TomsCore i7-4960X Preview: Ivy Bridge-E, Benchmarked

[Aikouka]

It's pretty sad that Intel's CPUs are so uninteresting these days that people care more about what Intel puts between the die and the heat spreader than about the actual multi-billion transistor chip itself...

I think that's really only the case for the enthusiast sector. The E-series chips are always a generation behind when it comes to the architecture, so what's there to really be surprised about? It's Ivy Bridge without an integrated GPU, but with more PCI-E lanes and quad-channel DDR3. There are more cores available too, but that's more of the same.

I guess it's just odd to clamor so much about the latest architecture that it's odd to cheer so much for what was hot last year. :|

 

[cytg111]

But the E's are not used for datacenters or supercomputergrids, you are not talking about 1000's of dies saving 30watts each - forecasting weather or nuclear bombs, you are talking about your upper end workstations and given the total output at the plug, i'd dare say that 30 watts is of no consequence.

 

[moonbogg]

All about OC. The new chip runs cooler, so it might turn out well. I'm not dropping a few hundred for another 200mhz though. PASS.

 

[sm625]

This is absolutely terrible. This is the kind of thing that screams "soon to be bankrupt". How can they mess up a cpu so badly that a minute-long operation takes 20 whole seconds longer. The fact that they are unable to profile these types of yawning performance chasms and actually fix them is appalling.

 

[crashtech]

Here comes the cherry-picking.

Overall, I see the performance of this chip is basically as expected. The only possibly exciting part is yet to come, when we see how IB-E overclocks.

It is still the overall fastest consumer level CPU on earth, even without OC, yet the whining will be unceasing as always.

 

[AdamK47]

I'll get it.

 

[Makaveli]

I'll get it.

Wouldn't expect anything less :biggrin:

Now someone in the FS forum will get a good deal on your current chip!

 

[crashtech]

I'd like to see if a few refreshed X79 boards come out with added features. I also wonder if an "old" X79 board will boot with these new CPUs, or if a BIOS update is mandatory using the older CPU to boot.

 

[exar333]

But the E's are not used for datacenters or supercomputergrids, you are not talking about 1000's of dies saving 30watts each - forecasting weather or nuclear bombs, you are talking about your upper end workstations and given the total output at the plug, i'd dare say that 30 watts is of no consequence.

These are server CPUs first and foremost. So 30% more efficient is HUGE.

 

[WhoBeDaPlaya]

I care about the overclocking results. Although this time I will pay a lot of attention to whether the reviewers have an "engineering" chip or a retail one. The shenanigans played last time with SB-E were downright dishonest for overclocking.

Didn't we get screwed with Haswell engineering vs. retail as well?

 

[ruhtraeel]

Hey, I'm not the only one thinking about the OC performance!

Seems like because they are sticking to solder again, it would be the first good upgrade since SB.

It sucked how IB ran hotter with less OC and Haswell ran even hotter than IB with even less OC and dishonesty in samples etc.

Maybe this will give incentive for people before SB to actually upgrade to the latest generation and not one that's 2 generations old but still does just as well for much less cash.

 

[Concillian]

It's pretty sad that Intel's CPUs are so uninteresting these days that people care more about what Intel puts between the die and the heat spreader than about the actual multi-billion transistor chip itself...

Especially when it's so easy to just pop the heat spreader off and put whatever you want in there. It's wierd why this is such a big deal. To an enthusiast this kind of thing should be considered par for the course.

 

[Idontcare]

Ivy Bridge was never meant to improve performance. It's basically a die-shrink of Sandy Bridge, but because of TIM instead of solder we never saw the benefit you normally associate with a die shrink (lower temps, higher OC). If IB-E uses solder that might finally allow the 22nm process to shrine, allowing for higher overclocks than Sandy Bridge.

Unfortunately while I agree with your sentiments regarding "that which one would normally associate with a die shrink", I disagree with your sentiments of what solder will add to the equation in terms of temperatures and overclocks.

As we have seen countless times with delidded IB's, while the TIM-paste and resultant IHS gap does create poor thermal performance, eliminating that gap and replacing the TIM-paste with the equivalent of solder only merely results in bringing IB's thermal performance and OC potential in-line with that of 32nm SB.

^ Once delidded and TIM replaced with liquid metal, the IB has lower power consumption at any given clockspeed, but no better OC potential than that of a 32nm SB.

