2011 started off great with a new toy for the hardware enthusiasts. Intel's new socket 1155 aka Sandy Bridge counts not less then 29 new CPUs.  Fourteen brand new babies for the desktop platform and 15 for the mobile market. Based on the 32nm architecture Sandy Bridge's technology is targeted at being more performance, better energy efficient and this all at a lower cost. Sounds too good to be true, doesn't it ?

 

 

Intel's roadmap, as usual is based on their infamous Tick/Tock release plan. Sandy Bridge is here to directly replace the older S775 and Clarkdale CPU's. Later on Lynnfield and even the Nehalem architecture will be phased out. Just to give you an idea how fast it all will happen. We were at Gigabyte's Sandy Bridge launch event, there was mentioned for those ( PC builders) that still want to build socket 1156 PC's, should immediately order motherboards now. This as they can't promise any more stock in the near future ( read February )

 

Intel still keeps the 1366 socket as High end/Enthusiast or as they call it performance platform. Now this is of course only the desktop roadmap. The Mobile roadmap is similar with a quick replacement of the Arrandale and Clarksfield lineup.

 

For those that want to know all the current and planned desktop models, here are their specs :

 

Intel chose to stick with the current i3/5 and i7 names. The newly designed logo, with the previous logo spilt open,  is clearly recognisable from the previous generation CPU's.

 

So still the same name branding, now the below picture will give you already a good idea how the lineup is.

A quick word on the  lineup : so the I7 variants will have 4 cores and 4 hyperthreaded cores allowing up to 8 simultaneous tasks. Core i5 and i3 will support up to max 4 threads. Caching wise i5 and i7 will have access to a max of 8Mb ( this differs on the model ) I3 will only be able to take benefit from max 3Mb cache. All CPUs have got an integrated graphics core. The 2 K  models (2500 and 2600K ) will have the 3000HD IGPU onboard. All the rest is equipped with the 2000HD IGPU.

 

The above diagram already gives away one of the biggest features on this particular CPU design. What Intel refers to as Last Level Cache is directly accessible by all the onboard CPU components ( cores, IGP, etc...) via a ring architecure. It's not directly comparable with the ATI design on it's graphics cards. but the concept is sort off the same : reduce cache access delays to the maximum.

We were all used to push the Front Side bus on the older Intel platforms and the Bclock out of the 1366 socket to gain some raw Mhz. This however is over and done with, as Intel integrated the clock generator into the CPU now. Sandy Bridge's bclock is controlling every component now and when anything goes out of sync the sytem will be very unstable. Forget about Bclocks of 240 and beyond. The stock Bclock is 100. Expect a max of 105-107ish for the enthousiast overclocking crowd. Theoretical maximum and highest Bclock spotted in the wild is 110. This puts a big damper on overclocking a low end CPU up to the speeds of it's bigger and normally much more expensive brother. If ever you want ot push to higher limits you have to opt for an unlocked multiplier K model. Comparable with the older Extreme models, yet Intel put them only 20 dollars more then the similar speed non K variant.

So we handled the new ring architecture, the shared cache. Last but not least is the dynamic frequency. Alike the previous S1156 platform, but Turbo Boost is enhanced by Intel. Allowing up to 400mhz more if one core is stressed, 300Mhz for dual core usage, 200Mhz for triple core and 100Mhz for quadcore optimised applications. Below you can see again the big difference between a locked and an unlocked K model.

The new Turbo , baptised Intel Turbo boost 2.0 is very impressive. Quickly it adapts to the load demand, thermal factors etc... If you have Intel's Turbo monitor tool open or Tmonitor you will notice constant speed changes between the cores.

 

Here's a small video of the turbo in action. Stressing with Prime95 v 26.3, first one core, two cores , 4 and finally 8. Look how the cores interact, fluctuate... and look at the CPU-Z CPU speed readout : 37X, 36X, 35X...

 

With every new socket, there is a new chipset design too. Initially the end user will only have the choice between the P67 and H67 chipset. To cut a long story short here is the diagram which clearly shows the main differences.

