Core i7-4960X "Ivy Bridge-E" Roughly 10% Faster than i7-3970X: Early Tests

[EarthDog]

Well i have known both of my i5 and i7 idle around 72w-77w. But no were near 100w for either of them idle.

You gotta be careful on how this is tested too as some mobo takes a load more power even at idle.. My Asus Maximus (x38) used to run 190w idle were with another board around 100w with the same chip.



It is about 4w when idle and that's if it disable HT or not compering my 2 chips..

Im more than certain that is your entire SYSTEM idling at that wattage. Its what I idle at with a 3770K at stock with power saving features on.

Were you measuring the 8-pin EPS connector to get those numbers? My rig idles at 200-watts but that doesn't mean the CPU is idling at that. According to Cadaveca's review of my board, the VRMs use very little power when the CPU is idle, so the majority of that must be my video cards and hard drives.

This is a good read: http://www.linuxfordevices.com/c/a/...PU-power-consumption-a-challenge-authors-say/

HDD's are NOTHING at idle (or when spun up for that matter, several watts). Your GPUs however, compared to the 7 series, dont drop to a 3W idle state, so I would imagine its that, the mobo itself, and the CPU

 

[LAN_deRf_HA]

It is about 4w when idle and that's if it disable HT or not compering my 2 chips..

I meant load. Ivy didn't do much for power savings, certainly less than what the TDP implies. http://media.bestofmicro.com/B/A/334774/original/average%20power.png

 

[Delta6326]

My opinion is totally different. ANY modern cpu will probably last as long, if not even longer because applications aren't getting much more demanding.

I mean how much can a game demand before any extra CPU is totally irrelevant? If you can map the game world at 60fps, maybe throw in a few cycles for AI and stuff that's in-motion, that's about all it's ever going to need. Everything else goes to the GPU.

Office apps, even more so. I can only imagine how many trillions of clock cycles are wasted while Word waits for your next keystroke. The fact of the matter is CPUs now are more than good enough for what we need, and Intel's direction in optimizing toward greater integration and lower power (as opposed to more outright computing power) is totally justified.

I agree with you any current or next gen CPU can realistically last for ages(4-8years) My Q6600 is still rocking. I'm a very light gamer and people need to realize that Intel is thinking about the other 90% of it's sells, Energy is a key factor now.

Side note jihadjoe what volts are you running on your Q6600?, I've been thinking of OC'ing mine, it should last me to the end of this year, then I will go Haswell then wait another 5years+

 

[NeoXF]

I agree with you any current or next gen CPU can realistically last for ages(4-8years) My Q6600 is still rocking. I'm a very light gamer and people need to realize that Intel is thinking about the other 90% of it's sells, Energy is a key factor now.

Side note jihadjoe what volts are you running on your Q6600?, I've been thinking of OC'ing mine, it should last me to the end of this year, then I will go Haswell then wait another 5years+

If you held out this "long", might as well wait for mainstream Intel hexacores (or AMD 5-module+) or at least DDR4... or see how A12 Kaveri turn out...

Otherwise I'd shrug thinking physically nothing consistent has changed from such an upgrade, just IPC, new instructions and added HyperThreading.


Then again, Q6600 to Haswell i5/i7 is a 5-6 generation jump... which then AGAIN, makes it even more sad that so little has changed.

 

[Nordic]

sad that so little has changed.

Change for change sake is pointless. What do you think has needs changing.

 

[de.das.dude]

if i take a guess, it will be 10% faster but should cost 50% more?

 

[Aquinus]

HDD's are NOTHING at idle (or when spun up for that matter, several watts). Your GPUs however, compared to the 7 series, dont drop to a 3W idle state, so I would imagine its that, the mobo itself, and the CPU

Last I checked the 6950 wasn't a 7-series card either. We're talking about ASRock's computer that has a 6000-series card in it, not your with your power sipping 7-series card.

The point is that despite the numbers being low, that's not all the CPU and there are other components that use power and even if you say those numbers are low, low numbers add up pretty quick when you're usage isn't a whole lot to begin with.

A hard drive consuming 7 watts on a machine that draws 200 like mine is nothing, but on a computer that draws 70 watts at idle (I'm assuming the drives aren't spinning down,) that 7 watts just went from being well under 4% of your power usage to 10% of your consumed power. So the ratio of those smaller usages impact you more because the number is already so low.

Also spinning up and slowing down drives a lot on a regular basis actually puts more stress on the motor in the drive. I've had the best luck leaving drives spun up because I'm perfectly willing to use the extra 25 watts to do it. (I'm rounding, I suspect 7200 RPM drives use more then 5400 ones like where i got the numbers from.)

