Anandtech Review of Ivy Bridge is Up now! Other sites added

[Gryz]

How big is the chance that the next stepping of Ivy Bridge would be better at dispersing heat ?

When is a new stepping expected ?

Does anyone remember how often new steppings were released with Sandy Bridge, Nehalem and other generations of CPUs ? Is there a trend here, or is it semi-random ?

Just trying to decide if it would be worth it to wait for the next stepping.

 

[BallaTheFeared]

Overall it's 0% faster when you take headroom into consideration. According to these transcode #'s that don't use GPU assist like the bench's you provided, a 4.5ghz Ivy is not comparable to a 4.7ghz/4.8ghz Sandy.

http://vr-zone.com/articles/ivy-bridge-vs-sandy-bridge--4.8ghz-quad-core-cpu-showdown/15637.html

I don't have an IGP so I'm just comparing processor to processor. If it's IGP you're after enjoy the complete lack of post processing .

It's faster, albeit most single digits, but the point then I guess for me is two key facts:

Exhibit A:

From the large pool of CPUs that we have tested/binned, Ivy Bridge uses 25% less power on average clock for clock due to the lower voltages required (in this case Sandy's 1.48v to Ivy's 1.27v).

That's 4.8GHz at 1.27v, their sample is worse than my i5-2500k though, mine still needs 1.4v for 4.8GHz.

Exhibit B:

Ivy Bridge is a (subzero) overclocker's wet dream, able to hit core frequencies near 7GHz with no cold bugs (unless motherboard induced).

My comp doesn't post to the dedicated gpu at sub 0 temps, this was both disheartening and a major problem in January when I bought it because I had just sold a 1090T that was running -10C @ 4.5GHz. Not only would it perform slower when colder, it would require me to use a blow dryer to heat the block just to get it to post.

Ivy solves key problems I've had with SB, high voltage, and issues with cold temps. :thumbsup:

 

[TridenT]

Still sitting here on a E7200.

I'm gonna keep waiting, I guess.

 

[Zucker2k]

How big is the chance that the next stepping of Ivy Bridge would be better at dispersing heat ?

When is a new stepping expected ?

Does anyone remember how often new steppings were released with Sandy Bridge, Nehalem and other generations of CPUs ? Is there a trend here, or is it semi-random ?

Just trying to decide if it would be worth it to wait for the next stepping.

The heat is being dispersed alright. If it wasn't, heat will quickly build up forcing throttling or a shutdown. What most are whining about are the core temperatures BEFORE the heat is dispersed. IVB produces LESS heat than SB, but the cores are registering a higher temp because of their close proximity to each other - creating a hotzone inside which Intel has conveniently, no, stupidly placed a sensor. Maybe Intel should take a page from AMD and move the sensor outside the die. Then we can do 4.5GHZ and 50c and throttle at 69c.

 

[blackened23]

The heat is being dispersed alright. If it wasn't, heat will quickly build up forcing throttling or a shutdown. What most are whining about are the core temperatures BEFORE the heat is dispersed. IVB produces LESS heat than SB, but the cores are registering a higher temp because of their close proximity to each other - creating a hotzone inside which Intel has conveniently, no, stupidly placed a sensor. Maybe Intel should take a page from AMD and move the sensor outside the die. Then we can do 4.5GHZ and 50c and throttle at 69c.

This obviously explains why so many review websites could not go higher than 4.5-4.6ghz on test samples without constant BSOD's on air cooling.

If you have proof of what you stated, i'd be interested, otherwise it sounds like optimism. Most reviewers still get BSOD's far quicker than they did on SB, hell, look at the reviews on this website. The overclocking review mentioned that most of their samples could not go past 4.5ghz.

edit : this is not to say IB is bad, its an excellent product - but I was hoping for more. I wanted to upgrade, but this doesn't seem to do much except have better iGPU performance which I don't care about. If it had been a 10% ipc increase with the same oc headroom as sandy bridge i'd be sold for sure. Thats just me though

 

[Don Karnage]

This obviously explains why so many review websites could not go higher than 4.5-4.6ghz on test samples without constant BSOD's on air cooling.

If you have proof of what you stated, i'd be interested, otherwise it sounds like optimism. Most reviewers still get BSOD's far quicker than they did on SB, hell, look at the reviews on this website. The overclocking review mentioned that most of their samples could not go past 4.5ghz.

Blackened is correct. Up to 4.5Ghz my 3570K is great on stock voltage but go beyond and you need to start turning up the voltage a lot. Look forward to seeing if the new batch's will be better at overclocking.

 

[Kenmitch]

How big is the chance that the next stepping of Ivy Bridge would be better at dispersing heat ?

When is a new stepping expected ?

