Intel i5 655K Review

Every once in a while, something truly unexpected comes along. A piece of hardware you thought you’d never see in a million years. A sub-kilobuck CPU with an unlocked multiplier from Intel used to be one of those things, at least until the processors with a “K” in their model number showed up.

What?!

Whoever it was at Intel that pushed the idea of releasing two CPUs with unlocked multipliers, you deserve a tip o’ the hat for sure. Previously, some of the most entertaining yet affordable overclocking odysseys occurred exclusively on the AMD platform. They have consistently had unlocked multipliers in their Black Edition processors with relatively little price premium over their locked-down brothers.

Now there are two new, unlocked, Intel processors available for the P55 platform with the coveted unlocked multiplier – the i7 875K and the i5 655K. At their hearts are the i7 870 and i5 650 and there is no indication that anything has changed except to unlock their multipliers.

That’s no small change though. Historically, Intel has saved unlocked multipliers for the best of the best of their CPU lines. They were designated Extreme and tended to cost at or over a thousand dollars.

Here’s some of what Intel has to say about their new, spiffy unlocked chips:

Up, Up & Away!

Up, Up & Away!

All true there. Pushing the CPU’s frequency without needing to worry about the rest of the system sounds great to me! Here are the specifications for both of the new CPUs:

i7 875K Specs

i7 875K Specs

i5 655K Specs

i5 655K Specs

A quick comparison of some of the Core i7 & Core i5 chips:

Chip Lineup

Chip Lineup

The most important thing to note here is pricing. The 655K is a reasonable $40 premium over the locked 650. However, the 875K is only $342 – a very sharp drop from the locked 870′s price tag. The numbers in the chart are per 1,000 units, but if you buy enough thousands, they’re apparently even less. Newegg has the 875K listed at $329.99 right now, while the 870 is still at $579.99. For reference, the 655K currently sells for $209.99 and the 650 for $179.99.

Here they are all boxed up and ready to be delivered to your door!

Retail Boxes

Retail Boxes

Testing – Stock & Overclocked

Today we have the i5 655K in-hand to put through its paces.

Intel i5 655K Marked

Intel i5 655K Marked

Intel i5 655K Front

Intel i5 655K Front

Intel i5 655K Back

Intel i5 655K Back

First up, we’ll go through the single test run at stock clocks only – the Everest suite of CPU & FPU tests. All of the other tests were run overclocked as well. For comparison, we’ll use results from another recently reviewed chip, the AMD Athlon II 640 (see the ASUS Crosshair IV Formula review) as well as those from a Phenom II 965 BE also here for testing.

You may notice a few minor differences in the stock numbers for the 655K as opposed to the CIVF revew. That’s because the chip was put through the standard three runs at stock testing after that review (for which they were run once just for comparison purposes). This gives an accurate representation of what you can expect from out-of-the-box performance at the rated speed of 3.2GHz.  The test system used in this review consists of:

  • CPU: Intel i5 655K
  • Motherboard: EVGA P55 FTW
  • RAM: Patriot Sector 5 DDR3-2400
  • GPU: Gigabyte 5870 SOC
  • PSU: Corsair TX650
  • OS for stock and stable OC testing: Windows 7 Professional x64.
  • OS for less than stable overclocking: Vista Enterprise x86.

The CPU was cooled with a custom water loop, in this order:

Note that this chip does have turbo boost, which allows up to a 260MHz boost for single-threaded applications. For our purposes, that feature was disabled during testing.

Now, the Everest results. Note that the CPU AES test is missing because of obviously flawed results. Even with the most recent version, they were in the 200,000 range, when the Athlon & Phenom results were in the mid 20-thousand range and all reference results were below 50,000.

Everest CPU Tests

Everest CPU Tests

Everest FPU Tests

Everest FPU Tests

A good showing by a dual core CPU for sure, with FPU results that were especially strong compared to the CPUs with four physical cores.

Overclocking

Overclocking this chip was an absolute blast. I haven’t had this much fun since I tested the first unlocked CPU I ever had, an AMD 5000+ Black Edition. Things have come a very long way since then.