H67 uses the CPU integrated HD2000 or HD3000 GPU as main display. Another GPU can be added for higher graphics performance into the PCIe 16x slot. But a H67 board does not allow any CPU overclocking via the multiplier. So buying a K CPU is worthless if ever you are gonna hook it up with a H67 board (unless you want the HD3000 GPU performance). Same for the RAM multipliers. The chipset supports up to max 1333Mhz ram speeds. While this might sound like more than enough , further tests will put another spotlight on this matter.

P67 is more versatile. Of course not taking any use of the CPUs Intel HD. But sporting multiple GPU solutions. Multiple ram dividers (800/1066/1333/1600/1866 and even 2400mhz )and best of all it allows CPU multiplier overclocking. 

Later this year the Z68 will be on the shelves, sporting CPU and IGPU overclocking. If it will be the next big thing is doubtful. Also other low end chipsets will pop up in the upcoming quarters, but not very interesting if you are gamer or bencher.

Okay that was a brief summation of the new features of this CPU and chipset, but there's more. New instructions boost the CPUs encoding performance. Intel's AVX, which requires Windows 7 with Service Pack 1, will give a very nice performance boost if the application supports it.

Here's what a 2500K retail sample looks like:

 

 

 

Before we start to look into the benches here's a quick rundown on the test setup :

Asus P8P67 Dlx motherboard (Bios 1053)
Intel i5 2500K and i7 2600K CPU
4Gb Corsair Dominator 1600Mhz ram CL8-8-8-24 1T
Nvidia GTX 285/480 GPU : 260.99 WHQL drivers used
Enermax Galaxy 1Kw DXX PSU
Windows 7 Professional 64Bit

 

 

Synthetic tests used:

3dmark01 and 06
SuperPi 1M and 32M
Wprime 32M and 1024M
AIDA 64 Memory benchmark ( replaced the phased out Everest Ultimate edition )
Cinebench R10 64 Bit
X264HD encoding test

For games we use :

Far Cry 2
Mafia II
Alien vs Predator DirectX11 benchmark
Dirt 2
For easier performance detection we limited the gaming resolutions to 1280 x 1024. This to keep the GPU sort of out of the equation.

 

What have we tested : Stock clock performance, ,all CPU's run with 1600Mhz ram speed, only the i760s rams run at 1333Mhz. (This as it would require to OC the CPU's Bclock to get 1600mhz ram speed )

Due to a multitude of RAM dividers it might be interesting to find the sweet-spot for your needs, so we tested almost each one of them. Then we go a bit deeper and investigate the main RAM timings.

Then a 4ghz show down to show you how efficient or performant the new platform might be compared to your current setup.

Last but not least some overclocking results of OC team BE for the HWbot Country Cup We tested several CPU's and found a one that did 5.4Ghz! Not bad but there are far better ones out there too.

Let's get cracking and see if all these tech innovations pay off.

Now let's see what Intel's latest generation has got to offer. Too bad we couldn't include the S1156 i7 875K CPU. As this is the socket, that 1155 Sandy Bridge will replace. Our 875K ES sample apparently decided to leave us unexpectedly, without a warning nor saying goodbye. As usual, our testing environment is a mix of synthetic and real life benchmarks. Let's get it going :

 

 

Wow, didn't really expect this kind of performance from what is referred as a mainstream platform. Intel's high end platform can't keep up. As a reminder the 2500K with Turbo function enabled can go up as high as 3.7Ghz, the 2600K up to 3.8Ghz. This all depending on the application load, thermal performance etc... For the AMD Thuban CPU we just enabled the Turbo function in the bios. Not pushed higher as some motherboards allow manual turbo intervention. This to keep the stock comparison as fair as possible. Compared to the S1156 Lynfield i760, the new features seems to provide a serious boost.

