multi anymore Its gone back to 60x with official 1.5. I'm not gonna mess around flashing the UEFI just for a screenshot, so heres the 60x from 1.5.

[mmorpgfocus]

any engineer can share at what nm size we start to ruin Moore's law based on the constraints of basic physics?

I mean how many nanometers across is say a three molecule thick piece of silicon?

How about a twenty molecule thick strand of silicon - how many nanometers thick is that?

You would expect as the conductive strands became incredibly thin that errors and power leakage would increase...

How LOW can Intel go before a new model is required for progress?

[mmorpgfocus]

I will answer my own question if anyone else is curious and did not already know this.

1 nanometer = 10 atoms.

So, obviously Intel is making transistors 220 atoms thick at 22nm - that is an incredible accomplishment when you think that they are producing a recordable on/off state for the processor to record at a size so incredibly small - I mean only 220 ATOMS across!

One problem though is apparently about to hit Intel and all computing enthusiasts in the face and it is not readily avoidable - it is called the laws of quantum physics.

Unless quantum physics collapses - there will NEVER be 1nm silicon transistors.

The reason is that at 1nm the on/off state can no longer be controlled with a flow of electrons (electricity).

I don't understand quantum physics - (few humans do or ever have) but apparently under the laws of things the size of 1nm, there is NO on or off for a transistor - electrons jump around, and dissapear and reappear all over the place and the very act of trying to observe whether the transistor is in an on or off state necessary to record for processing becomes an act that can change the state! Computing breaks down.

I don't know what the limit will be - but some suggest that unless we can start to reach for some new physics - transistors CANNOT get smaller than around 6nm.

So there are only a few shrinkages left before efficiency/architecture or operating frequency improvements will be all we have.

I feel eventually we will reach a scientific breakthrough that will allow progress to continue - but we may all be facing a 2020 where computer progress rapidly slows down for YEARS!