Ivy Bridge...

[fixall]

I've been seeing some articles floating around saying that Intel is shooting for a 20%-30% speed increase over the Sandy Bridge processors with their upcoming Ivy Bridge CPU. So going by the standard that most I5-2500k's can hit a rock solid 4.5Ghz with air cooling pretty easily... Are they saying the new Ivy Bridge CPU's could be able to hit 5.5Ghz-6Ghz...!? Am I looking at this the wrong way? Does anyone know of a good Intel rumors site? lol.

 

[lassik]

Nah, you're seeing it wrong. It'll have a tweaked architechture and maybe new extensions to give it more performance per Megahurt.

But yeah, it'll probably have a higher frequency, but only by around 100-200Mhz at stock.

 

[Toxxyc]

Speed increase does not only rely on an increase in the core clock speed, remember that. You can increase performance/speed in a number of ways, and the most common two is to push the clock speed (up the GHz) and to enable Hyper Threading on the CPU (one core is seen and used as two, making applications requiring more threads run much faster). Other methods include a refined architecture ("Sandy Bridge" design over something like "Nehalem" design), smaller manufacturing process (32nm improvement over 45nm), increased FSB speed (newer 1333MHz FSB over older 800MHz FSB from CPU's like Pentium 4) and physically increasing the core count on the CPU (Quad Core and Dual Core over single core CPU's).

Expect to see more of the refinements rather than an overall increased clock speed. A good comparison between Sandy Bridge and Ivy Bridge would be almost the same as comparing the Intel Core i7-975 to the Core i7-980X. Differences aren't in clock speed (both are 3.33GHz reference), but rather in design:

- 975 has 45nm manufacturing process, 980X has 32nm
- 975 has 8MB Cache, 980X has 12MB
- 975 has 4 cores, 980X has 6 cores
- Both CPU's have Hyper Threading Technology
- They have the same bus speed

But still the 980X is a much more powerful processor, only due to the design refinement.

 

[apache_lives]

lassik - its "megahertz"

toxxyc - 980x isnt more powerful because of refinement, its more cores and cache scaled down to 32nm, it hasnt been refined in reality, only modernized and added to.

pentium 4's also never featured a 1333mhz fsb - the Core 2 Duo E6x50 and Quad Extreme's were the first to have it standard.

sandybridge to nehalem is more refinement and redesign which is why its closer to the 6 core 980x in performance whilst only sporting 4 cores - ipc has increased

to the OP: that 20-30% speed increase could be because of clock speed, core count, ipc, pcie-lanes, link speeds or boost frequencies - wait for benchmarks for exacts - its not 20-30% in terms of clock speeds but it will probably have headroom like that anyhow - remember there target (intel) is more performance with similar or better thermal/power specs, intel could otherwise easily release ~4.5ghz (average sandybridge overclocks) cpuz with ~150-200w TDP's

 

[Toxxyc]

Yeah modernized is what I meant with "refined". But you said exactly what I meant only in different words...

I know P4 never had 1333MHz FSB (trust me, I ran one for AGES), that's why I said it only reached 800MHz (or 667MHz on the crappier ones).

But yes, what you said is a better description, I confess...

 

[fixall]

Ahhh... So most likely, when they say 20-30% increase in speed it will be in best case scenerios and may not always apply to the application I'm using. Like playing a game that can't make use of the hyper-threading/extra cores. But on the other hand it should still be a better built processor with more overclocking/performance potential.

Am I getting it now??
Thanks for the help everyone!

 

[Zenthar]

Could also be the quad-channel memory. Curious to see if it really makes a difference outside synthetic benchmarks.

 

[tecmo34]

I expect 4.8-5.0Ghz for daily use with IB

I think that would be nice but with there most likely only being 6-cores & 8-cores being released on the desktop version (LGA 2011 / Socket R), I don't expect to see those high of overclocks.

http://en.wikipedia.org/wiki/Sandy_Bridge

 

[Toxxyc]

Could also be the quad-channel memory. Curious to see if it really makes a difference outside synthetic benchmarks.

Well, seeing as how small the difference is between dual-channel and triple channel even in synthetic benchies, the CPU's will have to be truly epic to be able to use the difference in triple and quad channel in a benchmark productively and thus produce a better score.

But Sandy Bridge surprised all, why shouldn't Ivy Bridge do the same?

 

[dogman_1234]

But Sandy Bridge surprised all, why shouldn't Ivy Bridge do the same?

Not a new design.

 

[Toxxyc]

Not a new design.

True that.

I still believe Ivy Bridge will be awesome. "Big Daddy" stuffs, specially in the LGA2011 socket.

 

[Zenthar]

But Sandy Bridge surprised all, why shouldn't Ivy Bridge do the same?

I think what surprised most is how well it OCed and the fact that the unlocked clock multiplier doesn't come with a x5 price multiplier as well. Would Sansdy Bridge made as much noise if only the i5-2500 and i7-2600 had been released?

 

[Toxxyc]

I think what surprised most is how well it OCed and the fact that the unlocked clock multiplier doesn't come with a x5 price multiplier as well. Would Sansdy Bridge made as much noise if only the i5-2500 and i7-2600 had been released?

LOL no not at all, I can imagine how enthusiasts would rather stick to Core 2 Quad rather than go to i5-2500 for the purpose of overclocking.

Now we have to wait for the "E" versions of Ivy Bridge and see if they live up to the hype they are currently receiving...

 

[Wisecracker]

Nah, you're seeing it wrong. It'll have a tweaked architechture and maybe new extensions to give it more performance per Megahurt.

But yeah, it'll probably have a higher frequency, but only by around 100-200Mhz at stock.

You need to trademark that ... :lol:

 

[hardawayd]

I find the timing of this breakthrough announcement interesting since there are rumors of Apple going to ARM and at the CES Microsoft said the next Windows will also run on ARM and Nokia dumped Meego that would have used Atom processors and now they are going Windows 7. There is increasing attention given to having servers that use ARM---the facebook datacenter engineer is watching this---the Yahoo engineer over their datacenter is watching this and I'll bet Google's datacenter management is watching for ARM to become feasible in their datacenter. It is all about energy efficiency and costs given performance is adequate. Intel has a huge problem and unless the power and costs change dramatically for their processors it does not look good for them.

 

[Kewlx25]

Could also be the quad-channel memory. Curious to see if it really makes a difference outside synthetic benchmarks.

Probably what you're thinking for consumer stuff. It's a quad channel because it's based off the server line. For the server line, quad channels is awesome. If you had a quad socket motherboard and all sockets CPUs decided to access another CPU's memory, having quad channels would help tremendously. If you had 8 cores per cpu with 2 threads because of HT, that's 16 threads per socket. Multiply by 4 sockets and you got 64 threads trying to access quad channel memory. I'm sure it would bottleneck.

Another nifty thing the IB 2011 socket has is a 4x PCIe3.0 lanes built in. This is supposed to be used for a bunch of SATA6.0 ports. So, instead of just having your SATA6 ports on your chipset, you would also have a pack of SATA6.0 ports on their own direct dedicated 4x(4Gbytes/sec) PCIe3 connection to the CPU.

Servers are going to love that. I saw mention of something like 14 SATA6 ports for one setup. Imagine a server with 14 SSDs, each with a 750MB/sec SATA6 connection to a 4GB link strait to the CPU.. /drool Obviously 14 SATA ports is much more bandwidth than what the 4x PCIe can handle, but that also means you could saturate that 4GB/sec link quite easily. 4GB/sec of IO is awesome.

 

[Leaps-from-Shadows]

IVB isn't quad-channel. You're confusing IVB with SNB-E.