 

 

The two Sandy Bridge CPUs are SuperPi monsters. Over 100secs faster compared to the fastest 1366 CPU, our humble i970. The competitors just have to throw in the towel. There's no match (yet) for the "Sandies". Wprime is still ruled by the Gulftown CPU, thanks to its 12 thread capability. Maybe with a bit higher Turbo ratio the AMD 1090T could give a good showing here. But that's for later, when we test all CPUs at the same clock speeds. Till now jaw breaking performance of the 1155 platform !

 

We had to stop using Everest Ultimate Edition, as it was no longer updated. If we had used it for this review the graphs woulds be out of the margins of this page. We saw over 38K bandwidth. Luckily it's successor AIDA64 is here to save the day. No need for many words here. The Write, Read and Copy performance are out of this world. Bandwidth galore is not even the most suited expression. Imagine this technology, mixed with the 1366 socket : a lethal cocktail !

 

 

Techarps x264HD benchmark renders a 30 second High Definition clip into an x264 codec clip. Also supporting multithreading technology, making it an ideal benchmark for multi-core CPU's. The results are expressed in frames per second. Compared with the Bloomfield lineup (4 cores plus 4HT threads), even the 2500K ( only 4 cores ) still manages to render at least the same amount of frames per second. If we pop in the 2600K, we see the fastest platform in pass 1, pass2 is again Gulftown territory, benefiting more from the extra threads. But the 2600K retails at only a 3rd of the retail value of an 980X. Cheaper mobo's, dual channel ram solutions... and Intel still calls it a mainstream platform!

 

 

Cinebench R10's results are not different. Top notch performance in the single core thread ( over 300mhz-500 advantage here compared the other CPUs, due to the Turbo function ) But even the multi core rendering is blistering fast AMD's lineup is really getting hammered by Intel. I'm impressed by Sandy Bridge.

 

No this is not an error, the 3dMark01 results are fully correct. Maybe of less use to a daily user. But for the benching crowd this is really good news. The Core2 Duo S775 found its match. There's a new king in town lads for 3DMark01 benching. Even the 3DMark06 performance is quite good, especially looking at the score of the i970. Sandy Bridge seems to look as a solid game/benching platform. Let's take a look at the game tests :

 

 

We dropped the resolution to 1280 x 1024 and set the details to high or even very high for the older games. Why the lower resolution ? Just to be able to show the effect of a faster CPU in the game. If we put in a very high resolution,max detail and max texturing filtering the Graphics Card will become the bottleneck. Securing the top 2 spots, which is no surprise after all these tests.

 

 

 

Since the current Sandy Bridge platform is seriously limped in Bclock OCability, let's explore if we can get some extra performance by playing with the available RAM dividers. The tests on the previous pages were all done with 1600mhz ram speed. Only the 1156 i5 760 was tested at 1333Mhz. On my Asus P67 test board I had the choice between : 800/1066/1333/1600/1866/2133 and 2400mhz. (Though I never got the latter to work). With previous Intel chipsets you had to tinker with the Bclock to get the required high ram speeds. Sandy Bridge makes it all, maybe for some, too easy. Just select the required divider, save and reboot. I only tested the 1066/1333/1600/1866 and 2133 divider with the Corsair GTX2 rams. I skipped the 800mhz DDR3 ram as I don't see it useful, as it über cripples the bandwidth. Neither is this ram daily used, not even in laptops...

 

 

As usual we start off with Superpi 1M and Wprime 32. Both being very fast tests and mostly not clearly showing ram efficiency. Normally raw CPU Mhz rules over timings. But we already seem to detect up a small increase in performance as the ram speeds go up. Time to move onto the big brother superpi 32M.

 

 

Now this is more like it. Over 40 secs shaved off the 1066Mhz Superpi 32m time. Wprime is only gaining a few secs, but this benchmark is less influenced by ram speeds or timings. How big are the reads and writes then to accomplish this huge boost ?

