Intel 4960X, 4930K, 4820K Unboxing

[LinusTech]

The 4960X unboxing is up. Pretty much no surprises given how concrete the details usually are leading up to an Intel launch...

 

 

And now that we know exactly how a given enthusiast grade architecture is going to perform because we had it a year prior on the mainstream platform, the surprises are even less surprising.

 

At any rate if you want to buy one here are a couple of links:

 

4930K (the one I can honestly recommend if you really need more cores)

NCIX: http://ncix.com/search/?categoryid=0&q=4930k

Amazon: http://geni.us/iKVdS

 

4960X (if you really want the bragging rights)

NCIX: http://ncix.com/search/?categoryid=0&q=4960x

Amazon: http://geni.us/wtBhF

 

4820K (if you REALLY need more RAM or PCIe lanes, but don't need more cores)

NCIX: http://ncix.com/search/?categoryid=0&q=4820k

Amazon: http://geni.us/eX9Q4l

[CoolBeans]

This video is private

Also 3960X isn't in IvyE line up

[Tall Gora]

The 4960X unboxing is up. Pretty much no surprises given how concrete the details usually are leading up to an Intel launch...

 

 

And now that we know exactly how a given enthusiast grade architecture is going to perform because we had it a year prior on the mainstream platform, the surprises are even less surprising.

 

At any rate if you want to buy one here are a couple of links:

 

4930K (the one I can honestly recommend if you really need more cores)

NCIX: http://ncix.com/search/?categoryid=0&q=4930k

Amazon: http://amzn.to/17L2p1F

 

4960X (if you really want the bragging rights)

NCIX: http://ncix.com/search/?categoryid=0&q=4960x

Amazon: http://amzn.to/15Ncrid

 

4820K (if you REALLY need more RAM or PCIe lanes, but don't need more cores)

NCIX: http://ncix.com/search/?categoryid=0&q=4820k

Amazon: http://amzn.to/1aW463x

no surprises

[qwertywarrior]

[AraSarkisyan]

finally. when will we see some benchmarks? Also i wonder if these still bottleneck titans in SLI

[Maskot]

Bottleneck is such a stupid term.

 

In terms of gaming performance they won't be too far ahead of a 4770k anyway

They will be once games start using more than 4 cores.

[Dravic]

Bottleneck is such a stupid term.

 

In terms of gaming performance they won't be too far ahead of a 4770k anyway

 

They will be 1 step  behind 4770k actually except in games that use additional cores.

[AraSarkisyan]

Bottleneck is such a stupid term.

 

In terms of gaming performance they won't be too far ahead of a 4770k anyway

it's what came to my head at that moment, m'kay? 

[Dravic]

They won't be using any more than 8 threads for a long time yet. So for gaming 4770k is still the pick of CPU's for best performance.

 

 

 

 

Again, games wont use more than 8 threads 

 

Threads =/= Cores. Cores =/= Threads.

 

6-core 4930k will be MUCH faster running 8 threads than 4-core 4770k, because only 2 threads will be run on shared Hyper threading if anything.

 

While on 4770k all 4 additional threads will be virtual and that is much slower than real core.

[Hirad123]

Bottleneck is such a stupid term.

 

In terms of gaming performance they won't be too far ahead of a 4770k anyway

In elric's test the 4770k actually out performed it in some games. No surprises though.

[Dravic]

The difference in single threaded performance between a hyper threaded core and an actual core is almost zero

 

You are so wrong... I just... I don't even...

 

There's a reason why Windows had to be patched up to do core parking for i7s...

[qwertywarrior]

[WoodenMarker]

Have a look at a 4670k vs a 4770k in either Cinebench's single core bechmarks or Passmarks signle core benchmarks, or any single threaded benchmark for that matter.

 

There really is a reason why in every single thread you go on you get called out as the most unreliable source on this forum and quite frankly I'm sick of your crap, stop spreading BS around this forum and stop pretending you know what your talking about, because you very clearly don't

Actually, I agree with @Dravic on this one.

The single core benchmarks bench the single core and not all at once. When all the virtual HT cores are being benched, the threads have to share the resources for what would have been a single core without HT. 

There is a difference in performance between a hyperthreaded thread and a normal thread. A normal thread has access to all of the core's resources while 2 virtual cores would have to share. 

This is why an i7 doesn't make a full 100% increase in performance for rendering over the i5.

