AMD Phenom X4 9750 Quad Core CPU Review

[AkG]

​

AMD Phenom X4 9750 Quad Core CPU Review

​

Manufacturer Product Page: AMD Phenom? X4 Quad-Core Processor
Availability: Now
Price: Click Here to Compare Prices
Warranty: 3yr



When it comes to their CPU division AMD has been having a tough time of it lately. First, their long anticipated "native Quad core" which was only just released this year had a bug in it (aka the "TLB erratum"); causing many a long time AMD buyer (who had already held off on upgrading to 775 because of the "just around the corner" X4s) to flock in droves to the already mature Intel 775 Core 2 quad core technology. In retrospect announcing that they had a bug in their chips was probably just about the worst thing they could have done. Heck, by that point they may have been just as well off in postponing the Phenom release as releasing it in that condition to be hammered by Intel's marketing machine did them no good at all. Maybe it was the sudden rush of Canadian sensibilities via the ATI take-over which made the heads of AMD suddenly grow a conscience, or maybe not; whatever the reason and no matter how laudable their ideals, it was a devastating blow to an already rocky company who were both low on liquid assets and trying to merge two company cultures into one. However, above all else they no longer had the top end of the market (hurting already shaky morale) as the Q6600 from Intel was stomping all over even the fastest X2 they could make.

This perfect storm of tragedy definitely threw them for a loop and in my opinion what really hit them the hardest was confirmation they had overstated the upgradeability of early AM2 motherboards. In the early AM2 days AMD said most AM2 capable motherboards would be able to just drop in a X4 and work perfectly with just a BIOS upgrade, this was quickly debunked when the Phenoms were released as they require HT 3.0 (thus the new and “improved” X4 capable AM2 boards are called “AM2+”) to work best. This was the equivalent of a swift kick in the head to an already downed opponent and the future looked awfully bleak as even super die hard AMD customers felt betrayed, threw in the towel and went to the dark side (heck this is what did it for me as my super expensive ASUS motherboard would never run an X4 properly!).

However, AMD gave themselves a bit of breathing room to regroup and fix that horribly overblown bug on their B2 revision silicon that hardly effected anyone (yet the fix seemed to universally slam every X4 owner). Now like a street fighter who can take a hit, AMD has staggered to their feet, shaken off the past errors and have a new weapon to resume the fight with. The only question is whether this weapon is nothing more than a broken beer bottle (which is really a terrible weapon and your much better off using a UNbroken beer bottle as a bludgeon than a broken one as a cutting/stabbing implement) or is the geeky equivalent of a base ball bat and six "friends" similarly equipped?

Today we will be looking at one of AMD’s latest weapon in winning back the faithful who had lost their way and have been debating stepping into Intel's camp. The weapon that we are referring to is the bug fixed X4; or to be more precise the 2.4ghz 9750 X4 in its B3 revision. In this review we are going to put the X4 through its paces, both in Windows XP and Vista to see what it is made of and compare it against the venerable Q6600. We will of course be overclocking the X4 until is screams to not only make sure the TLB error is toast but also see what the new Phenoms are truly made of. One thing is for sure: it may not be a pretty sight at the end but we intend to see how effective this now TLB error free native quad core really is.

​

 

[AkG]

Specifications

Specifications

 

[AkG]

Fatal Flaw or Bad PR? Finally Fixed with B3.

Fatal Flaw or Bad PR? Finally Fixed with B3

​

As anyone who has worked in the computer industry knows, CPUs and their micro-architecture are always flawed. Most of these so called flaws or “errata” are so obscure that most never effect anyone in the real world; sure you can setup a scenario in a controlled environment that will highlight it but for the most part they are benign. Heck, most are the equivalent of saying: “stand on one foot, stick you other foot straight out, spin around in place while patting your head and tummy at the same time, all the while singing the Oh Canada!....backwards in Esperanto and while doing all this your ability to factor PI to 10,000 places may be negatively impacted.” Yeah, they usually are that vague and most every CPU has them. Take a look at the errata list for the Intel C2D and you will be astonished at the sheer number of these little nit-picks. Intel/AMD/etc are aware of them and simply fix them in future revisions and distribute BIOS level work arounds (or mirco code updates) to the motherboard manufacturers, and life goes on.

Unfortunately, not all errors are small, complex nor as benign as the example above; some errors are so mind-bogglingly huge and significant that the CPU maker issues a RECALL on all of them. The largest example of this was with the original Pentiums and their inability to do floating point math properly (in extremely rare and weird scenarios). Back then having a built in FPU (aka “math coprocessor”) was a big deal and yet Intel borked it so badly that ALL of those amazing (for the time) powerhouse chips were recalled. Yes an error in a mere 1 in 9 billion floating point divides was enough to recall them ALL.

Recently, AMD had a “major” error in their shiny new native quads with the B2 silicon that was probably much more unlikely to have any real world negative effect. This error was the translation lookaside buffer error or “TLB error” for short. To understand what a translation lookaside buffer error means, you first have to understand what the TLB does. This is going to be a cliff notes version of it, so please forgive us if we over simplify a very complex issue.