 

 

The message seems clear. Copy and Read performance are not brilliant at 1066 and 1333Mhz. (luckily I didn't test 800mhz speeds) Once we move up to the 1600mhz ram divider, the bandwidth department gets a nice boost. Scaling even better at the higher dividers. This socket has got a monstrous bandwidth when bundled with appropriate rams. It even trounces the triple channel performance of it's bigger S1366 brethren.

Let's do some encoding with the X264HD benchmark. Will the extra bandwith provided, give some more headroom ?

 

 

Again observing scaling with more RAM bandwidth. A few FPS more, is for a serious encoder a time saviour. Give your 1155 CPU what it needs : some nice rams running at at least 1600mhz.

 

Moving onto PCMark05, which tells you a bit more about the complete performance of your system.

 

 

Once ram speeds are at and beyond 1600mhz the performance picks up. Bundling such a high performance processor with rams at or below 1066Mhz is a crime. 1866 and 2133Mhz results are very close.

 

 

3DMark01 has always been the love of the S775 platform. Not even the almighty 1366 could take the performance crown. Intel's latest creation however provides some stiff competition. But that's for later. Observing the collected data, more ram speed reflects in better performance. Doesn't take a mad scientist to analyse what's going on here.

 

 

Even though 3DMark06 is less spectacular in scaling then the 01 version, it still manages to get a few points extra. Synthetic tests are nice, but a few games examples maybe will tell you a bit more.

 

 

Far Cry 2's integrated benchmark was run at 1280 x 1024 resolution. Benefiting big time from more ram speed. Rewarding you with a few extra FPS. Though starting off at 150FPS, which is already bloody fast , I don't think you will notice the difference between the 1066Mhz and 2133Mhz setting in games.

 

 

Mafia II seems to be maxed out on this setup. One FPS difference between 1066Mhz and 2133Mhz... the GTX480 runs the flat out...

 

 

I included the Final Fantasy XIV benchmark as it seems to pop up more and more on several forums. Though the outcome of the benchmark is just a number, no FPS. Scaling is there, though what it means for real FPS. absolutely no clue...

 

From the previous page we can conclude that Sandy Bridge loves high ram speeds. Though can we squeeze some more points or FPS out of it, when messing with the main RAM timings ? Let's start off with 1333Mhz. Yep I again ditched a RAM divider, as the 1066Mhz might be sort off interesting for laptop integrators. For desktop users it shouldn't even be an option to buy...

I used the 2500K on the same bench-suites of the previous pages. Testing the following timings :

For 1333Mhz and 1600mhz : CL6-6-6-18, CL7-7-7-21 and 8-8-8-24 

For 1866 : CL7-7-7-21, CL8-8-8-24 and CL9-9-9-27

For 2133 : CL7-8-7-21, CL8-8-8-24 and CL9-9-9-27

1333Mhz is up first :

Both SuperPi 1M and Wprime 32  tests are very short and more in favour of raw CPU clocks. The differences are really small between Cas 6 and 8.  

Superpi 32M almost gains about 7 seconds. Nothing shabby for a bencher, for a daily user the difference the gain is negligible.

AIDA64's memory benchmark only shows an improvement in the Read test. Write and Copy stay put. Maybe this will change with higher ram clocks.

Encoding a file into HD quality almost gains nothing either from faster timings. Not even a full FPS is won.

 Cinebench improves also a tiny bit with the Cas 6 setting. Nothing earth shattering though

 Finally a nice improvement : PCMark05 seems to like tighter timings. 300 points gain is really nice.

The Synthetic 3DMarks from Futuremark again show slightly better scores with tighter timings. Especially the 3DMark 01 score is improving with over 1000 points.

By limiting the resolution to 1280 x1024 our GTX 480 is CPU limited. The tiny gain observed will be not noticeable when upping resolutions and detail level. For a gamer if 1333Mhz is your weapon of choice , even Cas 8 will do nicely.

1600Mhz, it's your turn ! Show us what you got please

 

SuperPi 1M and Wprime 32 benefit very little from tighter timings, but every hundreds of a second counts for a bencher. For a daily user or abuser this test is pretty pointless...