The only real point I'm trying to make is the lower the draw, the more other components can impact that usage, such as add some hard drives or adding a PCI-E expansion card.

We already know how low it idles and that isn't in dispute, it's just the method.

 

[NeoXF]

Change for change sake is pointless. What do you think has needs changing.

What sake of change? We're busting quad-cores since 2006, within the same or slightly higher GHz range with almost the same tired instruction sets, with minimal IPC increase (if any) from generation to generation.

Meanwhile, things like ARM are catching up to x86 like fungus on a 3rd world gym ceiling.

Look at AMD's APU/heterogeneous initiative, at least on paper, it sounds like a huge f ing breakthrough, way bigger than gigahurtz wars and their diminishing returns or many-cores processing from awhile ago.

 

[Octavean]

What sake of change? We're busting quad-cores since 2006, within the same or slightly higher GHz range with almost the same tired instruction sets, with minimal IPC increase (if any) from generation to generation.

Meanwhile, things like ARM are catching up to x86 like fungus on a 3rd world gym ceiling.

Look at AMD's APU/heterogeneous initiative, at least on paper, it sounds like a huge f ing breakthrough, way bigger than gigahurtz wars and their diminishing returns or many-cores processing from awhile ago.

It's a phenomenon of convergence whereby Intel is attempting to increase power efficiency in order to move in to the same segment of the market as ARM and ARM is attempting to improve its performance in order to move into the segment of the market traditionally dominated by more powerful x84 / x64 Intel / AMD solutions.

 

[TheoneandonlyMrK]

Once your into this chip and a decent motherboard territory ($$££€€)its more about how high it will oc with crazy cooling imho that and the max is min crowd.

 

[Hilux SSRG]

Intel is not catering to people like me who don't give a fig about power efficiency. Like I and other gaming/tech users really care about spending $20 bucks more on electricity costs p/year.

I really want an eight [8] core, sixteen [16] hyper-threading beast of a x86 cpu running at 8k-10k gigahertz stock speed.

Forget the 1-10% power effeciency gains per "upgrade" cycle. These are not mobile chips Intel !!

Waiting forever to upgrade from my i7-920 and Intel has not moved the bar much in 5+ years.

 

[EarthDog]

Your bar... is high. LOL!

 

[Fourstaff]

Waiting forever to upgrade from my i7-920 and Intel has not moved the bar much in 5+ years.

3930K is quite a lot more powerful than 920

 

[Aquinus]

3930K is quite a lot more powerful than 920

+1: Even the 3820 isn't a bad step up from the 920 in terms of performance. It's not mindboggling but there is an IPC improvement and it also runs at higher clocks and has a more powerful IMC. In general I would say SB-E was a decent step up from skt1366 because in general it's faster across the board regardless what CPU you get compared to the 920, but it's not necessarily a reason to upgrade. It's a good platform if you're upgrading anyways imho but that depends on what you're using it for.

 

[Tatty_Two]

Sandy Bridge, in 2010, was a huge jump in performance, and I think it has skewed everyone's expectations.
Intel has been squeezing optimizations and modifications of the same basic architecture for quite a few years now. It should not be surprising that the performance bump is now in the 5-10% range for each new iteration. Especially when taking into consideration the problems encountered when shrinking the design to smaller and smaller process nodes.

No it was'nt..... look at the clock for clock performance as opposed to stock clocks which will give you an accurate picture of the architecture advancement, I think you will find across the board we would be talking 10-12%.

 

[Hilux SSRG]

3930K is quite a lot more powerful than 920

Now I know I should "try" to find a comparison between the 920 and the 3930k [if one exists!] but let's say its 20-30% faster overall. I'm wondering why anyone should shell out $500+ for a new mobo and chip for just 20-30% improvement in 5+ YEARS.

I'm not looking to start an arguement but am willing to say intel has coasted for a few years now. I'd rather spend the money on a better gpu from amd or nvidia.

 

[HumanSmoke]

Now I know I should "try" to find a comparison between the 920 and the 3930k [if one exists!] but let's say its 20-30% faster overall.

Then again, you could fire up the calculator...
45% better in Mental Ray, 44% better in V-Ray, 39% better in Visual Studio......

Depending on the users intended workload, it may (or may not) look advantageous - and that's the whole platform I mean (X79 vs X58)

 

[Tatty_Two]

Then again, you could fire up the calculator...
45% better in Mental Ray, 44% better in V-Ray, 39% better in Visual Studio......

