If geeks love it, we’re on it

AMD Phenom II X4 955

AMD Phenom II X4 955

When AMD released the Socket AM3 processors earlier this year, the 900-series was left out of the mix. We found the Phenom II X4 820 and Phenom II X3 710 to be pretty potent little chips, but the range topping 900-series chips, the Phenom II X4 920 and X4 940 Black Edition were stuck in the land of DDR2.

That time is over.

Today, AMD is releasing its new flagship processor, the Socket AM3 Phenom II X4 955. This new Phenom II is fully backwards compatible with AM2+ motherboards, but brings a DDR3 memory controller and a boost in clock speed over the previous top Phenom II, the X4 940 Black Edition. Also joining the Phenom II X4 955 BE is the Phenom II X4 945, a lower clocked and multiplier-locked version.


The new chip should hit shelves today priced at $245. This puts it at an awkward, but strategic position in the marketplace. Intel’s similarly priced Core 2 Q9450 was recently discontinued, so the chip has no direct competition at its price-point. Instead, it looks up Intel’s line to the Q9550, a $280 chip whose price nears that of the Core i7 920. AMD has set its sights high while keeping the price low enough that the chip doesn’t compete directly price-wise with the Nehalem.

AMD graciously sent the Phenom II X4 955 to us, along with a DDR3 motherboard and memory so we can test this update to the Dragon platform for ourselves. Unfortunately for today, MSI has not released a BIOS to support the new processor for our test bench, so we are unable to test its DDR2 performance. We did, however, secure a Q9550 for DDR2 and DDR3 comparison, giving us a look at how the processor performs compared to past models we’ve tested.

Detail and Specifications

The new AM3 Phenom II processors share many similarities with the early Phenom II models released back in January. As such, this preview primarily covers the architectural differences between the two models. Readers interested in learning more about their common architecture should read our Phenom II launch review.

Let’s take a closer look at the AM3 Phenom II and see what makes it unique.

Common AM3 Processor Specifications:

  • L1 Cache Sizes: 64K of L1 instruction and 64K of L1 data cache per core (512KB total L1 per processor)
  • L2 Cache Sizes: 512KB of L2 data cache per core (2MB total L2 per processor for X4s / 1.5MB total L2 per processor for X3s)
  • L3 Cache Size: 4MB or 6MB (shared)
  • Memory Controller Type: Integrated 128-bit wide memory controller *
  • Memory Controller Speed: Up to 2.0GHz with Dual Dynamic Power Management (all current AM3 processors)
  • Types of Memory Supported: Support for unregistered DIMMs up to PC2 8500 (DDR2-1066MHz) -AND- PC3 (DDR3-1333MHz)**
  • HyperTransport 3.0 Link: One 16-bit/16-bit link @ up to 4.0GHz full duplex (2.0GHz x2)
  • Total Processor Bandwidth: Up to 33.1 GB/s total bandwidth
  • Packaging: Socket AM3 938-pin organic micro pin grid array (micro-PGA)
  • Fab location: GLOBALFOUNDARIES Fab 1 module 1 in Dresden, Germany (formerly AMD Fab 36)
  • Process Technology: 45-nanometer DSL SOI (silicon-on-insulator) technology
  • Approximate Transistor count: ~758 million (45nm)
  • Approximate Die Size: 258 mm2 (45nm)
  • Max Ambient Case Temp / X4 CPUs: 71o Celsius
  • Max Ambient Case Temp / X3 CPUs: 73o Celsius
  • Max Temp Phenom II X4 955:62o Celsius
  • Nominal Voltage: 0.875-1.5V
  • Max TDP: 125 Watts

*Note: MC configurable for dual 64-bit channels for simultaneous read/writes
**Note: for DDR3-1333, AM3 boards will support 1-DIMM-per-channel @ 1333MHz

The biggest difference between the early Phenom II AM3 processors and the new Phenom II X4 900-series models is the L3 cache. Quad core 800-series chips are outfitted with only 4MB of L3 cache. The new Phenom II X4 955 and 945 feature 6MB of L3 cache. Since the transistor count is constant between both the 4MB and 6MB models, we can deduce that part of a 6MB processor was simply disabled to create the 4MB model. We’d encourage anyone interested in learning more about AMD’s stance on disabling CPU components to read our X3 review from last year.