Now things start to take off, quick comparison with the 1333Mhz ram results. SuperPi 32M is over 10 secs faster. The architecture clearly benefits form more ram bandwidth. How much it requires will become clear in the following pages. Even Wprime1024 is gaining a few secs. Even though this benchmark is not really timing, nor ram related. 

 

Write performance is close to the previous 1333Mhz tests, though Copy and Read are getting mind blowingly fast. Almost 2500Mb/s gain, by just using 1600Mhz rams. Maybe a golden tip already for future buyers : stay away from 1333Mhz rams, it slightly bottlenecks your brand new CPU.

 

Cinebench seems less ram bandwidth dependant as some of the previous tests. The final scores are a tiny bit better than the lower speed divider, but only very little when the timings are tightened.

 

.  

Once we are at 1600Mhz ram speed, the tight timings only give a little boost in Cinebench or PCMark05.

 

 

 Same story for 3dmark 01 and 06, tiny gains with lowering the timings. Score increase is far less than with the 1333Mhz tests. Indicating the bandwidth is pretty optimal for this particular tests, being it at loose or tight timings.

Far Cry 2 getting a few FPS more with the 1600mhz ram kit over the 1333Mhz results. Mafia II is being run at almost max speed. The GTX 480 is pumping out some ridicilous FPS. I reran the tests as with the GTX 285, I was struggling to breach 85FPS... with the extra GPU power differences are easier to spot.

 

1866Mhz up next, will it bring even more performance or is there a setback ?:

 

 

First test start off pretty good, a small improvement over the two benchmarks, CPU and RAM wise. Sandy seems to enjoy every minute of it.

 

 

Wprime 1024 is within the error of margin now, but SuperPi 32M keeps on putting out faster numbers. AIDA64 will probably indicate nicely what exactly is going on.

 

 

The Write tests shows a marginal improvement, but again Copy and Read get a healthy boost. It doesn't stop to amaze me !

 

 

X264HD gains zero to none with the decreased timings. Did we reach a sweet spot ? Let's continue with the tests, to see if there's more...

 

 

Yep, being repetitively as ever :  tiny gains, but nothing really earth-shattering to see here. It scales a little bit, but again far less as with the lower ram clocks.

 

Pcmark05 scoring less than 80 points better, similar to the encoding tests. Timings don't seem to matter that much anymore at these speeds.

 

 

Even the 3DMarks are neck to neck with the 1600Mhz results. Even sometimes being beaten by the 1600Mhz C6 setup. Let's see if the games get a few FPS more with the improved bandwidth.

 

 

Blimey around 1 FPS difference, that's all there's too it. Think we can almost conclude that for a daily user, 1600mhz ram speed is more than sufficient. Giving a great overall combination of speed and efficiency.

 

 

2133Mhz was the highest divider that allowed me to boot with the Corsair GTX2 Hyper kit at 1.67Vdimm. I had to loosen TRCD to 8 to get the ram kit Superpi 32M stable ( all the other tests ran fine with 7 setting ) The next divider 2400 was a no boot, not even with the loosest timings on the rams... Let's go on with the show :

 

 

SuperPi 1m and Wprime32 are almost identical to the lower 1866 divider. We need more raw CPU power to get better scores. But that's not the intention of this particular test. 

 

 

SuperPi 32M benefits nicely from the extra bandwidth provided. Over 4 secs compared to the  1866 ram divider. So it is crystal clear that those that want maximum performance clock for clock, high speed rams is the way to go ! And if you can or ya wallet allows it, opt for tight timings too !

 

 

AIDA64 shows that Copy, Read and Write still manage to increase with the 2133 divider. And it's all so easy to setup. With my 2500K ES sample I did not have to increase the VCSSA voltage. Some other claimed they needed to up VSCCA voltage a bit to get 2133 stable. Tighter timings still boost even further...

 

 

Tighter timings almost have got zero benefit for the HD file encoding X264HD test. We already saw sort of a limit being reached with the previous 1866 divider test.