Depending on the users intended workload, it may (or may not) look advantageous - and that's the whole platform I mean (X79 vs X58)

But as I said earlier, those differences are not a reflection of huge architectural improvements, when you actually look at the 2 offerings and see that the 920 stocks at 2.66Gig and had 4 cores/8 threads, the 3930 stocks at 3.2gig and has 6 cores/12 threads, it leaves me with the feeling that there is a scary amount of hype and very little real substance to the architecure, but then again I am a 930 owner so i perhaps would say that!

 

[Ikaruga]

There is a test on a Chinese page which claims almost no performance gain compared to Ivy Bridge. I hope it's wrong.

 

[radrok]

But as I said earlier, those differences are not a reflection of huge architectural improvements, when you actually look at the 2 offerings and see that the 920 stocks at 2.66Gig and had 4 cores/8 threads, the 3930 stocks at 3.2gig and has 6 cores/12 threads, it leaves me with the feeling that there is a scary amount of hype and very little real substance to the architecure, but then again I am a 930 owner so i perhaps would say that!

You got to count in that the 3930K overclocks way more than a Nehalem chip.

My old i7 920 couldn't do more than 4,2Ghz.

My 3930K does 5,1-5,2Ghz.

 

[Aquinus]

My 3930K does 5,1-5,2Ghz.

That's all your motherboard. My P9X79 Deluxe doesn't like 4.5Ghz or higher without some serious bumps in voltage (>1.425v). 4.92Ghz @ 1.51v was the highest I was ever able to achieve with this board and I've seen people use the RIVE to hit >5Ghz with the same CPU and similar voltages.

Not to say that I'm not happy with my 4.37Ghz under 1.4v, but it really depends on the motherboard. Unless I'm missing some important settings on my motherboard but I don't think that is the case.

You forget that SB-E has a better IPC than its 1366 predecessors, not by a lot but it starts counting more and more the higher the CPU clock goes because IPC scales linearly.

But as I said earlier, those differences are not a reflection of huge architectural improvements, when you actually look at the 2 offerings and see that the 920 stocks at 2.66Gig and had 4 cores/8 threads, the 3930 stocks at 3.2gig and has 6 cores/12 threads, it leaves me with the feeling that there is a scary amount of hype and very little real substance to the architecure, but then again I am a 930 owner so i perhaps would say that!

It was my impression that even the 3820 was a sizable improvement from a 920, forget a 3930k.

 

[HumanSmoke]

There is a test on a Chinese page which claims almost no performance gain compared to Ivy Bridge. I hope it's wrong.

They appear to be graphics (and GPU dependant) benchmarks, so hardly surprising.
Of more interest would be 5GHz at 0.9v

 

[radrok]

That's all your motherboard. My P9X79 Deluxe doesn't like 4.5Ghz or higher without some serious bumps in voltage (>1.425v). 4.92Ghz @ 1.51v was the highest I was ever able to achieve with this board and I've seen people use the RIVE to hit >5Ghz with the same CPU and similar voltages.

Not to say that I'm not happy with my 4.37Ghz under 1.4v, but it really depends on the motherboard. Unless I'm missing some important settings on my motherboard but I don't think that is the case.

You forget that SB-E has a better IPC than its 1366 predecessors, not by a lot but it starts counting more and more the higher the CPU clock goes because IPC scales linearly.



It was my impression that even the 3820 was a sizable improvement from a 920, forget a 3930k.

I've had quite a lot to tweak to reach those clocks stable.

I think your motherboard has the same settings hidden in somewhere but I won't deny it's easier to overclock on a RIVE than with another kind of motherboard.

On top of my head I had to set LLC for both VCCSA and CPU to Ultra, had to fiddle with CPU current capability, CPU switching frequency and some strange settings that are on BIOS that I never heard or remember of...

The voltage slope from 4,7/4,8 Ghz to 5,1/5,2Ghz is insane though, we are talking from a comfy 1,35v to 1,52-1,55v.

I wouldn't be surprised to see my motherboard VRM toast someday even though they are watercooled but the backside is not and you can't touch that backplate without getting almost burned.


Speaking of HEDT platform I wish Intel would BCLK unlock the upcoming Haswell-E Xeons to have back some glory like X58 westmeres had.

 

[TheoneandonlyMrK]

They appear to be graphics (and GPU dependant) benchmarks, so hardly surprising.
Of more interest would be 5GHz at 0.9v

Er didn't intel focus mostly on gpu improvements this time round and claimed a 50% improvement in that area though , I think given intels cpu ipc performance is bordering optimal already they are doing well but clearly they are as they should be, focused on power efficiency and gpu grunt.

 

[D007]

I'm glad I got my 960 when it came out..
Has been a little monster for me and a hell of a boost from a dual core.
This however doesn't make me want to upgrade at all..lol