AnandTech Ivy Bridge Performance Preview

[grkM3]

I think the hd4000 will fine in most laptops since those ultrabook are all mostly 1366x768? res anyways and at that resolution that gpu will run almost any game at decent framerates.

 

[Olikan]

I think the hd4000 will fine in most laptops since those ultrabook are all mostly 1366x768? res anyways and at that resolution that gpu will run almost any game at decent framerates.

yep...mine thought also...

ivy and trinity..... nvidia is going to get GTFO in laptops market... not good

 

[Vesku]

Seems like even budget notebooks will soon no longer be lap burners.

 

[Diogenes2]

You have highlighted the major flaw in Intel's current release cycle. Why buy an "-E" proc when you can wait 4 months and save a ton of money on something that will beat it everywhere except extremely threaded tasks.

Why ?

Uhhh, that would be because you run extremely threaded tasks..

 

[Arachnotronic]

Kinda of reminds me of people that built 1366 rigs just before SB's was launched.

What's wrong with 1366? For people who need serious CPU power, Gulftown was the only way to go for almost a year after SNB's launch.

Feel bad for guys like me who built 1156 rigs before SNB's launch. 1366 has had some real amazing staying power and when my 1156 setup died, I went with used 1366 + Gulfy over new 1155 + SNB. Do not regret the decision one bit.

 

[Makaveli]

Nothing is wrong with 1366 i'm on that socket right now.

My point was just that if you build a rig based on the socket late into its life you would feel alittle regret as SB came out and is a mid range product that was faster.

And I agree with you for those of us that need the cpu power Gulftown is still greater than SB just not SB-E.

As for staying power 100% I agree with you.

I'll will be on 1366 until haswell and my rig runs everything now great no performance complaints at all and I don't see any coming up before haswell is out.

If I could find a 980 used for a great price I would still update my 920.

 

[Paul98]

I will be looking forward to the overclocking results on this and the 3570k

 

[Rvenger]

I will be looking forward to the overclocking results on this and the 3570k

+1

Especially since I know the 3770k isn't going to adopt 2600k pricing. Newegg will have it at $369 or maybe even $379 at launch.

 

[Genx87]

Looks like a pretty good upgrade. But I will probably just ride it out until Haswell. Nothing on my radar that needs me to upgrade my computer. Pretty much expecting to play EVE and World of Tanks for the next year or so.

 

[sm625]

I love how he goes out of his way to state that this was not sanctioned by Intel, but then doesnt even say one word about overclocked performance. As if it would take more than 20 minutes to overclock the gpu and run a couple fps tests. We pretty much know the cpu is going to OC up around 5GHz, so there's no real need to get into that. But the gpu? Come on this is a big wildcard at this point. For all we know the thing could beat a llano when both are overclocked. For shame...

 

[Rvenger]

I love how he goes out of his way to state that this was not sanctioned by Intel, but then doesnt even say one word about overclocked performance. As if it would take more than 20 minutes to overclock the gpu and run a couple fps tests. We pretty much know the cpu is going to OC up around 5GHz, so there's no real need to get into that. But the gpu? Come on this is a big wildcard at this point. For all we know the thing could beat a llano when both are overclocked. For shame...

And we all know how well the unlocked Llanos overclock. Its no walk in the park either. 🙄

 

[Fjodor2001]

We're always looking for the next release to be as dramatic a performance improvement as we witnessed in going from P4 to Conroe. That was nearly a doubling of performance per clock.

Does it surprise anyone that the impressive jump in performance, power efficiency, and so on for the Intel CPUs happened at the same time as AMD was in its prime (i.e. when AMD introduced 64-bit, multicore, high power efficiency)?

If Intel was really being pushed again, what could we expect a year from now? 8 core, 5+ GHz, 2-3x IGP performance?

 

[Idontcare]

Does it surprise anyone that the impressive jump in performance, power efficiency, and so on for the Intel CPUs happened at the same time as AMD was in its prime (i.e. when AMD introduced 64-bit, multicore, high power efficiency)?

If Intel was really being pushed again, what could we expect a year from now? 8 core, 5+ GHz, 2-3x IGP performance?

Absolutely. No question, without an internal crisis to command priorities of "performance, at any expense!", management will naturally gravitate towards the prioritization of maximizing metrics of interest to the shareholders and those metrics come at the expense of deprioritizing aggressive R&D.

AMD did it when they were at the top, it in part led to their present cascade of stumbles. Intel did it as well, which led to the creation of the P4.

It would be nice to see Intel compelled to unleash the dogs of war. But they have no reason to, not now at least.

 

[Makaveli]

Does it surprise anyone that the impressive jump in performance, power efficiency, and so on for the Intel CPUs happened at the same time as AMD was in its prime (i.e. when AMD introduced 64-bit, multicore, high power efficiency)?

If Intel was really being pushed again, what could we expect a year from now? 8 core, 5+ GHz, 2-3x IGP performance?

Good point but to be honest i'm not sure.

We've come to the point now where quad cpu's are the sweet spot on the desktop with desktop workloads.

And if software continues to be single threaded for the post part. IPC matters more than core count. It may be a few more years before you actually see 8 core chips that are affordable in the customer market.

The need just isn't there right now for it.