The first thing we’ll look at is 24/7 stability. Prime95 Small FFT was the stress test of choice, run only for an hour due to time constraints. These are good starting numbers. More time may be needed to ensure rock-sold 24/7 stability.

This chip is no slouch at all. It started at a modest 4GHz to start getting a feel for how it behaved.

4GHz Stable

4GHz Stable

Once it felt pretty good, I set an arbitrary, reasonable maximum voltage of 1.35Vcore and 1.35V VTT. After that, BCLK was set to 160 (it actually ended up being raised to 161 later)  to get the RAM up a little bit to DDR3-1600/8-8-8-24. The system booted just fine so it was time to crank the multiplier and start stress testing. Here’s how it turned out.

4.5GHz Stable

4.5GHz Stable

Nothing less than a stunning 4.5GHz, stable overclock! These are 24/7 voltages too. So you’re talking about a 1.3GHz overclock from stock speed, and 4.5GHz 24/7 operating speed. Nothing short of amazing.

Performance Results

Gaming Performance

Real quickly, we’ll go over a game test. Our main audience is obviously overclockers, but a lot of us (myself included) wouldn’t be here if we didn’t start by pushing our systems to get that extra few FPS.

The game benchmark of choice was Stalker: Call of Pripyat. It was tested at 1280×1024 on the “Ultra” setting with both MSAA turned off and 4x MSAA.

Stalker: Call of Pripyat Benchmark

Stalker: Call of Pripyat Benchmark

As you can see, without MSAA, increased clock speed helps quite a bit. Overclocked, the 655K runs away with it. With MSAA turned on though, it would seem GPU power is the name of that game.

On to the benchmarks!

2D Performance

To flesh out the performance numbers a bit, we are able to include results from an i7 870 clocked at 4GHz or the 2D and 3D performance graphs, with the exception of PiFast.

The first 2D graphs are the fast and fun benches – SuperPi 1M, WPrime 32M & PiFast.

As expected, this CPU shines at single-threaded applications, beating out every competitor in SuperPi 1M. Even in the multi-threaded WPrime 32M & PiFast it held its own, with overclocked results beating out both stock-clocked AMD chips WPrime32M and trouncing both in PiFast.

SuperPi 1M, WPrime 32M & PiFast

SuperPi 1M, WPrime 32M & PiFast

Next up are the longer running brothers, SuperPi 32M and WPrime 1024M. The SuperPi 32M results are graphed in seconds, but the actual score in minutes:seconds is displayed in the data labels.

SuperPi 32M & WPrime 1024M

SuperPi 32M & WPrime 1024M

With single-threaded SuperPi, the 655K beat up on the competition again, except for the 870, which somehow pulled out a win with 500MHz less clock speed. This caused a re-check of the benches, but it’s correct. After being overclocked, this little dualie showed it has some guts again in multi-threaded apps, beating out the Athlon II & Phenom (both at stock speeds of 3.0GHz and 3.4GHz, respectively), just being edged out by the overclocked Athlon II.

To round out 2D performance, we’ll look at something indicative of more ‘real-world’ performance – rendering. The metrics are supplied by Cinebench, both R10 and R11.5.

Cinebench R10

Cinebench R10

Cinebench R11.5

Cinebench R11.5

Well, it’s not going to set any rendering records, but when overclocked it can definitely hang with a couple of quad cores. Considering this chip is just ripe for the clocking, that’s not a drawback at all.

3D Performance

For 3D testing, the two benches with CPU scores separated out as well as applied to the final score were used – 3DMark06 and 3DMark Vantage. The 655K results are the two brighter colors. The Phenom II 965BE was not compared in these.

3DMark06 & 3DMark Vantage

3DMark06 & 3DMark Vantage

While eight threads makes a ton of difference, it would seem a large overclock makes one as well. Not bad performance at all considering it’s compared to both a true quad core and a quad with HT!

