Intel Ivy Bridge-E Slated for Q3-2013

good that Intel intends to fulfill an obligation - enable its current SB-E users to upgrade..

 

I'm confused.

Headline says Q4-2013, article text says Q4-2012. So which is it?

 

IB-E would finally have 8-cores (which aren't server CPUs), I presume?

 

very long time for their release

 

this will be my next cpu and socket

 

so x79 still have long life time work

 

And no fucks were given. Haswell will be out by then. The relationship between the midrange and the highend will remain unchanged. I don't understand why Intel keeps putting the best technology in the middle product, or why they maintain a highend product that's perpetually slower than their midrange.

This situation ruins it for anyone wanting to do extreme builds. You can't just plop in one super proc anymore and have it all. You have to choose between extreme multi-threaded app performance and everything else. I wish they'd kill this dual consumer platform crap and keep the obscure multi-threaded procs relegated to workstations.

 

True. They started this trend with the first Core ix processors and it seems to have become a habit now. Although I am not complaining seeing as the price/performance ratio to their mid-range CPUs is very reasonable and makes sense for money-restrained consumers who want some bang for their bucks. I just hope that their extreme models are well worth the wait for people who are already on socket 2011, otherwise Intel's gonna have some pissed off customers.

 

Shhh! Don't tell them that. I am really liking my i5-2500k, and hope its successor will be just as good in price/performance and have a ton of performance.

 

How does Ivy-E release ruin that? You look at the relation ship between 1155/1156 against the 2011/1366 sockets the wrong way. It's not mainstream/poor socket versus ENTHUSIAST/rich socket. The bigger sockets were meant primarily for heavy use of multi-threaded applications and the smaller sockets were meant for kicking ass in gaming. A 3770k system seems pretty damn enthusiast to me as will an 8-core 2011 processor EVEN AFTER Haswell comes out. Those that bought a 1366 socket early on were pretty happy when the could plop in a i7-980 processor years later and still hang with the best of them. I have a feeling that the socket 2011 owners will be equally as pleased. Let's not even mention that AMD has nothing even close.

 

I love being able to upgrade on the same socket like I did with 1366
Let's see if Intel will pull out that 10c/20t CPU that has been spotted some time ago

 

Awesome, guess I will be using my Rampage for a long while

 

If they decide to keep using the same internal TIM (3570K/3770K) those Ivy-E of 6 and 8 cores will be a HUGE fail, the temperatures under OC will be terribly high. In the other hand if they decide to go back to the fluxless solder it will just prove they fucked up intentionally with the current Ivy Bridges (1155), obviously a mainstream platform sabotaging the sells of the high end platform doesn't look nice.

 

The successor to the Core i5 2500K is already here. IIRC it’s the Core i5-3570K. Not a massive upgrade though IMO. I have a Core i5 2500K as well as a Core i7 3930K (among others). I don’t think these chips / platforms are in need an upgrade any time soon as they are reasonably powerful.

Would I like an Ivy Bridge-E upgrade in Q3 2013 and or a Haswell upgrade in Q2 2013,….sure but will it be necessary, cost effective or even yield a sizable performance improvement,…. I doubt it,...but maybe,...

 

I'm divided.

I own something in the 2011 socket family right now. Yes, the "Hexa" core chips are a neutered octo core chip, due to thermal issues. Yes, the performance is great for a SB based processor. The issue is that IB comes out just a little while later, and beats SB at lightly threaded applications hands down.

Then we come to find that IB absolutely bakes when you overclock. The problem is that Intel decided to change their manufacturing, for whatever reason I don't know (my theory is cost and decreasing the incursion of midrange CPUs onto the highend market). IB came out as a significantly poorer overclocker, with one very easy to fix flaw.

Flash forward from Q4 2011, to Q3 2013. It's nearly been two years, and the "improvement" that Intel makes is to fix the manufacturing flaw, and just scale the chips up correctly. IB-e is a good performer, isn't neutered, and offers SB-e adopters a big upgrade. Intel is doing some shady things to make IB-e relevant, despite openly stating that desktop CPUs are a dying breed. My optimistic side wants this to be paranoia, but Intel has been guilty of...inappropriate...business practices in the past. There is no indication that they didn't intentionally nerf IB once the AMD offering bombed and there was no real pressure to make the next significant step forward...

 

Do I need to remind everyone that IVB-E was only a die shrink? Manufacturing issues aside, it's practically the same chip on a smaller process that uses less power. Only the skt2011 extreme editions are gimped 8-core CPUs with the two extra cores laser cut. IIRC the 3930k is only 6-cores.

