Intel Ivy Bridge Processor IPC and Overclocking @ [H]

sunama said:
How can you say that budget is irrelevant?
R+D budgets (generally) dictate how quickly you can develop/improve a product.
In general, the more money you throw at a problem, the faster it gets solved.
Click to expand...

oh boy.... Project Management is not in your future I see.


In other news, I just recently purchased an i5 2500k and after seeing these results, I'm even more happy about the purchase. Go Intel!
 
I think the main point of IB is not succeeding SB in strict sense, I think it was introduced as a test piece for new 22nm SOI and it is focused mainly on improved iGP, and continued focus on ultrabooks. We all know that there is hardly any difference in IB-SB in raw performance.
The reason why they actually just shrinked IB die is because with 22nm they can make more CPUs on one wafer and than entire production will become cheaper while prices same as SB.
 
Thanks for the review, was F5ing a lot to see this posted. Despite the marginal performance gains, I would surely chose IB over SB, was I building an Intel system tomorrow. There is still a lot to like (improved efficiency; lower TDP is especially important to me since my power company charges insane rates, which I why I am also not folding btw), though I had hoped for even more. Very good to see that tri-gate seems bug free.

Could not help seeing this:

"I recently built a new system for my personal usage. I was aware of what was coming down the pike from Intel"

This is supposed to be "pipe" right? Or my non native English is just unfamiliar with this term.
 
nakedhand said:
"I recently built a new system for my personal usage. I was aware of what was coming down the pike from Intel"

This is supposed to be "pipe" right? Or my non native English is just unfamiliar with this term.
Click to expand...

Either would work. Pike sounds more natural in American English. I believe it is a colloquial term that refers to the "turnpike", often abbreviated as pike, but I am not 100% sure.
 
gunbust3r said:
I would like to see benchmarks of a "best overclock 2700k" vs a "best overclock 3770K"
Click to expand...

Anand tested a 4.8 3770K against a 5.0 2600K and they were neck and neck. Hard to compare max overclocks because it varies so much by processor (and we don't know what is typical of Ivy yet).
 
Yeah its going to take more than this to pull me away from my air cooler 860 @ 4.2ghz, which I have had since launch. That is the most staying power a component has ever had for me. The only way I would move to something else would be if there was a very good deal on a setup with an SLI motherboard, as I currently have a crossfire mobo but have a GTX680.

So, good deals and desire for SLI could move me in the direction of a new mobo/cpu. It will most likely be for the next go around though. I will say that I do like these chips for laptops though. The 2.3ghz quad, 12gb ram, and 660m setup that asus is bringing out is very tempting.
 
XamediX said:
I just purchased an i5 2500k and after reading this review, I couldn't be happier with my purchase. This should hold for at least 2 more generations.
Click to expand...

I purchased an i5 2500k recently, along with an SE motherboard. I'm overclocked at 4.2-4.6Ghz stable and can push it to 4.8Ghz without any heat issue. IB is only beneficial for integrated graphics.
 
RCantw3ll said:
Yeah its going to take more than this to pull me away from my air cooler 860 @ 4.2ghz, which I have had since launch. That is the most staying power a component has ever had for me.
Click to expand...

Agreed, the Lynnfield and Bloomfield cores have had an amazing staying power.

I got a Core i7-920 D0 stepping when they first came out, and seeing that it was able to hit over 4Ghz stably, I really could still have been using it today. I didn't need to upgrade when I did (I thought I was CPU limited, but it turned out I wasn't)

Seeing that I got it in Spring 2009, if I had kept it it would likely have been more than sufficient until the launch of Haswell. Seeing that Haswell is predicted to launch in spring 2013, that would have been 4 years out of a CPU, without it feeling old. That HAS to be unprecedented...
 
One positive to come out of this is that the limited increase in performance from IB might just let AMD catch up to the point where they are relevant in the enthusiast market again.

I kind of doubt they will take any performance crowns between now and Haswell, but they might just be able to play enough catch-up so that an IB vs. AMD contruction equipment of the day doesn't look silly and lopsided.

