Expected Ivy Bridge performance

[Puppies04]

Well I have hit a thermal ceiling on this rig (and from reading these forums I think a lot of other people are in the same boat), I don't want to add any more to my max temps so I can't go higher than 4.5ghz stabily. If IVB requires less voltage (and it will) to achieve a similar overclock then I would happily run it faster.

A 10% increase on my 2500k would be 4.95ghz and I am fully expecting the IB 2500k equivalent to hit 4.6-4.8 at the same kind of max temps coupled with the 5%ish IPC inprovement that would equate to around a 10% overall improvement.

 

[Nemesis 1]

Prescott (Pentium 4) has a 31-stage pipeline while Sandy Bridge/Ivy Bridge has a 14-stage pipeline. The number of pipeline stages in Intel CPUs hasn't changed for around 5 years now, when Conroe was introduced.

Well your as wrong about this as you were BD and thats a fact.

Also your quote above who was talking a P4p Not I not ever piece of junk intel produced. I clearly said P4C that the 18 stage pipeline . P4P indeeed . Pure crap equaled only by PH1 and BD.

 

[LOL_Wut_Axel]

Well your as wrong about this as you were BD and thats a fact.

Also your quote above who was talking a P4p Not I not ever piece of junk intel produced. I clearly said P4C that the 18 stage pipeline . P4P indeeed . Pure crap equaled only by PH1 and BD.

Um, what? I said Bulldozer would focus on improving clock speeds and if there were gonna be IPC improvements they would be near nothing. And then what happened?

And you can keep yapping on about how I'm wrong, but the highest OC to be seen right now by almost-production ready Ivy Bridge engineering samples is 4.8GHz. Ivy Bridge won't bring about a significant difference in clock speeds. The only thing Intel focused on improving with it is the IGP and performance/watt.

 

[Nemesis 1]

Yes, that is what I'm saying. The difference will be 100-200MHz more headroom than Sandy Bridge and no more.

Prescott (Pentium 4) has a 31-stage pipeline while Sandy Bridge/Ivy Bridge has a 14-stage pipeline. The number of pipeline stages in Intel CPUs hasn't changed for around 5 years now, when Conroe was introduced.

For Intel higher clock speeds is nowhere near as important as lowering the voltage needed to get to a target clock speed.

Were do you get this stuff from . Intels is interested in performance as well as lowering voltage . But not so much as you would like people to believe. Intel Owns server market . Intel owns highend desktop middle desktop and are pushing AMD hard on lowend desktop. . Intel owns Highend mobile . Intel owns middle mobile . Intels medfield is no joke . Intels Knights corner is no joke .

Intel can do what ever intel wants to do . Its AMD that has limited choices. Intel has $$$$ Talent Best fabs in the world . AMD has none of these things except a over priced GPU house . 5 billion dolllars LOL.

 

[Nemesis 1]

quote=LOL_Wut_Axel;32951041]Um, what? I said Bulldozer would focus on improving clock speeds and if there were gonna be IPC improvements they would be near nothing. And then what happened?[/quote]

Well its a good thing all those threads are here than isn't it. Because as I recall you were all over the place on BD . I was the only forum member here at Anands that had BD right and I knew that in Dec of 2010, I even posted it in dec of 2010

 

[Idontcare]

Um, what? I said Bulldozer would focus on improving clock speeds and if there were gonna be IPC improvements they would be near nothing. And then what happened?

Well its a good thing all those threads are here than isn't it. Because as I recall you were all over the place on BD . I was the only forum member here at Anands that had BD right and I knew that in Dec of 2010, I even posted it in dec of 2010

Being right for all the wrong reasons is the uninteresting result because it says nothing of your capability to be right in the future. :\

In Dec 2010 not even AMD's own people knew how bulldozer was going to perform in reality, claiming you somehow "knew" is just a lucky guess masquerading as a fact, don't fool yourself into thinking people are so easily duped or swayed :|

 

[lol123]

Prescott (Pentium 4) has a 31-stage pipeline while Sandy Bridge/Ivy Bridge has a 14-stage pipeline. The number of pipeline stages in Intel CPUs hasn't changed for around 5 years now, when Conroe was introduced.

That's not accurate. Real World Technologies puts the SB pipeline length at 2 stages more than that of Nehalem in case of a micro-op cache miss, which would make the number 18. For reference, the original NetBurst had a 20 stage pipeline.

 

[psolord]

Nope. The Sandy Bridge architecture won't go much higher in terms of clocks, the reason being the short pipeline.

Yes, that is what I'm saying. The difference will be 100-200MHz more headroom than Sandy Bridge and no more.

Prescott (Pentium 4) has a 31-stage pipeline while Sandy Bridge/Ivy Bridge has a 14-stage pipeline. The number of pipeline stages in Intel CPUs hasn't changed for around 5 years now, when Conroe was introduced.

For Intel higher clock speeds is nowhere near as important as lowering the voltage needed to get to a target clock speed.

Thanks for the info.