Pushing the Envelope

In typical fashion, literally the first thing I did after pulling this CPU out of its box was overclock the snot out of it. Didn’t even see what it would do with HT off, just started cranking. For benching on water, the maximum voltage allowed was 1.45V VTT and 1.5Vcore. This is a 32nm chip after all, and I would strongly caution against running even at those levels for any length of time.

In any case, this is what a mere couple of hours managed to get for SuperPi 1M & 32M, WPrime 32M & 1024M and max CPU-z validation. The OS for these runs was Vista x86.

As a side note, I did try finding the max BCLK for this chip. Unfortunately it was right around 215…the same as the i5 750, two i7 860s and the i7 870 that came before it. I think it’s safe to say that’s a board problem at this point!

SuperPi 1M was of course the bench that completed at the highest clocks. Ironically, the faster time (on the right) was completed at lower clock speed. Both the SuperPi 1M and 32M times were a personal best.

SuperPi1M - Max Freq 5020MHz

SuperPi1M - Max Freq @ 5020MHz

SuperPi1M - Fastest Time @ 4928MHz

SuperPi1M - Fastest Time @ 4928MHz

SuperPi 32M - Completed @ 4928MHz

SuperPi 32M - Completed @ 4928MHz

Moving on to WPrime, both of these times earned a few global points on HWBot. Fun stuff!

WPrime32M - Completed @ 4884MHz

WPrime32M - Completed @ 4884MHz

WPrime1024M - Completed @ 4806MHz

WPrime1024M - Completed @ 4806MHz

Last, but not least, we have the max CPUz validation.

Yep, with HT on, this chip managed to validate at a whopping 5155MHz!

Final Thoughts & Conclusions

What is there to say really? Here we’ve seen a stellar Intel dual core with HT and an unlocked multiplier absolutely blow through benches, overclock like mad (with relative ease) and do it all without breaking the bank.

Now, keeping things in perspective, we should note the results on HWBot for the 655K are pretty darn close to those of the i5 650, further indication that this really is what its name applies – a 650 but with an unlocked multi. This is not a bad thing, especially if you don’t already have a 650.

If there is anything bad to say about this processor, I’m having a hard time finding it. Um…it’s not a quad core, there. But reading this, you weren’t looking for a quad core anyway. Fast, fun and full of potential, this chip has Overclockers Approved written all over it.

If you find yourself in the market for an Intel dual core, this one deserves a look or three, followed by a press of the “Add to Cart” button. Then hours of overclocking bliss!

Tags: , , , ,

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

Discussion
  1. EarthDog
    This. :clap:


    This.

    hokiealumnus
    I wish. Just not in the cards at this point. We'll just have to see what Intel has up their sleeve for the future...and whether we can get our hands on one. :thup:


    Looks like you were typing while I was posting. :)

    Wish I had one though, I really do!
    Thanks all!

    wingman99
    I like the review, maybe you could do the i7-875K next.:)


    I wish. Just not in the cards at this point. We'll just have to see what Intel has up their sleeve for the future...and whether we can get our hands on one. :thup:

    SamSaveMax
    Very nice review.

    The only thing that I've noticed is that the SuperPI was good for 5020MHz finished in 8.687s. But 32M SuperPI was really slow for 4928MHz finished in 8m 59.203s...especially for 7-7-7 timings.

    How was the temps?....could it be heat related?


    Aye, I noticed that too, but the 7-7-7 timings were at DDR3-1333, so that's what I attribute it to. The i7 870 time was at DDR3-2000 / 9-10-9 (from this review), so I guess it was that. The result was also surprising to me. Guess RAM really does make as much of a difference as people say.

    Temperatures definitely weren't a problem. Max Prime95 Small FFTs temp at that speed (4.5GHz) was 59°C.
    Very nice review.

    The only thing that I've noticed is that the SuperPI was good for 5020MHz finished in 8.687s. But 32M SuperPI was really slow for 4928MHz finished in 8m 59.203s...especially for 7-7-7 timings.

    How was the temps?....could it be heat related?