The biggest change side from cache is the increase in frequency. The Phenom II X4 955 is the fastest Phenom II offered to date with a clock speed of 3.2GHz. That’s only a 200mhz step up from the X4 940, so we hope to see just how much of a difference a little speed and DDR3 make for the chip.

Here are the requisite CPU-Z screenshots for the Phenom II X4 955.


Also, for comparison sake, here are screenshots for its opponent today, the Q9550 in both DDR2 and DDR3 environments.



Despite being clocked lower, the Q9550 packs 12MB of L2 cache.

Next page »


  1. Leonardo
    Leonardo Thank you for the review. It's good to see AMD with a successful CPU in this processor range. Looks like a fair price/performance platform to be based on this model.
  2. DrLiam
    DrLiam Great review, this looks like a very nice addition to AMD's quad core inventory.
  3. Cliff_Forster
    Cliff_Forster This is really exciting. The results at stock are solid, and with it being an unlocked chip its going to be an easy overclock.

    AMD has even posted a guide to overclocking using these unlocked chips. AMD really embraces this concept as you can see a number of black edition chips as various price points.

    AMD is making the more tweaker friendly options of the major players and the 955 is another fine example.

  4. Snarkasm
    Snarkasm Pete, I may have missed it, but do we have any kind of ballpark comparison on temps between the 955 and the Q9550 and i7 920?
  5. lemonlime
    lemonlime Great review, Pete :)

    Nice to finally see a non-disabled quad in AMD's AM3 lineup.
  6. Mirage I was reading the PII-955 reviews today and I enjoyed this one as well, thanks for the review. I have a couple of comments.

    - I think the comparison of DDR3 and DDR2 on a Q9550 is not very useful. Q9550 does not have a memory controller. It connects to the north bridge to access memory via quad-pumped 333MHz bus. As long as both memories can saturate this bandwidth, DDR3 will be slower (or at most equal) due to higher latency of DDR3 than DDR2. DDR3 advantage can only be enabling higher 1:1 FSB for overclocking.
    - If you can add a i7-920 with dual-channel memory configuration (instead of triple-channel) that would be very interesting to compare with PII-955, IMHO. Many reviewers do not even mention that i7-920 works with dual-channel configuration perfectly fine at an almost same performance level compared to triple channel and i7 system cost will be a little lower. Triple channel gives a very nice advantage for expanding the memory later by adding only one stick (of the same specs) without any performance penalty (actually performance bonus).

    Thanks again
  7. BuddyJ
    BuddyJ Thanks for the comments all. Your points are all well taken Mirage. I'd love to add more i7-920 benchmarks in dual and triple channel configurations.

    For those of you looking for additional overclocking info, chew* at XS has some great screen shots. That should hold you over ;)

    Snark, we don't have accurate enough temp numbers for all three. I'll get temps in the overclocking article since that's where they really come into play.
  8. Cliff_Forster
    Cliff_Forster XS has been doing that forever, they just slap up a bunch of scree grabs so they can say "First!"

    Its annoying, I'd just as well wait a week or two, have someone really put the chip through its proper paces, and write something coherent.
  9. Heath Heya, great review.

    I very much cant wait to see what the numbers look like on a AM2+ board.

    I have been reading how the Phenom II's will work on the M2N32-SLI and a few other AM2 (non-plus) boards with a new bios, some officially (bios reports chip correctly), some unofficially (bios reports unknown processor)

    I was wondering if you could throw an older AM2 (non-plus) board in to the mix when you run the AM2+ to give people with older boards an idea of how well the new chips run on the old boards. (probably even with ddr2-800 as that's what most people with the older boards run).

    I would like to see if I will get much benefit upgrading from my 9750 to a phenom II. The plan is to get the chip in a few months time and run it in the M2N32 till the end of the year when I will put it in a new AM3 board.

    The first Athlon x2 that lived in the M2N32 got put in a 780G and turned in to a HTPC. The 9750 will go back in to the M2N32 and replace my aging server.....