^ Once delidded and TIM replaced with liquid metal, the IB has only slightly lower temperatures under load at any given clockspeed relative to that of a 32nm SB.

In general all that delidding does for the IB is get it back to the point where an enthusiast can push it to the same place that they can push a SB, but not any further. (and in fact there appears to be a slight regression in technical capability of the 22nm versus 32nm as we see both IB and HW tend to not reach the same OC clockspeeds that 32nm was capable of achieving, 4.9GHz was the absolute limit for my delidded IB whereas 5GHz was easily in reach for my SB)

And we wouldn't expect a soldered IBE to be any different in that regard when compared to a soldered SBE.

In this case all that we (the end-consumer) really got out of the die shrink was a lower power-consumption profile. Intel got a smaller chip that was cheaper to produce and has better margins.

I personally suspect that Intel is grappling with a situation with their 14nm where the performance curve has actually regressed even further, where hitting 4GHz (at any power-consumption level and operating temperature) is currently unyieldable. But that is just my personal opinion and speculation.

 

[CHADBOGA]

Another marginal product which fails to inspire.

What a sad time for people who consider themselves Desktop Users, primarily. D:

 

[crashtech]

Another marginal product which fails to inspire.

What a sad time for people who consider themselves Desktop Users, primarily. D:

You mean a product with marginal improvements, right? Because calling the soon-to-be fastest desktop CPU on earth "marginal" does not compute.

It is tough to be inspired by such small gains, but then again those gains are getting tougher to come by all the time. It's as good as can be expected. Let's see how they OC.

 

[tweakboy]

I wonder why Intel bother with these tiny incremental improvements?
I hope they overclock better but I'm increasingly tempted to stick with my 920 until I die at this rate.

I like that post nice one jack

 

[Arachnotronic]

Another marginal product which fails to inspire.

What a sad time for people who consider themselves Desktop Users, primarily. D:

If people actually bought bulky desktops and weren't enamored with iTabs, then we'd get more in this space.

We are a niche market, and we will get the scaled up mobile chip or the crumbs from the server space. It sucks, but such is the way of things.

 

[PeteRoy]

Where is the 8 cores?

Samsung Galaxy S4 already has 8 cores on a mobile phone, why Intel refuses to release 8 cores to consumers?

We should have had 8 cores back in 2009, Intel already showed 8 cores Nehelam.

In 2013 I would expect to have 16 cores in desktops.

 

[pandemonium]

What purpose would 8 or 16 cores serve a personal desktop CPU? Clocks >>>>>> Cores.

 

[joshhedge]

Where is the 8 cores?

Samsung Galaxy S4 already has 8 cores on a mobile phone, why Intel refuses to release 8 cores to consumers?

We should have had 8 cores back in 2009, Intel already showed 8 cores Nehelam.

In 2013 I would expect to have 16 cores in desktops.

Pretty sure only 4 of the cores can be working at any point of time though due to how big.LITTLE works.

 

[Idontcare]

What purpose would 8 or 16 cores serve a personal desktop CPU? Clocks >>>>>> Cores.

More cores would be useful if people actually did stuff with their computers. We don't really do stuff with our home PCs though.

Usually we are using them as rather expensive entertainment systems, and that can be done with just a few fast cores, an internet connection, and a decent graphical output device.

More cores is great if you are really into multi-tasking low-threaded apps that are doing stuff in the background. Gaming while video transcoding in the background is an example of that.

But how many people realistically try and load up their computer to do as much as possible in a finite period of time?

It just isn't a common user scenario with today's entertainment-seeking consumer.

 

[pandemonium]

That's sort-of the point I was making. In the PC experience, this isn't common. So the market for more cores isn't necessary. Not only that, but threading isn't handled intuitively to utilize a growing core personal computing industry.

If you're into multi-core tasking, you're better off just buying a server platform IMHO.

 

[Idontcare]

If you're into multi-core tasking, you're better off just buying a server platform IMHO.

Yes.

But...server platform is silly more expensive than consumer platform, it destroys its own performance/price justification

BP6 FTW.

 

[pandemonium]

Oh, ABIT. I miss you.

 

[Gikaseixas]

Because its targetted at the workstation and server segment with completely different demands than an "enthutiast".

So if one buy one of these, the purpose will be to build a server? Not a gaming machine?
Stop defending these smallish improvements just because it's Intel. PC advancements have been piss poor on the performance front. Efficiency improvements are great for most of the Intel lineup but performance for desktops are not that great.