 

 

PCMark05 however does gain a mind blowing 250 points (sorry bit of sarcasm involved).

 

 

With Cinebench we ran into a few discrepancies at these speeds. Cas8-9 results very very close to 1866. Cas 7 was faster in the single core test at 2133, though multi core was time after time faster at 1866Mhz.

 

 

Our two futuremark 3D tests, still continue to improve with more ram speed. 3dmark06 is barely improving (GPU limited with the GTX 285) , but hey every gain counts. 3DMark01 tops at 59.6K. Compared to the slowest 1333mhz setting almost 2000 points gain.

 

 

Ubisoft's Far Cry 2 continues to impress me. Give it more CPU, more bandwidth and it still continues to pump out more FPS. Of Course we are running only at 1280 x 1024, albe it  with very high detail setting selected. Adding more AA, higher resolutions will put a damper on our joy, as the frames will stabilise sooner and the GPU will turn into the bottleneck. Mafia II finally reaches scores over 127FPS.

Glad you made it through the previous pages. What happens if we put all of the CPUs at the same clockspeed and disable the turbo function. We will perform some basic tests to see how the young and the old perform clock for clock. 

Here's a rundown on the specs :

AMD 965BE/1090T@4ghz : 20 x 200HT, 2600NB speed and 4Gb DDR3 1600Mhz CL 8-8-8-24 1T

Intel QX9650@4Ghz : 10 x 400FSB, PL9 and 4Gb DDR3 1600Mhz CL 8-8-8-24 1T

Intel i760@4ghz : 20 x 200 Bclock, 4gb DDR3 1600Mhz CL8-8-8-24 1T

Intel 920/950 and 970@4Ghz : 20 x 200 Bclock , 3200 uncore and 6Gb DDR3 1600Mhz CL 8-8-8-24 1T

Sadly our i7 1156 system wasn't tested at 4ghz as it failed to boot. After numerous LN2 sessions the 875K  Lynfield gave in... RIP

I hope for some that there are no real surprises in the above specs. I would even refer to them, as possible daily 24/7 clocks. Enough with the chit chat. Starting off with the ram SuperPi test :

 

The new Sandy Bridge generation seriously ass-whooping all the competitors. Intel did one hell of a job here to even beat their current flagship ( which is no slouch either and famous for it's raw power ) SuperPi has never been AMD territory, maybe Bulldozer can finally change that...

 

 

SuperPi 32M shows the new technology improvements even better. Over 50secs gain over the I7 1366 CPUs. Who would have thought that a dual channel setup would ever beat triple channel ? Not even that Octopus would have ever picked that :p

 

 

Look at them speeds, now this is what we call über bandwidth. Again, not even the I7, our former bandwidth king of the last years can touch it. Heck it's not even coming close. Sandy bridge FTW !

 

Time for a Cinebench R10 64bit showdown :

 

Single core wise, the new "Sandies" are the best performers. Multi core wise they are just beaten by the Gulftown hexacore cruncher. For the other competitors, sorry lads, SB clearly dominates.

 

At the Tones Sandy Bridge brunch, lots of visitors asked me about the encoding performance of Intel's new born. Same story as with Cinebench, only Gulftown will be the weapon of choice for X264HD alike. Of course it all will depend a lot on your encoding software if 1155 or 1366 socket is preferred, or if it renders on your GPUs. This little bugger is no slouch at all ! Mainstream my ... ( fill in dirty word for booty on the dotted line please )

 

 

Let's throw a bunch of games at our humble overclocked CPUs. Sandy Bridge again right up in front of the pack. Take note that these are average FPS. Sometimes the minimum FPS was slightly higher on the 1366 platform. The DX11 Alien Vs Predator is completely GPU limited. Even the slowest CPU in our test manages to pump out a decent 75 FPS.

 

 The DirectX10 runs of Dirt 2 and Far Cry 2 put the new K CPU's right were they belong. On top of the charts. For gamers there's no need to cash out for a power-hungry 1366 socket equipped rig. At first I expected the Sandy Bridge lineup to be just a small evolution of the 1156 socket. Intel's engineers have really surprised this sceptic reviewer big time.