    Man this old board has had a good long life. =)
  10. Insight-Driver
    Insight-Driver I read a number of views on the Phenom X4 955. I then went to Newegg and put some wish list systems together. I am of the mind that I would go to an I-7 processor because I could, at the same price as the Phenom system get the same performance, with more margin for future upgrades. I think this is because I want to build a system that will last a few years. I can't seem to be satisfied with one system more than a couple of years before I get the itch to build a new one.
  11. Thrax
    Thrax You and I are of the same mind, Insight Driver. While I respect AMD's value angle, I simply cannot find it in me to get behind it.

    AMD is recommending a motherboard that tips the scales at $180 for the X4 955. That's not cheap, and it's in the same price range as any number of high-performance X58 boards. Comparing the cost of AMD's recommended motherboard with a 955 to one of those X58 boards and a Core i7 920 reveals that there really isn't an unbridged chasm in platform value.

    I'm not sold on this front.

    Let's step away from value and look at possible upgrades during the expected lifetime for each company's socket:

    AMD Socket AM3: Phenom II until 2011.
    Intel LGA1366: Nehalem until 4Q09, then Westmere until 2011.

    LGA1366 immediately presents an obvious upgrade path with the Westmere. While the Nehalem already offers a performance lead on the Phenom II X4, the Westmere will widen it further while reducing size, heat and power consumption. Unless AMD has some magnificant new stepping of the Phenom II lurking in the wings, it has no direct way to bridge that performance gap until 2011.

    So I have to ask myself: How can I support this processor when it's comparable in price, loses in the benchmarks, and does not present a clear upgrade path like its competitor?

    /me shrugs.

    Maybe AMD has an answer that will catch us entirely off our guard. In the mean time, selling the value angle to me will be virtually impossible.
  12. Cliff_Forster
    Cliff_Forster Robert,

    You at least have to appreciate the dual memory controllers making the platform so scalable. Part of the value is in being less "boxed in" on how you leverage the Phenom II as a value proposition.

    Have an existing AM2+ board that supports the recommended TDP? You can upgrade to the latest and greatest in AMD's line, you cant do that going from Core 2 to i7.

    Lets say your not an enthusiast but you still demand a serviceable level of multimedia performance, AMD's board IGP's are far superior to anything Intel offers, a cost savings can be there for certain users.

    Lets say your a tweaker and you just want a reasonably priced unlocked chip to play with, AMD is the only company offering this.

    In total, the AMD platform is just more flexible, its all in how you leverage it. You don't have to use the 790FX/750SB combo AM3 board, you have a wide range of options.
  13. Thrax
    Thrax Cliff,

    The Core 2 Duo/Quad is 3 years old. Let's go back that far and look at AMD's socket: AM2. AM2+ and AM3 processors do not work in an AM2 socket. Both manufacturers are on even ground for anyone coming from an architecture released in 2006 or 2007, and any customer from this generation is in for a new motherboard and processor.

    And let's be a little honest about AMD, here. Selling a lack of innovation as "flexibility" is a bit disinenuous. In the lifetime of the Phenom II, Intel will have had four separate generations of processors, each faster than the last: Conroe, Penryn, Nehalem and Westmere. AMD has had Agena and Deneb.

    Not only has AMD stood on the same ground as Intel with socket changes, it's done less innovation in the same amount of time.

    You sort of got me on the IGP business, but we both know that Intel GMA is crap, and any HTPC with an Intel chip would be running on an NVIDIA IGP.

    Lastly, any overclocker knows that locked vs. unlocked is hardly a matter for concern any more. Modern FSBs make FSB tweaks as easy as multiplier adjustments. I can set a Core i7 920 to 4GHz just as quickly as you can take a Phenom II X4 to 4GHz. The approach is difference, but the time and the cooling requirements are identical.
  14. BuddyJ
    Thrax wrote:
    AMD is recommending a motherboard that tips the scales at $180 for the X4 955. That's not cheap, and it's in the same price range as any number of high-performance X58 boards. Comparing the cost of AMD's recommended motherboard with a 955 to one of those X58 boards and a Core i7 920 reveals that there really isn't an unbridged chasm in platform value.

    Not entirely true. AMD's list of recommended mobos have prices starting at just under $100 and peaking at $190. And you can't get a good Core i7 board for that. The Gigabyte X58-UD3R and DFI X58-T3eH6 are the bottom two enthusiast boards; the Gigabyte is slower than most but overclocks well, and the DFI is hit-or-miss. The core of enthusiast boards seem to start in the $240 range and go well over $300.