Does anyone remember how often new steppings were released with Sandy Bridge, Nehalem and other generations of CPUs ? Is there a trend here, or is it semi-random ?

Just trying to decide if it would be worth it to wait for the next stepping.

Unless there is a major flaw/bug most likely there will be no new stepping. I think SB has one stepping from launch till now. SB-E had a flaw so it has had a new stepping.

IB is just like the rest of the cpu's as far as the silicone lottery goes.

 

[Zucker2k]

Blackened is correct. Up to 4.5Ghz my 3570K is great on stock voltage but go beyond and you need to start turning up the voltage a lot. Look forward to seeing if the new batch's will be better at overclocking.

Yes, it's physics. Hotter cores will need disproportionally more volts to push them even higher. 22nm is a low voltage process designed with mobile in mind, not high performance/high frequency like 32nm. That we're seeing these speeds is testament to a lot of work done tuning the process while reining in voltage/power consumption. 77w is really nearing the upper limit of tolerance for the process; anything beyond that needs better cooling, which every transistor loves.

 

[LOL_Wut_Axel]

The real story here is mobile, gents.

For the first time laptop processors have reached the performance of high-end desktop processors. The Core i7-3920XM is as fast as the i7-3770K, both having a 4C/8T Turbo Boost of 3.6GHz. Even the "slower" i7-3820QM is as fast as the i7-3770, both running at 3.5GHz at full load. The 3720QM, one of the lower tier models in the lineup and the one you'll be able to configure to in $1000 laptops right when it's available, is less than 5% slower than the 3770 due to running at 3.4GHz when its four cores are pegged.

Insane amount of CPU performance for laptops now.

 

[IGemini]

Agreed, this'll be a good thing for mobile processors. And with good timing now that ultrabooks are starting to hit the market.

Have no plans to upgrade to IB, yet I like seeing the IPC improvements with each family, makes me look forward to Haswell. Over my C2Q it'll amount to something like a 40-100% performance increase at almost 50% power draw reduction.

 

[jmarsiglio]

How good is the graphics capability of this chip? I'm not a huge gamer, but I do have a 30" LCD running at high resolution and would like to play the occasional current PC game.

Many thanks.

On native resolution, almost impossible for anything more than semi-demanding games. But at 1366x768 or lower you can play all games on low-medium settings.

note: At 2560 x 1600, even the mighty gtx 680 will have trouble on a lot of games.

 

[rgallant]

I think I'll wait for a batch spread sheet just to make sure we are not looking at the best ib has to offer.

-so if 80-85c is safe ? that's ok with me should I run at the high clocks.[4.7-4.9]

-still better or the same as running high volts [1.45+] on sb that people except as norm.\safe now days.

-roll of the dice what kills a chip faster - excessive temps or volts.

 

[Tempered81]

roll of the dice what kills a chip faster - excessive temps or volts.

Both are bad. Could run 1.9Pll/1.9vcc/101C constant all day. Electromigration, brittle silicon, degradation, voltage response, overclockability, lifespan/longevity decrease. Probably just "wear it out a 'lil" over time.
Few are actually killing Intel's tank-like chips, it's difficult to gauge. Temp is not a definite killer because THERMTRIP & Prochot will save the chip, like OCP saves the board, voltage spikes 2000mv vccpll have been known to do it. Motherboard malfunctions, pwm burnouts, power spikes, cheap components, phase explosions, bent pins, bad install/fups, mods, brittle solder, user error.
This forum is for 'safe overclocks'. There are other outlets where you can see 2200mv+ like Hwbot.

 

[Minot]

Seeing that anandtech was able to run Ivy at 1.1V @ 4.4Ghz at 65C, that seems like a fantastic sweet spot for a daily build.

I was most intrigued by this comment above after reading the comments here and some of the Anandtech articles. I agree that 1.1v @ 4.5 GHz seems to be the IB sweet spot for speed and heating according to this article: http://www.anandtech.com/show/5763/undervolting-and-overclocking-on-ivy-bridge. But what I want to know is how they achieved this sort of OC at 1.1v. All of the comments from Don, LagunaX and others seemed to be well above 1.2v. In fact, Don's current sig has 4.6 GHz @ 1.23v on his 3570K.

What settings did Anandtech use to achieve 4.5 GHz at 1.1v?

 

[Smoblikat]

any point upgrading from an i7 950? Primary functions gaming and 1080p movie watching

Not really, keep your graphics card updated and youll be fine. Games wont be limited by your 950 for a while.

 

[Olikan]

Ivy Bridge is a (subzero) overclocker's wet dream, able to hit core frequencies near 7GHz with no cold bugs (unless motherboard induced).

bulldozer hits 8Ghz

 

[Smoblikat]

try 14% in x264...