With each new CPU generation, the expectations of the enthusiast community gets higher. While with I7 s1366 4.5Ghz was doable for some. Not everybody reached them clocks for daily use ( this with Hyperthreading on)

The early non NDA results showed in some benchmarks clocks of 5ghz and beyond on air for the Sandy Bridge CPUs. Some tests were even executed on the boxed cooler (with a little help form cool ambient or a portable AC unit).

Time to explore what can be done on a Thermalright Venomous X with one 120mm fan running at 1600rpm.

For testing I ran a two different scenarios :

First test with stock volts (my 2 CPUs were both detected set at 1.2 Vcore by the Asus bios) and the P8P67 DLx equipped with the Thermalright Venomous X

 

 

Not bad at all : 4.2ghz for the 2500K and 4100Mhz for the 2600K. OC potential will depend a lot on ya CPU binning, motherboards power circuitry and co.So don't shoot me if your CPU might not reach the above clocks. if yours do better I'm in for a trade :)

How about 4.5Ghz maybe for daily clocks ? After some fiddling around in the bios both CPU's required about +/- 1.3Vcore to maintain prime stability. Main key on my setup is to keep the CPU voltage stable via the motherboard loadline calibration settings. Here's a screenshot of my bios settings for the P8P67 Dlx (latest bios 1253) :

This setting allows me an OC of 4.5Ghz ( so with reasonable volts and temps ) and yet the CPU steps down when being unused or light loads to eg 1600mhz at 1.0 Vcore. Ofcourse these settings can differ from board to board and CPU. The same CPU is benchable at 1.485Vcore at 5200mhz with just a few alterations in the bios.

 

Now why not push for higher ? Well take a look at the below charts for the temperatures with my VenomousX and a single fan spinning at 1600rpm :

 

I'm not the type of guy that likes high temps and if an hour prime95 ( Custom test 12K ), can cause this heat output. Then you think LinX or another Intel Burn test program can do even worse. Well lads, it's the inverse world now. With the 2600K, Prime95 version 36 is more stressful then the latest LinX version. For proper Intel Burn test stress testing Windows7 has to be updated with Service pack 1 ( it's required to use AVX instructions ) In real applications your CPU will never ever get this hot. But I'm aware that some will push for a 100 - 200 LinX run or a 24 hour Prime test. Till now, we got no idea if CPUs will degrade  with high Vcore under air or exotic cooling. Only time will tell.

There's already one death reported,  but that owner pushed 1.72Vcore through the CPU, just for a CPU-Z validation screenshot. Think he's the cause of his loss, yet he screams for RMA as he isn't happy...
It isn't because someones CPU can do 5ghz , that yours can do too on the same setup. We tested a few CPUs at the Tones Sandy Brunch event. Out of 5 ( all same batch) when benching there was already a discrepancy  of 300Mhz. Hwbot is starting a batch thread right here

A golden rule : Take it slow, monitor temps, there's more than enough OC potential in these beauties.

And then some fun screenshots, these can be obtained after just 15 mins of tweaking: this platform rocks !! 

 

   

 

 

More overclocking results by Massman on the next page

 

We finished some nice benching sessions with the Single Stage Phase unit to defend our national colours in the Hwbot country Cup. Benching with the Sandy Bridge CPUs is almost child's play. It's so much easier to get high speeds and pretty amazing performance bang for the buck.

The CPU that we used was one of the Tones demo CPU's and was found okay for 5.4Ghz ops. Leeghoofd's sample CPU goes up to max 5250Mhz... anything above results in BSODs, reboots and freezes. With Sandy Bridge you can find the max potential clocks pretty fast. Previous Intel CPUs involved a lot of tweaking : Ram dividers, Uncores,  PLLs, etc...