At least we can hope. Some more competition in this market wouldn't hurt.
 
Zarathustra[H];1038646299 said:
I agree that this would be nice, but you'd ahve to realize that tit would be representative of the tested parts only. If you wanted to say something conclusively about SB and IB in general, you'd need a random sample of ~30 CPU's each, bought in stores across batches to make sure that your sample is sufficiently randomized.

Then you'd need some poor bastard to overclock 60 CPU's jot down the data, and crunch the numbers statistically, in order to say which is faster.


It seems to me right now that SB is going to be faster overall OC vs. OC, but IB will be cooler.

The reason I say this is because the IPC of IB is only between 2.3% and 7.2% higher (average of 4.0%), but the clock speed differential is almost 10%.

In the end, the differences will be very small though.
Click to expand...

http://vr-zone.com/articles/ivy-bridge-vs-sandy-bridge--4.8ghz-quad-core-cpu-showdown/15637.html
 
Forceman said:
Anand tested a 4.8 3770K against a 5.0 2600K and they were neck and neck. Hard to compare max overclocks because it varies so much by processor (and we don't know what is typical of Ivy yet).
Click to expand...

Seems like 4.6ghz is doable with a minimal bump in voltage. Either way, what I would like to see is i7 2600k and i7 3770k at 4.6ghz with GTX680 sli. Mostly due to two reasons. It will minimize the video card bottleneck and test the difference between pci-e 2.0 x8/x8 and pci-e 3.0 x8/x8 in real world usage. I think that would be for more of a solid comparison between the two platforms in what you could expect in real world usage.
 
Zarathustra[H];1038646941 said:
Agreed, the Lynnfield and Bloomfield cores have had an amazing staying power.

I got a Core i7-920 D0 stepping when they first came out, and seeing that it was able to hit over 4Ghz stably, I really could still have been using it today. I didn't need to upgrade when I did (I thought I was CPU limited, but it turned out I wasn't)

Seeing that I got it in Spring 2009, if I had kept it it would likely have been more than sufficient until the launch of Haswell. Seeing that Haswell is predicted to launch in spring 2013, that would have been 4 years out of a CPU, without it feeling old. That HAS to be unprecedented...
Click to expand...

^^this

I bought my 920 in January of 2009, which was the initial C0 stepping. I never would have believed it, but in all this time I've not had any serious urge to upgrade my CPU or platform. To me, that is unprecedented, since I've been building myself a new rig about every 12-18 months since my Athlon 700/Asus K7V. And with every new build I realized significant performance increase as well as newer features. But all of the new stuff coming from Intel since Nehalem, while certainly great, hasn't changed the fact that LGA1366 is more than capable of handling all but the most extreme (and rare) usage scenarios, even more than 3 years after its release.

The only thing I feel I'm missing out on at this moment is native SATA III. But that alone is not enough of a reason for me to build a completely new rig. A good reason would be, say, any feeling that I'm ever waiting on my CPU. But it's never a bottleneck. (I definitely need to upgrade my GPU, though. New GTX680 in a month or two.)
 
Nice review.

Still no urge to upgrade my ol'trusty 920 though... It just keeps on giving this one...

Lack of competition really seems to be showing with the slowness of improvements by Intel.

Everything just runs really smooth, and... skipping another generation I guess.
 
Yakk said:
Nice review.

Still no urge to upgrade my ol'trusty 920 though... It just keeps on giving this one...

Lack of competition really seems to be showing with the slowness of improvements by Intel.

Everything just runs really smooth, and... skipping another generation I guess.
Click to expand...

I think it's really more down to there not being much reason to improve the CPU portion instead of a lack of competition. As you pointed out, even 4 year old 920s are more than enough CPU power for almost all tasks, so there really isn't much incentive to make them faster. Hence the focus on size (to reduce costs), power use, and IGP.
 
SB FTW for now. Temps are huge compared to SB, esc. at round 5ghz.