I have a skt2011 rig, no 6-core for me though, and I've been incredibly satisfied with my 3820. It's not IVB, but it does have quad-channel memory, it has 40 PCI-E lanes, and the lid is still soldered to the CPU. I can't complain, it's a well performing CPU with more features than you need.

Is IVB faster? More or less, but it is. Does it have more features and capabilities as X79/SB-E? That's another story and I'm not convinced that it does. For example, you can't get an Intel CPU with VT-d that can overclock until you get at least a 3820. No k-edition intel chip has VT-d. The 3820 has 10Mb of L3 as opposed to 8Mb. The 3820 has supports quad-channel (or triple, dual, or single depending on your memory placement.) Finally (a sizable deciding factor for me,) SB-E has a flexible bclk. How much it will overclock really depends on the CPU. Dumo got the 3820 running at 166mhz bclk. The best mine wants to do is 135, but honestly, 125x36 gives me a 4.5ghz without even touching the Turbo (in fact I have turbo disabled.)

So yeah, there are benefits to going skt2011, it just depends if you want them or not and if you're willing to take a very minor, light threaded performance loss for the platform. I think it's worth it and I bet you that most other skt2011 owners will agree with me. I just hope that IVB-E will have the lid soldered to the chip instead of using a TIM.

 

Let's see.

 

It is mostly but it has some architectural improvement hence the 5-6% gain in performance when compared clock for clock and it is the best tick Intel has released (performance/watt especially).


Anyway, I could care less about that marginal difference in IPC when my workload is heavily threaded, I can tell that if they made an extreme edition with dual QPI I'd be jumping on a dual socket mobo.

 

Mine does 171 BCLK and I'm sure it would go higher but the heat was too great.
Haven't tried the 250 strap as I doubt it would boot at 250.

I'm definitely happy with my LGA2011 rig and can't say I have one bit of buyer's remorse. A little disappointed that they pushed back IVB-E almost a year but that just gives me a reason to push this "little" 3820 further.
And yeah, if IVB-E has that crappy TIM under the lid, you can damn well bet that mine(and others I'm sure) will be de-lidded. No sense in spending $500+ on an enthusiast-class chip if you can't overclock the damn thing because of heat...

 

I have to say, there's a reason for owners liking SB-e, but the price point generally makes it only for those that have serious work to do on a computer. It isn't because of a budget, or performance for single threading, or power consumption. The reason it is enjoyed is highly threaded performance only.

I've ran the 2500k, 3570k, and 3930k. None of them are hugely different in real life scenarios, where programs aren't running for days on end to process something. The difference comes in heavier programs. The 3570 can beat the 2500, but only by a small margin (24 hours to encode with 2500k, 23 with a 3570). The 3570 is beaten by the 3930 resoundingly. A 23 hour encode in the 3570 takes a little under 14 with the 3930.


As far as the bclk, that's largely about the motherboard. My Gigabyte x79-ud5 is an absolute pile of turds. The board crashes when you hit the 1.25, though it is stable enough on the 1.00 setting. Each new revision of the firmware took away more of the overclocking potential, so stay away fro Gigabyte x79 boards if you intend to do anything more than play with the multipliers for each core.


All of this said, it still seems like Intel is pulling the punches on IB. They take nearly two years to update the high end platform, and do so only after their mainstream offering has already moved on to the newest architecture. WTF? Why would you buy the BMW SUV from two years ago, when the BMW roadster from this year is already available? The only people looking at IB-e will be those who bought into SB-e. There will be a small market, which will both make the platform languish and drive the price up. I hate that Intel is doing this, but it makes business sense. The average consumer is willing to deal with less power, as long as the size goes down. We, at this forum, are exceptions to this rule (yeah, it's a cop out, but its true).

 

Only half true. There will likely be a smattering of new motherboards in the LGA2011 architecture released around the launch of IVB-E, meaning more people have the ability to go LGA2011.
People are still building LGA1366 rigs, which should go to show just how well Intel's "enthusiast class" parts age in relation to performance. Hell, an i7 920 with a mild oc will still keep up with if not beat most of the current mainstream offerings.
The biggest reason I see that would keep people from building LGA2011 systems is price. A decent board is around $300 and you only have three(now four) processors to choose from, with the cheapest coming in at about $300 and the next choice up is almost double that price.

 

Ivy Bridge has a much stronger IMC than Sandy Bridge, memory clocks of 1300MHz+ are common. I can imagine Ivy Bridge-E would do the same. I drool at the prospect of quad channel DDR3-2800.