Still I can't help but wonder, why my Q9550 runs happily at 4Ghz (my i7 860 can do the same as well-also a 45nm part) while a friend's 2500k runs happily at 5Ghz, if they have similar pipelines?

From what I gather, their distinct difference as far as clock is concerned, is their fab process, right?

 

[grkM3]

Well intel raised max multy from 57 to 63

Nuff said

And lol my x79 setup has 40 lanes that can do pcie 2.0 at 16x

Find me a board that will be bottlenecked at 2.0 speeds on it and ill give you a 2600k

Pcie 3.0 is marketing get over it.

 

[TakeNoPrisoners]

read what I said again,you are saying pcie 2.0 has only 8x when you can run it at 16x and its more than enough to run 2 7970s at full speed

1 card 16x 2.0 and the other card 16x 2.0 with another 8 lanes doing nothing on a x79 board with a sandy E cpu

do you honeslty think some one with a x79 setup is going to downgrade there 6 core sandy E to take advantage of an ivy with pcie 3.0 support?

and your 8x to 16x scores are less then 1fps in most games and can easily be throw away do to + or - results from running the benchmarks at different times.

dont forget driver tweaks,whats more crazzy is how well 4x runs,a simple 2-3mhz bclk over clock would get rid of that bottle neck even at 4x

so tech we can run 3 7970s on an x79 setup

16x
16x
4x

and 4 lanes left for usb and sata etc.

I refuse to read anything you say. You have severe trouble putting a sentence together.

Did you even graduate high school?

 

[Puppies04]

Well intel raised max multy from 57 to 63

Nuff said

And lol my x79 setup has 40 lanes that can do pcie 2.0 at 16x

Find me a board that will be bottlenecked at 2.0 speeds on it and ill give you a 2600k

Pcie 3.0 is marketing get over it.

What are you talking about? PCIE 3.0 might be less useful on a setup with 40 lanes but who in their right mind is going to "upgrade" from a 2011 to an IB rig. 1155 (16 lanes) has been shown to bottleneck top end cards on x8 x8 so anyone running SLI/crossfire with the top cards that come out this year will relish using a PCIE 3.0 setup. Likewise anyone running quad SLI with 580's or above will be bottlenecking their cards on 2011 so i'm sure they will be waiting for IBE to drop. Just because you don't need it the way you have your rig set up doesn't mean you can call it a gimmick.

 

[grkM3]

this was taken right from the anandtech review

Ultimately what is clear is that 8GB/sec of bandwidth, either in the form of PCIe 2 x16 or PCIe 3 x8, will be necessary to completely feed the 7970

pcie 2.0 @16x will feed a 7970 all the bandwidth it needs and going sli/crossfire does not double exact need for bandwith,you dont get 2x the performance with 2 cards.Its closer to 1.7 ish going sli/cross fire and even on a 1155 setup 2 7970s would run just fine.

No one is dropping 1200 bucks on gpus and running 1155 boards,they are in the high end segment and are already on socket 2011.

main stream is 1 gpu and pcie 2.0 @ 16 will run a 7970 JUST FINE.

 

[Arzachel]

With the best IB integrated GPU + a slight overclock I'd bet Skyrim could run at low or even medium with ~30fps.

From the looks of it, Ivy Bridge will still trail Llano in GPU performance, so I wouldn't count on it unless Skyrim is even more CPU limited than I thought.

 

[grkM3]

I refuse to read anything you say. You have severe trouble putting a sentence together.

Did you even graduate high school?

Im sorry you cant read what I typed,it was on my cell phone.No I never went to highschool,im only in 8th grade :\

 

[Idontcare]

Nope. The Sandy Bridge architecture won't go much higher in terms of clocks, the reason being the short pipeline.

Pipeline limitations on clockspeed are not a given. It comes down to the underlying process technology (drive current, RC delay) as well as microarchitecture.

Intel has been making marked gains in driver current these past generations, and 3D xtors are only going to improve on that rate of gain because they have much better control over the physical and electrical boundaries that define the channel itself.

If Intel did a straight shrink of Sandy Bridge (not dummy shrink, but a reasonable straight shrink) and dialed in the voltage such that the power consumption was normalized between Sandy Bridge and a hypothetical 22nm Sandy Bridge then I would expect the 22nm version to clock roughly 15-20% higher than the 32nm microarchitectural equivalent based on the pipeline-limited-clockspeed argument you are taking.

 

[Idontcare]

Thanks for the info.

Still I can't help but wonder, why my Q9550 runs happily at 4Ghz (my i7 860 can do the same as well-also a 45nm part) while a friend's 2500k runs happily at 5Ghz, if they have similar pipelines?

From what I gather, their distinct difference as far as clock is concerned, is their fab process, right?

Yep. Pipeline is part of the equation that determines operating clockspeed of the IC, but it is only part of that equation.