I will wait for future revisions of i7 3770k..Maybe things will get cooler..
 
I'm in the i7-920 boat as well. I have 2 c0 processors and 1 d0, all pretty much stock. Haven't had much incentive to to overclock them yet. I actually have a 2500k rig that I could move my main machine to, but don't want to go though the process of redoing everything to make it my main. I do use the 2500k for large file compression and video stuff.

I think the thing for me is the only thing that really pushes these things right now is gaming. Since most games these days are console centered they don't really push the hardware that much.

So yeah, switching the whole family over to i7-920's has ended up being a good investment. Here's hoping to squeak out a couple more years from them before I feel like they are getting "old".
 
And Im still sitting smitten with my 2600k WC'ed at 4.4. Really not sure why everyone was waiting for this, all evidence showed that it wasnt going to be anything special. The summary of saying it was a "push" pretty much summed it up precisely.

All its deadon in saying if you dont already have a 2500/2600 and are looking for an upgrade then you may as well go with the 3500/3700 series.
 
Forceman said:
I think it's really more down to there not being much reason to improve the CPU portion instead of a lack of competition. As you pointed out, even 4 year old 920s are more than enough CPU power for almost all tasks, so there really isn't much incentive to make them faster. Hence the focus on size (to reduce costs), power use, and IGP.
Click to expand...

True, but I still think if there was increased competition, then a company would force themselves and come out with something original to distinguish their product.

AMD seemed to have tried and failed to a certain extent in the consumer segment. So Intel has no need to push right now... just keep their fabs running and upgraded to the latest tech. ARM is still way out there...
 
I was interested in seeing a power comparison between ivy and sandy at same clocks, but I didn't notice it while I was skimming the review. All the info that was there was useful for deciding if it was a good idea to switch from sandy.
 
Yakk said:
True, but I still think if there was increased competition, then a company would force themselves and come out with something original to distinguish their product.
Click to expand...

I dunno, something like Hyperthreading or Quicksync or being the first to introduce C-states with clock-gating?

Nahh, too conventional :rolleyes:

AMD seemed to have tried and failed to a certain extent in the consumer segment. So Intel has no need to push right now... just keep their fabs running and upgraded to the latest tech. ARM is still way out there...
Click to expand...

Will you people STOP WITH THE DOOM AND GLOOM?

STOP THE DOOM AND GLOOM.

I REPEAT:

STOP THE DOOM AND GLOOM.

Intel typically takes it slow introducing new micro-architectures, and there's no way they can accelerate new FAB processes beyond their current schedule. Intel has never successfully delivered a major architectural revision AND a die shrink at the same time WITHOUT ISSUES. Neither has AMD.

Intel's die shrinks since 1997:

250nm: Pentium and Pentium II die shrink, no new features.

180nm: Coppermine. Only new feature was on-die cache (and that was already tested with Mendocino Celeron on 250nm)

130nm: The straight die shrink Tualatin was released 6 months before Northwood. Northwood only introduced Hyperthreading to the P4, no other major changes.

90nm: Complete failure with Prescott, which added a VERY deep instruction pipe, better branch prediction, SSE3 and improved hardware prefetch. Doing so much at the same time as a brand-new process (their first with strained-silicon) was a catastrophe.

65nm: Intel learned their lessons and launched with Cedarmill (P4 shrink) and Yonah (Pentium M shrink) well before they launched Core 2.

45nm: Released 45nm Core 2 shrinks 8 months before the 45nm Bloomfield.

32nm: released 32nm Westmere shrink a year before Sandy Bridge.

22nm: Intel takes the smart road and leaves the Core architecture untouched for Ivy Bridge, but takes a risk on the GPU side - their biggest risk on a process node since Prescott. Despite this they managed to please almost everyone (except some hardcore enthusiasts who somehow convince themselves that this was more than a die shrink).

Process improvements learned from Ivy Bridge should pave the way for a drama-free Haswell launch next year.
 
