Review Index:

Intel Sandy Bridge-E Review - Core i7-3960X and X79 Chipset Tested

Author: Ryan Shrout
Subject: Processors
Manufacturer: Intel

The Processor Lineup, Memory Kits and Cooling Considerations

For the initial release, there will only be a pair of Sandy Bridge-E processors available for sale; a third will be coming early in 2012 according to Intel's schedule.

View Full Size

The Core i7-3960X is the new top-of-the-line Extreme Edition processor coming in at that classic $990 price point.  It has a base clock speed of 3.3 GHz with a maximum Turbo speed of 3.9 GHz when utilizing only 1 or 2 of the processing cores.  The CPU has 6 cores and supports HyperThreading for a total of 12 addressable threads along with the full 15MB of shared L3 cache.  

Just under it is the Core i7-3930K will be priced at $555 and run just about 100 MHz slower than the i7-3960X across the board.  It will still have 6 cores and 12 threads of processing but lowers the L3 cache to 12MB.  Both of these CPUs have the quad-channel memory controller with support for DDR3-1600 memory and have 130 watt TDPs. 

Missing in action today but due in Q1 of 2012 is the Core i7-3820 that runs at a base clock of 3.6 GHz and the top Turbo speed of 3.9 GHz.  This part is a quad-core / 8 thread processor with 10MB of L3 cache and the same 130 watt TDP.  While the price isn't known yet, I would hazard a guess of a ~$300 MSRP.  What is interesting here is that I really believe that THIS Sandy Bridge-E CPU would have easily been the best selling day one - most gamers can't afford the $990 item and will hesitate to upgrade even at the $555 level.  The Core i7-3820 could be the new Core i7-920, the most popular processor of its generation.

View Full Size

Left: Core i7-2600k (Sandy Bridge)  Center: Core i7-975 (Nehalem)  Right: Core i7-3960X (Sandy Bridge-E)

Let's compare the Core i7-2700K- the fastest Sandy Bridge processor using the LGA1155 socket, to the Sandy Bridge-E SKUs we see above.  With a price at just $332, a base clock speed of 3.5 GHz and a top Turbo frequency of 3.9 GHz, the quad-core Hyper Threaded processor looks like it might be very competitive with the Core i7-3820 in most applications.  Intel is likely thinking that the i7-3820 would cannibalize the sales of the i7-2600 and i7-2700 parts and wants to give them as much time as possible on the market (alone). 

And, as we might see in our benchmarks, the benchmarks and software that only utilize a single thread (or are lightly threaded), could run just as fast on the higher priced Sandy Bridge-E CPUs.

With the retail release of the new Core i7-3960X, there are some other changes to the plan from Intel.  For one, the processor will no longer come with a heatsink in the box of any kind.  Instead, Intel has been teasing an Asetek-built self contained water cooler that will be available for use with LGA2011 CPUs as well as LGA1366 and LGA1156/1155.

View Full Size

This is the cooler we used in our processor testing over the last couple of weeks and have been pretty impressed with both the performance and the sound levels that the cooler generates.  With 130 watt TDP processors (even higher when overclocked) having these improved coolers is a big plus though I can't shake the feeling that Intel is screwing people here by not including SOMETHING in the box with the processor.  Now, if you want that $555 processor you should expect to spend another $80-120 on a water cooler from either Intel, Corsair or someone else.  

View Full Size

Even more odd, is that Intel has built and is going to sell an OEM cooler that looks very similar to previous designs that works on the LGA2011 socket, seen above.  Priced at around $20 this cooler will likely not be a big seller except in workstation and server markets.  This slide also mentioned Ivy Bridge-E for the first time - ready to get excited for that yet??

Along with the changes in the retail configuration of the processors, memory packaged for Sandy Bridge-E also gets revamped as companies like Corsair, Kingston, G.Skill and everyone else start to offer quad-channel memory kits with four DIMMs. 

View Full Size

Memory is dirt cheap and we love it - we are going to start seeing 16GB of memory as the norm!

Testing Configuration

For our testing we did the majority of the benchmarking on Intel's DX79SI and the board was very impressive in terms of its stability and flexibility.  Because of the new platform and memory configurations our hardware setup does differ a bit from previous reviews.  For example, we are using 8GB of memory on our Core i7-3960X CPU (4 x 2GB modules) to keep the capacity close to the 6GB used with Nehalem (3 x 2GB) and 4GB on Sandy Bridge and others (2 x 2GB).  

View Full Size

  • Intel Core i7-3960X (Sandy Bridge-E)
  • Intel DX79SI X79 Motherboard
  • Intel water cooler
  • 4 x 2GB DDR-1866 Corsair Vengeance (running at DDR3-1333)
  • Intel X25-M G2 160 GB SSD
  • GeForce GTX 285 Graphics card
  • PC Power and Cooling 1200 watt PSU
  • Windows 7 SP1 64-bit

View Full Size

There are few key comparison we will be keeping an eye on with all of our benchmark data.