    If you cheap out with a 955 build, you're $245 + $120 + $30 (2GB DDR3-1333) = $395
    If you cheap out an i7 920 build, you're $288 + $180 + $50 (3GB DDR3-1333) = $518

    The question becomes, does the 25 percent increase in cost to go to Core i7 give me at least that much performance increase? That's where the value argument takes place.

    But most of us don't want systems that are shady tweakers. We want the fancy motherboards, fast memory in large capacities, and the whole enchilada. If the prices for the best AM3 boards fell like the prices for the nice Core i7 boards have, perhaps they'd have a chance at this system level, but in reality, the difference still is around $100. And when you're talking overclocking systems, a 4GHz Core i7 will rock a 4GHz Phenom II. Without a boost in clock speed, the Phenom II will remain "that chip that the LN2 guys use"
  15. BuddyJ
    BuddyJ Or, if you're poor like me, it's the chip I got so I could have fun tweaking stuff and still save $100.
  16. Cliff_Forster
    Cliff_Forster AMD is not lacking innovation, Intel is in fact robbing AMD's

    Killing the old fashioned FSB, AMD got there first, True quad core designs, AMD got there first, integrated memory controllers, AMD got there first.

    AMD did the innovating, Intel just has an edge in resources.

    Other than its benchmarks, tell me what is innovative about the i7 platform that Intel did not borrow from a prior advancement from AMD?
  17. BuddyJ
    BuddyJ The innovation comes in combining all the wonderful technologies that allow the chip to perform. Why doesn't the Phenom II have hyperthreading-esque tehcnology, and if it did, would its performance improve enough that we wouldn't be having this discussion? Also, who was first to 45nm? cough*Xeon5400*cough
  18. Cliff_Forster
    Cliff_Forster A die shrink is barely innovation, its just the typical path of progress.

    Dual core was AMD's superior answer to hyper threading.
  19. BuddyJ
    BuddyJ Great. Let's see their answer to fourhyperthreadedcores.
  20. Thrax
    Thrax A die shrink is a huge innovation. It's the culmination of years of R&D and numerous process breakthroughs. It's why AMD took time out of their day to piss on Intel's 32nm parade with their own 28nm GPU parade.

    But Intel didn't just shrink the die from Conroe to Penryn. Clock for clock, Penryn is 10-15% faster than its predecessor. AMD heralds clock-for-clock improvements in evolutionary chips like the Phenom II as well. ;)

    Call it innovation or call it robbing, but Intel is winning with it. AMD isn't.

    Regarding hyperthreading: It's an incredible idea. Really. It's truly brilliant. It's just a damn shame that it had a poor introduction because Netburst was awfulbad.
  21. Insight-Driver
    Insight-Driver I agree with Buddy J on the numbers as far as build costs are concerned. In my thinking, though, I think the LGA 1366 and DDR3 will have legs for a few years. New generations of the Intel chips area coming down the pike so prices on the current I-7's will trend down. DDR3 prices will trend down. I do think, though that the price difference is nearly proportional to the performance difference based on reviews and comparisons I've seen so far. I still think, for the kind of computing I do, it becomes a better choice for me to go the I-7 route.
  22. Leonardo
    Leonardo Great discussion, guys, keep it going.
    So I have to ask myself: How can I support this processor when it's comparable in price, loses in the benchmarks, and does not present a clear upgrade path like its competitor?
    Fair enough. Without putting my finger on the exact data evidence and without taking a side on the "innovation" debate, I can though, say that I think AMD's lag behind Intel is not as severe as it was a year and 18 months ago. AMD has advanced from yawn to hmm, maybe something's happening here.
  23. Cliff_Forster
    Cliff_Forster I'm not saying die shrinks are not important. They obviously are critical to processor design, reduce current leakage, increase efficiency, reduce heat output, normally improve frequency.

    The point I am making is saying that one or the other go to a certain nm process first, while great for them, its hardly an innovation. They are both chasing each other to shrink, its part of the design paradigm, die shrinks are good, everyone knows it. Important as hell, yes, an innovation in design, not really.