Last edited:
Mozex said:
I was interested in seeing a power comparison between ivy and sandy at same clocks, but I didn't notice it while I was skimming the review. All the info that was there was useful for deciding if it was a good idea to switch from sandy.
Click to expand...

Anand had some overclocked power numbers, and pretty much all the reviews had stock power numbers. Power consumption is down across the board.
 
I'm still running a 920 at 3.2 on stock voltage settings...I thought I would get Ivy Bridge but now I don't see that it's worth the effort, never mind the cost.
 
Have you heard about this? Ivy chips IHS is NOT soldered, instead it has TIM in there.
http://translate.google.com/translate?sl=auto&tl=en&js=n&prev=_t&hl=fi&ie=UTF-8&layout=2&eotf=1&u=http%3A%2F%2Fwww.sf3d.fi%2Fuutiset%2FIvy-Bridge-prosessoreiden-HS-ei-ole-juotettu-kiinni
http://www.techpowerup.com/164858/Ivy-Bridge-Temperatures-Could-Be-Linked-To-TIM-Inside-Integrated-Heatspreader-Report.html
http://www.overclockers.com/ivy-bridge-temperatures

oc-ivydie-2.jpg
 
Perhaps a hsf on the core directly would cool better or better applied tim
 
Zarathustra[H];1038646299 said:
I agree that this would be nice, but you'd ahve to realize that tit would be representative of the tested parts only. If you wanted to say something conclusively about SB and IB in general, you'd need a random sample of ~30 CPU's each, bought in stores across batches to make sure that your sample is sufficiently randomized.

Then you'd need some poor bastard to overclock 60 CPU's jot down the data, and crunch the numbers statistically, in order to say which is faster.


It seems to me right now that SB is going to be faster overall OC vs. OC, but IB will be cooler.

The reason I say this is because the IPC of IB is only between 2.3% and 7.2% higher (average of 4.0%), but the clock speed differential is almost 10%.

In the end, the differences will be very small though.
Click to expand...

Or just look through forums and figure that out through user posted results. It will be randomized, but it does introduce tester error and variability in the other components. The IPC improvement is about what is expected (or a little better for the +7%) in a tick. But what really disappoints is the lack of expected results from a die shrink: Higher clocks, lower heat output for very similar same hardware, and/or additional hardware features like additional cache. More cache really makes a difference in games. ;)
 
I think the biggest problem with running without the IHS would just be making good contact, and getting a good enough thermal transfer to make it work. Intel and AMD went to IHS in part because it helps with temps (also to prevent damage) as the die gets smaller. I'd like to see someone try though.
 
I think the biggest problem with running without the IHS would just be making good contact, and getting a good enough thermal transfer to make it work. Intel and AMD went to IHS in part because it helps with temps (also to prevent damage) as the die gets smaller. I'd like to see someone try though.
Click to expand...

It is not done for better temps. It is to protect the cpu die. Amd used to have open die designs, back with socket A. A lot of people killed their cpus when installing a heatsink. For the cost of adding a heat spreader is rather low, and prevents death of a good cpu with improper installation of a heatsink.

The cpu core itself can be cooled by heatsink/water block, if you want to void your warranty and take a chance at killing your cpu during removal of heat spreader, and installation of your cooling device. I suppose you could make a copper shim to aid you.
 
ebduncan said:
It is not done for better temps. It is to protect the cpu die. Amd used to have open die designs, back with socket A. A lot of people killed their cpus when installing a heatsink. For the cost of adding a heat spreader is rather low, and prevents death of a good cpu with improper installation of a heatsink.

The cpu core itself can be cooled by heatsink/water block, if you want to void your warranty and take a chance at killing your cpu during removal of heat spreader, and installation of your cooling device. I suppose you could make a copper shim to aid you.
Click to expand...

While that's true it also means that it's a slightly smaller surface area to attach your HSF/block directly. And that's given a uniform clean fit. In general it's something that should be avoided unless you can do it correctly.
 
You must log in or register to reply here.
Back
Top