So the ONLY thing ANYONE will do with their computer is x264?

 

[aman74]

Overall it's 0% faster when you take headroom into consideration. According to these transcode #'s that don't use GPU assist like the bench's you provided, a 4.5ghz Ivy is not comparable to a 4.7ghz/4.8ghz Sandy.

http://vr-zone.com/articles/ivy-bridge-vs-sandy-bridge--4.8ghz-quad-core-cpu-showdown/15637.html

I don't have an IGP so I'm just comparing processor to processor. If it's IGP you're after enjoy the complete lack of post processing .

I assume you're referring to Quick Sync? That seems to be the one area of substantial benefit with the new IGP. I'm just not sure how it will benefit me.

What do you mean about the post processing?

When I first read about Quick Sync the reviews seemed to imply it was doing it's own form of compression and there were quality issues with what was rendered. I thought initially that you could just use it with any encoding process you would normally use as long as it was supported. If that's the case, why would it affect quality?

Right now I'm using AnyDVD/CloneDVD and I'm wondering if it would benefit me there. Since I'm not doing a lot of gaming or a ton of video rendering, if I can get away with not getting a discrete GPU then IB would represent a savings. Supposedly it also beats any current discrete card for encoding so even if you have a discrete card for gaming, etc.. this would still represent a performance increase when using Quick Sync?

Sorry, just trying to understand all this and the practical applications. One concern though is that it still doesn't support 24p properly as far as I know.

Thanks,

Anthony

 

[garagisti]

The heat is being dispersed alright. If it wasn't, heat will quickly build up forcing throttling or a shutdown. What most are whining about are the core temperatures BEFORE the heat is dispersed. IVB produces LESS heat than SB, but the cores are registering a higher temp because of their close proximity to each other - creating a hotzone inside which Intel has conveniently, no, stupidly placed a sensor. Maybe Intel should take a page from AMD and move the sensor outside the die. Then we can do 4.5GHZ and 50c and throttle at 69c.

Actually you're right about the lower temp reading that you'll get, but the cores will still be that close to each other, and may be you could see chips toasting proper then. As right now, BSOD's keeping chips from toasting themselves.

 

[georgec84]

Is it me or is the IB iGPU not as bad as some people are making it out to be?

This review from notebookcheck has a fairly positive assessment:

http://www.notebookcheck.net/Intel-HD-Graphics-4000-benchmarked.73567.0.html

I guess the benchmarks from AMD's Llano will soon be replaced by Trinity, but still it looks like for midsize laptops you can game at decent FPS with medium settings at 720p or so.

 

[ElFenix]

why don't the K parts support VT-d?

 

[Atreidin]

why don't the K parts support VT-d?

"Market segmentation"

 

[Khato]

Is it me or is the IB iGPU not as bad as some people are making it out to be?

This review from notebookcheck has a fairly positive assessment:

http://www.notebookcheck.net/Intel-HD-Graphics-4000-benchmarked.73567.0.html

I guess the benchmarks from AMD's Llano will soon be replaced by Trinity, but still it looks like for midsize laptops you can game at decent FPS with medium settings at 720p or so.

This is the side that all too often seems to be overlooked - there's a rather marked difference between the desktop and mobile Llano SKUs. Drop from the 600 MHz GPU clock of the desktop A8 line to the 444 MHz GPU clock of the mobile A8 and suddenly Intel's HD 4000 is competitive... especially since some of the Intel mobile SKUs actually have a higher maximum graphics turbo frequency than the highest desktop (1250 MHz vs 1150 MHz, I'm not counting the 'extreme edition' mobile SKU.)

Regardless, the real question is how exactly Intel's upcoming dual-core mobile Ivybridge will compare to AMD's upcoming Trinity at each TDP target. There's little question that Trinity will win out on the desktop, but once constrained to mobile TDP's will that still be the case? Just what kind of frequencies will be possible in the 17W TDP with a dual core Ivybridge running on 0.9V instead of the ~1V that Sandybridge ULV SKUs seem to use under load? Or will they go even lower than 0.9V? The plot of transistor gate delay vs operating voltage that Intel presented back when it unveiled their tri-gate 22nm process claims similar gate delay at 0.8V on 22nm as 32nm had at 1.0V.

 

[frostedflakes]

Will definitely be interesting to see how they stack up, I didn't realize mobile IB was that competitive with mobile Llano. Pretty impressive results in that notebookcheck review.

Is Trinity supposed to have graphics turbo? That would help if they added that for the mobile parts. I know some of their C and E series APUs support graphics turbo, but I don't think I've read anything about mobile Trinity supporting a similar feature.

 

[Iketh]

So the ONLY thing ANYONE will do with their computer is x264?

How in the world did you extract that information from my post? Please, I'm dying to know.