As I am sure you all know, most (if not all modern) Operating systems use “virtual memory” in which program A thinks it is accessing blocks A through D and program B thinks it is also has using these blocks, when in fact neither are necessarily using these memory address blocks. Virtualization of the memory allows multiple programs to all run happily together without the need for complex or messy memory managers to be built into said programs (rather its done at an OS level). What the OS does is store a table of all these mappings in memory and to make things even more quick the CPU can also store a copy of this table in its internal cache.

(Diagram courtesy of Wikipedia, used under the GFDL)
​

Now, since the lookup table is in cache its faster than accessing an external memory manager but searching this table (which can get LONG) and then using said table takes valuable clock cycles that could be used for other things, so what the CPU engineers decided was to make a buffer of the cache which has a list of the actual one to one mappings already done. The TLB is none other than this table of translations of virtual to physical mappings and since the mappings are already done the CPU doesn’t need to go do a cache lookup making it even faster again!

As we use the OS, programs are going to be swapped in and out of memory and thus the cache and TLB have to updated on a regular basis. This sounds simple but cache table entries have to be handled very carefully otherwise you could end up in a situation where two CPUs have different data….which is a bad thing and can result in a hard lockup of the computer, or even worse still silent data corruption! This is where the notorious TLB errata comes into play. Under certain circumstances the logic that handles how the complex dance of marking and moving and updating of the page table can go awry. In a nut shell, you could end up in a bad situation where two CPUs have conflicting lookup data in there L2 cache. At best this can result in a hard lock BUT the more sinister problem is silent data corruption. From an enthusiasts point of view things are made even worse as this error becomes more and more likely the faster the cores run. This is why AMD set the upper cutoff at 2.4GHz, for their first Phenom CPU release schedule.

The funny thing is that Intel found a similar bug in its C2D line last year but because they can do micro code updates on their chips (bug fixes in the CPUs logic. Think of it as being able to update the BIOS of the chip itself….’ish) they were able to handle it. Unfortunately, AMD does not have this ability and as such got skewered by the Intel marketing machine.

​

(diagram courtesy of University of Wisconsin CS 537 online lecture notes)
​

Because AMD was unable to do a machine code update they had to resort to a BIOS level kludge. This so called “fix” for all intents and purposes turned OFF the TLB’s ability to look in the cache. Needless to say this was the sledgehammer / nuclear bomb approach to fixing the error. Heck, some programs suffered some down right HUGE performance hits to the point where a cheaper and slower DUAL X2 would outperform it but even when you take the average of a 10% performance hit, that is one honkin’ big hit.

For the B3 line (aka new and improved version) which we are reviewing here today, it is a hardware fix which is still more of a workaround than a real fix as it does incur a performance hit. It is no longer a huge hit and is in fact very minor, but it is still a hit. The upside is the TLB can now access the cache properly and thus be a lot more efficient.

​

Well that is the past and this is now. AMD has fixed this error and hopefully, in time, will be able to recover from this devastating blow. Things do look grim but AMD is a fighter which has overcome longer odds in the past. So lets now take a close look at what AMD has brought to the table!

 

[AkG]

Up Close and Personal

Up Close and Personal

​

For a good portion of 2007 AMD were evangelizing to their faithful about the wonders the K10 architecture would bring “really soon now” and how it would bring a true “native” quad core to the desktop. Before we get into the nitty-gritty of what makes a Phenom, a Phenom lets dispel the myth that AMD were bull headed about not putting out a cut ‘n’ paste job like the wonderful Intel Q6600 was (in that it was nothing more than two dual cores crosslinked and put unto one chip, but because it was the FIRST quad core on the market it was by default the best). For a long while now AMD has had an integrated memory controller built right unto the die. This made for some amazingly low latency and efficiency levels, but the down side is to update the controller to take advantage of more advanced speeds (e.g. DDR2-1066) AMD had to update the whole CPU. Worse still from a quad core point of view is they couldn’t just slap two X2s together and call it an X4, as the two memory controllers wouldn’t play nice with each other. The additional engineering resources (and an external North Bridge) allowed Intel the ability to do this, but as AMD is a smaller company, it would more than likely have meant pulling people working on their native quad core. Should they have done it anyways? This is open to debate but it is too late now to worry about spilled milk.

​

As with the TLB errata, which has been beaten to death, we are just going to give you some of the important highlights on what makes the K10 (Phenom) a superior architecture when compared to the older and venerable K8 (X2). The biggest most obvious change is in the shear size of the K10; the new Phenom CPU weighs in at a huge 463 million transistors, and is a 11 metal layer chip. This does add to the complexity of manufacturing the Phenom but does make it a solid feeling chip in ones hands, one worthy of respect. 463 million transistors may seem like a lot but it still is less than Intel is using, of course most of those extra transistors in the Intel quads are taken up by the massive amount of cache and don’t really count. AMD knows they cannot compete against Intel in the onboard cache arena as Intel has more cash to spend (pardon the pun) and can simply outspend them on die shrinks (which frees up room for more and more cache for example).