 

[Olikan]

from Kyle...

Ivy Bridge will clock no better than Sandy Bridge, in fact, it will likely clock worse as it will not be able to handle voltages much above 1.35 without a good chance of burning up the processor. 5 to 15% IPC improvement as clock scales. You will likely have a lot better shot of pushing a SB processor across the 4.8GHz threshold and keeping it stable than you will with Ivy Bridge. However, it looks to be pretty easy clocking up to the 4.5GHz level or so many times requiring little voltage tweaks at all.

http://hardforum.com/showpost.php?p=1038465252&postcount=85

 

[pelov]

It would be nice to see Intel compelled to unleash the dogs of war. But they have no reason to, not now at least.

I suppose that's a pretty good reason 😛

 

[Concillian]

from Kyle...


http://hardforum.com/showpost.php?p=1038465252&postcount=85

I know a lot of people probably don't like the sound of that, but it sounds really good to me. I buy ticks for the power efficiency and I my current i3-530 @ 4GHz is only a slight voltage increase over stock. If an IB i5 is going to get 4.5 with minimal voltage, it could end up at 50-100W less than OCed SB at load.

It's the kind of thing though that might get significantly better after a new stepping, like the way Wolfdales were so-so and had to be pushed to 3.5 GHz... until the stepping where all the e8xxx CPUs OCed to ~4 GHz.

 

[gplnpsb]

I know a lot of people probably don't like the sound of that, but it sounds really good to me. I buy ticks for the power efficiency and I my current i3-530 @ 4GHz is only a slight voltage increase over stock. If an IB i5 is going to get 4.5 with minimal voltage, it could end up at 50-100W less than OCed SB at load.

It's the kind of thing though that might get significantly better after a new stepping, like the way Wolfdales were so-so and had to be pushed to 3.5 GHz... until the stepping where all the e8xxx CPUs OCed to ~4 GHz.

On that note it's interesting that Ivy has already been through nine respins and is at the E1 stepping. After all this time and work I wonder if Intel would bother doing another stepping after launch. For comparison Wolfdale launched after 6 respins at C0, and Gulftown launched after just two respins at B1.

 

[mak360]

Wheres the overclocking?

 

[Concillian]

On that note it's interesting that Ivy has already been through nine respins and is at the E1 stepping. After all this time and work I wonder if Intel would bother doing another stepping after launch. For comparison Wolfdale launched after 6 respins at C0, and Gulftown launched after just two respins at B1.

"Bothering" is largely a function of where they are in terms of yield and performance. It's not a matter of how many re-spins happened previous, it's a matter of whether they stand to gain (monetarily) from another stepping.

When you're shipping hundreds of thousands or potentially millions of dies a month, it doesn't take much per unit cost change in order to have really, really short return on investment.

I know in the media industry we'd spend pretty large sums going after one cent per platter cost improvement.

 

[Ryun]

I just picked up a 2600K for $265 after taxes from Microcenter. I think I made the right decision to pull the trigger rather than to wait for Ivy. Much thanks for the preview all the same.

 

[tweakboy]

IMO Ivy Bridge is just Intel buying time for Haswell.

Its very similar to sandy bridge which is why it will fit Sandy mobos with bios flashed.

If you have a Sandy now forget about IB and its 10 percent performance increase which you will never see in real world app....

If you dont have a Sandy, wait for Mandy,,, or I mean Haswell and do a nice 2013 upgrade. gl

Im holding up until this albino and GPU inside a CPU craze dies and they start focusing on adding more cores. I need cores for my Sonar X1 , and I need 128GB ram for all the samples I load in template, thx

 

[Fjodor2001]

IMO Ivy Bridge is just Intel buying time for Haswell.

But will the IB->Haswell jump really be that much larger than the SB->IB jump? Aren't we looking at the same ~5-15% CPU performance increase, and perhaps 30-50% IGP performance increase, going from Ivy Bridge->Haswell?

 

[ElFenix]

i disagree with the assertion in one of the early paragraphs that sandy came out with a conroe sized beatdown. compare a conroe like the E6550 to the previous pentium extreme. even with just 66% of the clockspeed the conroe was beating the pentium extreme by 50% to 80% on sysmark, nearly 50% better on 3dsmax composite, and 50% to well over 200% on games. comparing similarly spec'd nehalem to sandy, it's more like 10% for most stuff. a nehalem with a big clock lead isn't losing to sandy.

But will the IB->Haswell jump really be that much larger than the SB->IB jump? Aren't we looking at the same ~5-15% CPU performance increase, and perhaps 30-50% IGP performance increase, going from Ivy Bridge->Haswell?

anyone have an overview of haswell architecture? from what i've heard haswell is more about power consumption than overall speed improvements. i have to wonder if intel will be doing a bigLITTLE type architecture for it.

 

[Makaveli]

But will the IB->Haswell jump really be that much larger than the SB->IB jump? Aren't we looking at the same ~5-15% CPU performance increase, and perhaps 30-50% IGP performance increase, going from Ivy Bridge->Haswell?

After I used my universal translator to decipher tweakboys post he is correct Ivy is just a small upgrade from SB.

However I expect a bigger increase from IVY to haswell as we are going tick to tock.

where as IVY is a tick and SB was the tock.