Upcoming IB Xeons

[ShintaiDK]

Not so much to say, coming in Q3.

 

[videogames101]

ohhhh yes

 

[Arachnotronic]

I will be getting an Ivy Bridge E.

 

[NTMBK]

10 core EP? Oh hell yeah. That thing is going to fly in a dual socket machine.

 

[T_Yamamoto]

10 core EP? Oh hell yeah. That thing is going to fly in a dual socket machine.

F@H

 

[NTMBK]

F@H

Sod that, just massively parallel compilation

 

[vertika]

right saying @NTMBK.
and this is also right "Just insult each other all you want until one of your keyboards explode. Last keyboard functioning will surely show who's the biggest, baddest internet tough guy. Or not

 

[tweakboy]

I will be getting an Ivy Bridge E.

Me too my friend. lets see what cores they come in. Cuz they said 6 to 12 cores Q3 2013 I dont even have to upgrade but I will always have the choice of grabbing a 8 core Ivy E in 2 years or so and slapping on my mobo.

 

[T_Yamamoto]

right saying @NTMBK.
and this is also right "Just insult each other all you want until one of your keyboards explode. Last keyboard functioning will surely show who's the biggest, baddest internet tough guy. Or not

One of our lovely admins posted that in a ATOT thread

 

[bunnyfubbles]

hopefully this means native 10 core and native 6 core and we get an 8-10 core i7 for the same $600-1000 price range of the current 6 core SB-Es and hopefully even see 6 core options in the $300-500 range

 

[ShintaiDK]

hopefully this means native 10 core and native 6 core and we get an 8-10 core i7 for the same $600-1000 price range of the current 6 core SB-Es and hopefully even see 6 core options in the $300-500 range

A 10C IB-E i7 would most likely be too slow clocked for users to accept. Expect 6 and 8 cores.

 

[cytg111]

hopefully this means native 10 core and native 6 core and we get an 8-10 core i7 for the same $600-1000 price range of the current 6 core SB-Es and hopefully even see 6 core options in the $300-500 range

I just cant help but imagine that such a 10 core beast would benefit massively from TX memory as in upcoming haswell .. until reviews hit the street we wont know, but just how much faster a 8 core lower clocked IB will be than a higher clocked 4 core haswell ? I'm not so sure it will run circles around it even in highly threaded apps.

 

[ShintaiDK]

I just cant help but imagine that such a 10 core beast would benefit massively from TX memory as in upcoming haswell .. until reviews hit the street we wont know, but just how much faster a 8 core lower clocked IB will be than a higher clocked 4 core haswell ? I'm not so sure it will run circles around it even in highly threaded apps.

It depends a lot on the instruction mix. In Linx for example, a quadcore Haswell will beat a similar clocked octocore IB. For anything at or below 4 threads Haswell will (always) win.

 

[exar333]

A 10C IB-E i7 would most likely be too slow clocked for users to accept. Expect 6 and 8 cores.

If I can get 8C/16T at similar speeds of my SB-E, I will be very happy. That is 2 more real cores for my VMs and some extra cache as well. Throw-in a small bump in IPC, and highly-threaded applications will see some great improvements from SB-E.

Edit: I would do a happy-dance if Intel gave us a 'present' and released a monster 10C EE CPU running at a good clock speed with a high TDP (e.g. 165w). That would rock...

 

[bunnyfubbles]

A 10C IB-E i7 would most likely be too slow clocked for users to accept. Expect 6 and 8 cores.

an overclocker doesn't have to accept anything

 

[cytg111]

It depends a lot on the instruction mix. In Linx for example, a quadcore Haswell will beat a similar clocked octocore IB. For anything at or below 4 threads Haswell will (always) win.

yea? Ill take that wager and make it "For anything at or below 5 threads Haswell will (always) win" .. and maybe even gaining ground on the 6th .. But it's just guesstimates, but reviewing the anand article on TX again and i do expect some small 'big' changes with haswell when the thread count rises. Of course, add special apps for avx2 in the threadmix and even the 10 core IB-E is in danger of being beat by a 4 core haswell ..... (says my gutfeeling which is as scientific and predictable as the next crystalball(tm) of your choice).

 

[BrightCandle]

The lack of additional cores for Haswell is a bit disappointing, despite being known for a long time. We have had quad cores as the top mainstream since the C2Q, 6 years ago. Back then Intel was telling us we would be running thousands of cores and here we are 3 process jumps later and it hasn't happened. 6 cores would have been nice, there are plenty of applications that benefit from that already.

No one is going to write software that requires 10-20-100's of cores until it looks like the hardware will be there. The software follows the hardware everytime.

 

[Ferzerp]

The lack of additional cores for Haswell is a bit disappointing, despite being known for a long time. We have had quad cores as the top mainstream since the C2Q, 6 years ago. Back then Intel was telling us we would be running thousands of cores and here we are 3 process jumps later and it hasn't happened. 6 cores would have been nice, there are plenty of applications that benefit from that already.

No one is going to write software that requires 10-20-100's of cores until it looks like the hardware will be there. The software follows the hardware everytime.

Amdahl's law along with few problems being inherently parallel enough to warrant MOAR CORES. It's mostly wasted silicon (and power that can be used for faster, fewer cores) except for edge cases (and that's even if we pretend programmers are up to snuff to extract all the theoretical performance out of every problem: something far from true)

 

[jpiniero]

You won't see >4 cores in the mainstream line until it's feasible in laptops to do so. So most likely Skymont at the earliest.

 

[Ajay]

Amdahl's law along with few problems being inherently parallel enough to warrant MOAR CORES. It's mostly wasted silicon (and power that can be used for faster, fewer cores) except for edge cases (and that's even if we pretend programmers are up to snuff to extract all the theoretical performance out of every problem: something far from true)

So, you didn't get the memo that Amdahl's law has been broken? Amdahl's law looks at problems of a fix size. If the problem is looked at in the time domain, then it becomes apparent that the scale of the parallel portion of thw work done can grow enormously. Therefor the parallel portion of work done can dwarf the serial portion of work done for a large variety of problem domains. Otherwise, Massively Parallel High Performance Computing would never have evolved.

 

[Ferzerp]

Only for specific problems, and addressing each one in such a way to scale to that level is a massive task in itself. Your web browser, office suite, and games just aren't well suited to scaling to massive core counts, it doesn't matter how much hardware is available. 6 cores is pushing it for consumer space. Heck, even having 16 available is plenty for running 50ish general purpose vms. There are few problems that almost all consumer users address that are embarassingly parallel (the types of problems you're speaking of). Few people spend all day rendering 3d images, cracking passwords, encoding video or the like.

edit: To be clear, the types of problems you are talking about approach 100% parallel portions in terms of Amdahl's law. It's not "broken" at all.