[Dravic]

Have a look at a 4670k vs a 4770k in either Cinebench's single core bechmarks or Passmarks signle core benchmarks, or any single threaded benchmark for that matter.

 

There really is a reason why in every single thread you go on you get called out as the most unreliable source on this forum and quite frankly I'm sick of your crap, stop spreading BS around this forum and stop pretending you know what your talking about, because you very clearly don't

 

 

Yes exactly qwerty.

 

I can also add this:

 

 

Check out the gains from FX-63xx -> FX-83xx, and compare them to the gains of i5->i7...

[target39]

You are so wrong... I just... I don't even...

 

There's a reason why Windows had to be patched up to do core parking for i7s...

 

 

Have a look at a 4670k vs a 4770k in either Cinebench's single core bechmarks or Passmarks signle core benchmarks, or any single threaded benchmark for that matter.

 

There really is a reason why in every single thread you go on you get called out as the most unreliable source on this forum and quite frankly I'm sick of your crap, stop spreading BS around this forum and stop pretending you know what your talking about, because you very clearly don't

Please keep it classy. I don't think Linus would be very happy if he had to end up deleting both of your posts and giving warning points and/or locking his own thread. I know someone who was just given two warning points because of a disagreement over a hypothetical question. I don't think either of you would want this to happen to yourselves.

[WoodenMarker]

In a hyperthreaded situation a single core is getting fed twice which utilizes the clock cycles that cannot be fed in a normal scenario.

 

http://www.anandtech.com/show/7003/the-haswell-review-intel-core-i74770k-i54560k-tested/6

 

As you can see there the single threaded performance of a 4770k is actually better than a 4670k

You just showed what we said. There's a difference between a virtual core and a physical core.

[qwertywarrior]

Yes exactly qwerty.

 

I can also add this:

 

 

 

Check out the gains from FX-63xx -> FX-83xx, and compare them to the gains of i5->i7...

huh i thought u guys were arguing about HT

[WoodenMarker]

My point was that a hyperthreaded core isn't worse than a single core, which is what those benchmarks show.

 

In a case where only 1 core is being utilized then a single hyperthreaded core may outperform a traditional core seeing as it is fed twice. A point I didn't bring up earlier, but still I guess is a point towards hyperthreading tech

A hyperthreaded core is no different than a physical core. The threads are just different. In the case you showed, a single thread from a single core will perform better than a thread from a hyperthreaded core.

Cinebench didn't  measure how a single thread fared when all threads were on load--it measured how it fared on it's own so the results are from a thread uninhibited by another thread hogging it's resources.

 

Edits: Of course a core with hyperthreading would perform better than the same core without HT. Otherwise, HT would be pointless.

The problem is that you said a hyperthreaded core instead of a core with HT.

One refers to a virtual core (a thread) and the other refers to a physical core.

[M-ursu]

This video is private

image.jpg

Also 3960X isn't in IvyE line up

[Dravic]

My point was that a hyperthreaded core isn't worse than a single core, which is what those benchmarks show.

 

In a case where only 1 core is being utilized then a single hyperthreaded core may outperform a traditional core seeing as it is fed twice. A point I didn't bring up earlier, but still I guess is a point towards hyperthreading tech

 

A hyperthreaded core is no different than a physical core. The threads are just different. In the case you showed, a single thread from a single core will perform better than a thread from a hyperthreaded core.

Cinebench didn't  measure how a single thread fared when all threads were on load--it measured how it fared on it's own so the results are from a thread uninhibited by another thread hogging it's resources.

 

Of course a core with hyperthreading would perform better than the same core without HT. Otherwise, HT would be pointless.

 

I never said hyperthreading is bad. But you try to argue that hyperthreaded thread is as fast as non-hyperthreaded core. It isn't.

 

That's why we can't see 100% scaling from 4 to 8 threads (i5 vs i7).

 

Also i7 is faster single-thread than i5 because i7 has 2mb more l3 cache which makes it that tiny bit more suited towards long calculations as it can save 2mb of data more in l3 cache which is a lot faster than RAM.

[Reuben]

They will be once games start using more than 4 cores.

A 4770K has 8 threads, and even a 4670K has such powerful cores multi-core optimization doesn't really matter.

[Yoinkerman]

I wish they'd just start programming for n threads, where n > 1.

 

That way more is better!

[wpirobotbuilder]

I wish they'd just start programming for n threads, where n > 1.

 

That way more is better!

There are very few things you can program for an arbitrary number of threads, and a lot of things can't be programmed for multiple threads.