    Mildly off topic, but not really. Watch this video. This to me is the future of benchmarking reviews for whoever can get it right. Direct side by side A/B comparison in video, real time. That tells a story greater than any screen grab from XS ever could.

  24. Mirage Shrinks actually increase current leakage and consequently increase heat output in the modern processes. That is why Intel used high-k oxide and metal-gate transistors first. Chip design was the biggest challenge of the past and it is still as important but today's biggest challenge is process design. If AMD had been selling 45nm chips not since the beginning of 2009 but since late 2007, like Intel, their financial situation could have been much better today. They could have stayed as a "chip-maker" instead of transitioning into a "chip-designer" by spinning-off their fabs.

    See the following links
  25. Khaos
    Khaos Pete; can has i7 dual channel benchmarks for you tonight if you would like.
  26. Thrax
    Thrax (They're +/-2% of tri. Count on it).
  27. Cliff_Forster
    Cliff_Forster Mirage,

    As far as current leakage goes, it would generally hold true for the die shrink, but your right, when they hit 45nm, the space between transistors got so small that they needed to innovate using SOI, so good point.

    Still, point I am making, Die shrinks are something they are each chasing each other on. Its not a secret that its something that each firm wants to do for various reasons / cost/efficiency/product improvements.
  28. mirage
    mirage Cliff,

    Do you remember the recall of 1.13 GHz PIII Coppermine by Intel?


    It was because Intel decided to shrink the Al interconnects to 0.18 micron, while IBM and AMD were switching to Cu interconnects at 0.18 micron in 2000. Additionally, when IBM and AMD were using SOI with 0.13 micron in 2001, Intel never used SOI. However, Intel seems to be a quick learner and takes lessons from the past mistakes. They have used high-k and metal gate transistor most extensively. Although those are not Intel's inventions, implementation of these inventions at production scale is a huge achievement. Intel and AMD have very different cost/efficiency criteria because of the difference in their production models and volume. I am wishing/hoping that AMD will get back to full competition with their new production model. Phenom II X4 starting at a price below $200 is a spectacular choice for the consumer.
  29. oliver lets compare effiency.. how much money does intel spend every year?? how much does AMD put in?? with the billions burned by intel their products need to 3x faster but they aren't...
    like the fake story about NASA spending 1Mio. to invent the ballpen that works in space/no gravity vs. the russians using a pencil.
    i know why i prefer AMD; i dont need to spend much more money on an intel-sys, only to be 5% ahead of a Phenom. im talking about mixed-real-life-stock-clocks...
  30. Cliff_Forster
    oliver wrote:
    lets compare effiency.. how much money does intel spend every year?? how much does AMD put in?? with the billions burned by intel their products need to 3x faster but they aren't...
    like the fake story about NASA spending 1Mio. to invent the ballpen that works in space/no gravity vs. the russians using a pencil.
    i know why i prefer AMD; i dont need to spend much more money on an intel-sys, only to be 5% ahead of a Phenom. im talking about mixed-real-life-stock-clocks...

    Oliver, your a man that knows what he is talking about!
  31. SGT I have both Q9550 which was about $600 bucks 3 years ago and i also have a 955 Phenom II Both are great system but for Game play you cant beat a Phenom II Intel just cost to much just to own one even there Motherboard cost a lot and for what...tell me why intel cost so much it no better than AMD just look at the Intel Core i7 975 Extreme Processor it over $1000 give it about 5 year and you be able to get it for $200 or less technology change every day...But money in the Bank never change.....
  32. Snarkasm
  33. Cliff_Forster
    Snarkasm wrote:

    I'll interpret for our guest.

    Overpaying for an i7 is something only a knucklehead would do. ;D
  34. Snarkasm
    Snarkasm But his argument was more along the lines of "something so expensive isn't any good because in the future it'll be less expensive."

    Right. I stand by my "wut."
  35. _k
    _k I don't get it either. I bought a q9450 when a retailer got the first shipment, 400$ something. I don't plan to replace anything for another 2 years at least unless something blows up. It carried a big price tag but it runs 3.6 on air with a 1800FSB plus my 300$ is running my quad-sli. I guess I just over pay for everything I buy.

Howdy, Stranger!

You found the friendliest gaming & tech geeks around. Say hello!