  • Core i7-3960X vs Core i7-990X - These are the competing "world's fastest CPUs" and we are interested to see how the 6-cores of Sandy Bridge-E compare to the 6-cores of Nehalem.
  • Core i7-3960X vs Core i-2600k - Both are based on the same CPU architecture but the new Core i7-3960X has 50% more cores with 100% more memory channels and we want to know how much performance you get with all those changes. 
  • Core i7-3960X vs AMD FX-8150 - The best from Intel against the best from AMD; we pretty much know how this is going to pan out but many of you will want to see it for yourself. 

I did want to let everyone know that we had intended to test the performance levels of the much lower priced Core i7-3930K processor today but ran into a bug with our normal way of accomplishing it.  Intel only sent along the Core i7-3960X but we planned on utilizing the multiplier adjustments in the BIOS to run at the lower clock speeds.  The problem lied in the inability for the BIOS on either the Intel DX79SI or other retail options to properly set the 6-core and 5-core Turbo multipliers to 35x.  Instead, they continued to run at 36x and thus at an incorrect speed based on the steppings the i7-3930K retail will run at.

View Full Size

If you are reading other reviews that claim to have Core i7-3930K results, be sure they had an actual sample of that CPU as we think performance will be skewed without it.  We are planning on getting a retail version of the Core i7-3930K this week to update you!!

November 15, 2011 | 06:06 AM - Posted by Sihastru

They did that already (and other sites too), 2500K and 2600K vs. Bulldozer, it lost there too.

The idea of the article here is to compare the best offerings from the main (only?) two competitors on the market. When we're in this segment, we're interested about performance more and less about the price. We don't really care about performance/monetary unit, we just want the best there is, period.

November 16, 2011 | 01:34 PM - Posted by AParsh335i (not verified)

Ryan you seem to be attracting a lot of trolls on this review.

November 16, 2011 | 01:58 PM - Posted by Ryan Shrout

I have always gone under the assumption that:

More Trolls = More Correct in my Views

November 15, 2011 | 08:49 PM - Posted by J.P. (not verified)

Why does the i5 2500K perform better then the i7 2600k on so many of these tests am I missing something. Is there other reasons that I would spend $100 more one a CPU not to mention would the 2700K be worth the $30 more than the 2600K. I was waiting for the SB-e processor but I just cant justify the price/performance BIGTIME FAIL on Intel, just wasted my time waiting for a product that barely outperforms I7 990X.

Is LGA2011 backwards compatible with the I7-2600?

November 16, 2011 | 01:33 PM - Posted by AParsh335i (not verified)

LGA2011 is completely different than LGA1155 (AKA I7-2600 you inquired about).

Seems a lot of people are mad about this...Ryan has said it like 10 times - this is not for most of "us." This is for a small market of power users, not the people that want the best frame rates on Battlefield 3.

November 16, 2011 | 02:01 PM - Posted by Ryan Shrout

First, as to why the 2500K might outperform the 2600K in some select instances: because a four-threaded app has 4 dedicated cores on the 2500K and though there ARE 4 dedicated cores on the 2600K, it is possible that the OS doesn't put each thread on its own core and instead you see a bit of HyperThreading i the mix, which isn't nearly as efficient as single core computing. That being said, usually the OS figures this out pretty quickly, so differences are actually minimal.

As for the 2700K - I would actually recommend it over the 2600K if that was your choice before, yes.

November 16, 2011 | 10:05 PM - Posted by J.P. (not verified)

Thanks for your response Ryan, if I hear you right the I5 doesn't have HT. so turning off HT one the 2600K or 2700K may improve some of the non-multi-threaded reliant applications. As for your suggesting the 2600K over the 2700K for the $22 difference would I be getting a sightly higher binned chip with better TDP, clock speed, etc.. I guess I am know wounding id the 2500K is the right choice @ more than $100 less( I believe I heard you guys say as much on the Podcast) do I loose any else then HT with the 2500K.

Thanks Again

November 29, 2011 | 09:01 PM - Posted by ftimster

So question?? If you want the fastest gaming pc or just the fastest system go with 3960x and dont wate for ive?? Asus rampage iiii 32gb or 16gb and keep your graphics top of the line. Money is not a problem but ive bridge whats the deal? not going to be all arround better than 3690x dont want to upgrade again i am currently running 990x asus rampage iii formula three gtx 580's Would like to here a really good breakdown on sandy-e vs ivy??? Thanks watch the show allways!!!! Tim in alasks!! :-)

December 5, 2011 | 03:28 PM - Posted by Steve-S (not verified)

Great review Ryan! Have you ever done a benchmark using virtual PCs? It would be interesting to put 32gb of ram in one of these systems and see how many virtual PC's you could effectively run. Throw different loads on each of the machines and see what brings it to knees.

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

By submitting this form, you accept the Mollom privacy policy.