Me and Massman benched on the Asus Maximus IV Extreme. Equipped with the Gskill Flare kit.  We got some hefty ram clocks going at sub 1.7Vdimm volts. How does 2133Mhz 6-9-6-21 1T Command Rate sound to you ?

Anyway without further ado here are some screenshots of last night's bench session :

 

 

 

This new breed really is stellar in 3DMark01 and 03. It ain't bad either in Aquamark3. For daily applications it screams, no matter if it runs stock or overclocked. For the benchers it really is a good alternative to the rare E8600 and Rampage Extreme boards. No more LN2 needed, if you have a good SB CPU, you can even do the above clocks easily on Air. Much will depend on your CPU. After 15 mins you are already aware if it will be a top or not such a top CPU. Let the binning game begin ( again ) :p

Well engineers of Intel, hats off. This for creating such a number crunching monster and daring it to launch at a mainstream price. Well that's the only thing that really is mainstream for this CPU. Its performance is top notch and will create big envy to those that have recently bought an I7 socket 1366 setup. Sandy Bridge is in 80% of our tests the best performer without any doubt. Let's make a small summary of my page filing tests :

Even if you don't buy one of the two K models, the performance of this new platform is overwhelming. But for a mere 10-20 euros you have the liberty to get some more speed out of it. Now the word overclocking might scare some, but it's so dead easy.  We are seeing 4.5 to even 5ghz stable daily clocks. On air lads ! Just a beefy air cooler is required.

Whatever model you will choose, these CPUs will not disappoint! 

 

 

As overclocking is Bclock limited, the best way to squeeze that little more performance out of it is to use high speed rams. Anything below 1333Mhz is not a viable option in my book. From 1600mhz ram speed on, this platform starts to feel at ease and only lingers for more ram speed.

For daily user/gamers/encoders : buy a 1600Mhz or better  kit. Timings won't do much for you, CAS 8 or 9 is more than suitable. The bandwith provided is phenomenal.

For the benchers : you know the drill :  the higher the speed the better, the tighter the timings the better...

For the motherboards : unless you are using this as an office PC : there's only option for the chipsets : P67. For the office or Home HTPC : H67 is a viable option. This only as the latter chipset ,has got limited ram divider support. Clock for clock it's as fast as its full born brother, the P67. But the 1333mhz ram support and zero Bclock OC'ability cripple it. Making it really a no go for gamers and/or enthusiasts.

We haven't included any IGP tests, yet, as the boards we have are constantly getting bios updates and performance is improving all the time. Also we still need to get our hands on a non K CPU so we can put 2000 and 3000HD head to head. In the upcoming Asus P8H67 M Evo review, IGP tests will be included for sure :)

So do you need to upgrade ? It all will depend a lot on your current configuration. Socket 1366 users won't have to. The ones on other platforms can give it some thought. Think of Sandy Bridge like this: you get better performance clock per clock, far easier setup with high ram speeds and it's very power efficient. And in case you opt for a K model it will really blow you away !

I wanted to give this new born at least the performance award for its blistering speed. From a benchers perspective maybe a value award, as it's a good alternative to the existing Gulftown 1366 setups... The later costing plenty more and need Liquid Nitrogen to get them soaring... Socket 1155 is a more cost effective way to have some really big fun !

 

 

Intel's Sandy Bridge CPU Recommended For

 

 

 

It's really hard not to be astonished by Intel's latest creation. I really hope for AMD that their upcoming Bulldozer is what most AMD fans expect it to be, or even better. But don't forget Intel is not resting on its laurels either. Ivy Bridge is just around the corner. And if the latter boasts higher performance than Sandy Bridge (a metal cocktail of 1155 and 1366), then we all will have more processor power then we can ever use... 

 

And please Intel spokesmen, stop referring to Sandy Bridge as mainstream ! It's far from that.

 

I wish to thank the following people and companies who made this review possible:

Milan and Rogier from Asus for the sample boards

 

 

Gareth Ogden from Corsair for the Corsair Dominators 1600C8 RAMS

Manu from Tones for the retail 2600K CPU's, we owe you big time !