 

[grkM3]

What are you talking about? PCIE 3.0 might be less useful on a setup with 40 lanes but who in their right mind is going to "upgrade" from a 2011 to an IB rig. 1155 (16 lanes) has been shown to bottleneck top end cards on x8 x8 so anyone running SLI/crossfire with the top cards that come out this year will relish using a PCIE 3.0 setup. Likewise anyone running quad SLI with 580's or above will be bottlenecking their cards on 2011 so i'm sure they will be waiting for IBE to drop. Just because you don't need it the way you have your rig set up doesn't mean you can call it a gimmick.

That was my whole point,lol was saying pcie 3.0 is a reason to grab a ivy setup.

The people that bought a setup to run proper 2/3 way sli are already on socket 2011 and would NOT downgrade to an ivy to get pcie 3.0 spec.But it looks like some sandy e's are supporting 3.0 spec and intel has still not officially anounced what chips will support it on socket 2011.Im betting its going to be with the launch of the c2 steppings and the vt-x support.

 

[grkM3]

I also think ivy will clock a little better than LOL is guessing,why else would intel raise the multy from 57x to 63x

Someone at intel must of gotten a few chips to hit that max multy right?

 

[Rvenger]

I also think ivy will clock a little better than LOL is guessing,why else would intel raise the multy from 57x to 63x

Someone at intel must of gotten a few chips to hit that max multy right?

As the 22nm process matures, the clockspeeds should go up. Thats why they enabled the max multi because there were some SB chips that did hit the 57x multi, although very rare and probably unstable.

 

[formulav8]

The original P4 and Northwood was a literal double of the P3's 10 stage pipe. So pre-prescott p4's were all 20. Prescott is 'said' to be 31. I'm not sure if Intel flat out said the 'exact' length back when it was released? :\

 

[LOL_Wut_Axel]

Thanks for the info.

Still I can't help but wonder, why my Q9550 runs happily at 4Ghz (my i7 860 can do the same as well-also a 45nm part) while a friend's 2500k runs happily at 5Ghz, if they have similar pipelines?

From what I gather, their distinct difference as far as clock is concerned, is their fab process, right?

Because 5GHz is about the upper limit you can get out of a pipeline with this many stages. That, and obviously to get higher than 4GHz on Penryn requires a lot of voltage, which then causes more degradation and more instability. Looking ahead, the revisions of the Bulldozer architecture will probably clock higher than what Intel can muster, but IPC will still be crap so it doesn't matter. AMD's current moto is "moar corez, moar megahertz!!!"

I don't know why people are so obsessed with clock speeds. Look at Zambezi.

 

[LOL_Wut_Axel]

Pipeline limitations on clockspeed are not a given. It comes down to the underlying process technology (drive current, RC delay) as well as microarchitecture.

Intel has been making marked gains in driver current these past generations, and 3D xtors are only going to improve on that rate of gain because they have much better control over the physical and electrical boundaries that define the channel itself.

If Intel did a straight shrink of Sandy Bridge (not dummy shrink, but a reasonable straight shrink) and dialed in the voltage such that the power consumption was normalized between Sandy Bridge and a hypothetical 22nm Sandy Bridge then I would expect the 22nm version to clock roughly 15-20% higher than the 32nm microarchitectural equivalent based on the pipeline-limited-clockspeed argument you are taking.

Right, but because it's 22nm voltage cannot be raised as high as Sandy Bridge. The smaller the process, the more silicon is degraded from high voltages. 1.4V is about the max you want to go with on Sandy Bridge for 24/7; that limit will definitely be lower with Ivy Bridge.

There's a reason why the new Ivy Bridge models will only be clocked 100MHz higher stock than the Sandy Bridge ones they're replacing.

 

[Puppies04]

That was my whole point,lol was saying pcie 3.0 is a reason to grab a ivy setup.

Yes it is if you are on a SB 1155 setup atm and want to run/already run SLI'd top end cards

The people that bought a setup to run proper 2/3 way sli are already on socket 2011

1. I have seen plenty of rigs running dual 580's on 1155

2. 2011 is a glorified server socket with a price tag to match, not a gaming setup, I would take a quad IB unlocked chip over any 2011 chip and spend the vast quantities of money I saved on something that would actually improve my gaming experience if you have already built the 2011 rig either out of necessity or just for E peen status I see absolutly no reason whatsoever to upgrade to IB though.

BTW I know IB isn't out yet but you could quite conceivably buy a SB setup today then sell the cpu second hand and upgrade to IB cpu and still spend less than a 2011 setup is going to cost you.

 

[gorydetails]

sound like it need a deeper pipeline,,,max clock is around 4.8ghz as we have seen,,,

 

[Nemesis 1]

Right, but because it's 22nm voltage cannot be raised as high as Sandy Bridge. The smaller the process, the more silicon is degraded from high voltages. 1.4V is about the max you want to go with on Sandy Bridge for 24/7; that limit will definitely be lower with Ivy Bridge

I am likeing that ya schooled us on that. So if I were on a path this would be the dead end. This 1.4 you speak of on 32n . What voltage is a 2700k at when idling . Whats a 3700k idle voltage. This driver current IDC spoke to. kinda sounds like a high calaber shot how. how fast will bullet pass threw 3D. How fast are 3Dgates