​

When you get beyond transistor counts and chip complexity, the big thing to remember is that when AMD’s engineers sat down to design the K10 they didn’t throw the baby out with the bathwater so to speak and the X2s and X4s share some features and are still very different CPUs. For example, both the older X2s and newer Phenoms both have 128KB of L1 cache. Some X2s only have 512Kb of L2 cache, while others have 1024Kb; this was very confusing for consumers as you could easily get into a situation where two X2s running at the same speed were rated differently. To avoid confusion all Phenoms only come with 512Kb. The biggest difference when it comes to cache is the inclusion of relatively slow 2MB of L3 cache.

<img src="http://images.hardwarecanucks.com/image/akg/CPU/9750/diagram.jpg" border="0" alt="" />
​

One great improvement that was sorely needed was updating the HyperTransport protocol to version 3. Unlike HT 2 which operated at 5x multiplier of the 200mhz clock speed; HT 3.0 runs at either 9x multiplier up and down (though it is also referred to as "18x" as its 9x up and 9x down at the same time) or 10x multiplier (for twice the speed of HT2). As we said this extra boost in performance was sorely needed as anyone who tried sticking a Phenom in an older HT 2 compatible only motherboard knows, the 4 cores become starved for data under HT2 bandwidth restrictions! In the case of the 9750 which we are reviewing it uses the 18x multiplier as the 20x is reserved for even faster models.

As mentioned earlier the “Northbridge” or memory controller is built right on the die and once again AMD took the opportunity to improve its performance by giving it native support for DD2-1066. Also unlike the older K8 which had a single 128 bit wide controller, the K10s have two 64 bit wide controllers which can either be “ganged” together to make a virtual 128bit controller or you can leave them separate and operate independently of one another.

Another great improvement over the older architecture is SSE128 support. Unlike the older K8 architecture which could handle two 64bit SSE instructions per cycle the newer K10 can handle two 128bit SSE instructions at a time. On the surface this doesn’t sound like a big deal but you have to realize that when an older X2 is faced with a SSE128 instruction set it first has to decode it into two 64bit chunks and then run it. To put it another way not only can the K10 handle more data its is more efficient in how it handles it. SSE instructions are very prevalent in resource intensive / time constrained processes (like video encoding/ decoding for example) as they make things so much more efficient and quicker; and the inability to handle the newer and larger 128 instruction set was a huge handicap for the older X2s when compared to Intel’s C2D architecture. This is one area AMD had to step up and improve upon or be relegated to low end systems only. This on the surface is just one of the numerous changes that AMD did with regards to data capabilities, and some of the others include doubling the instruction fetch size for 16bytes per cycle to 32 bytes which when combined should result in one slick, fast and (hopefully) competitive CPU.

In general the K10 is an extremely well designed architecture that unfortunately only catches up with Intel’s older C2D quad line. Unlike when the X2s launched and devastated the competition (performance wise) the Phenom is more about evening things up than trying to surpass Intel. This is actually a good thing as Intel really did pull an Israeli rabbit out of their hat with the C2D. As we said earlier AMD is a fighter and when they needed to pull off an engineering miracle they dug in took their hits and came out with an awfully impressive design which should only get better with time and a die shrink.

 

[AkG]

Motherboard Options

Motherboard Options

Since AMD’s Phenom CPUs require a motherboard and chipset capable of handling the new HyperTransport 3.0 protocol, options are limited. This is where AMD dropped the ball (again) as they did for all intents and purposes tell early AM2 adopters that as soon as the new K10s (AKA Phenom) quads came out they would be a drop in replacement with just a BIOS update. I personally was burned over this and its still a sore spot for a lot of early adopters. Maybe it was wishful thinking, but in the end you are going to need a AM2+ motherboard with the appropriate chipset to run a Phenom properly, as most older AM2 boards just can’t handle them. Granted, some will but you will be missing out on some much-needed performance. Please take it on faith that the 9750 (or any Phenom) will not work in an older AM2 motherboard, if the quad is even recognized and boots up it will run like the proverbial dog. Phenom’s need the extra bandwidth HT 3 affords and it would be a lot like putting an elevator in an outhouse.

That being said, in this section we will be looking at some of the motherboard and chipset options which go hand in hand with the new Phenom processors.

Please Note:

We are very hesitant on lower end chipset based motherboards and this monster 9750 Phenom. We have nothing against the chipset per say, rather term like “entry level” or even “mid grade” and a powerful quad like the Phenom may end in disaster as you saw happen to us! To this end we will not be giving examples of the motherboards for these chipsets nor going over their pros and cons. You can consider their inclusion in this review for "information purposes only".


And in the Green Corner

No matter their track record with some circles, Nvidia have some great chipsets that will run the Phenoms properly. On the green side the 700a series (750a & 780a for example) are very good chipsets that allow you to get your SLI fix.

For all you wondering what is the difference between the 730a, 750a, 780a, and the newer 8200/8300s, the short answer is: not much. Technically speaking, the 730 is the entry level, the 750 is the mid grade and the 780s are the enthusiast level; and the 8200/8300s are a newer single chip solutions which have an integrated GPU built right on the chip.

In reality both the 720, 730 and 750 chipset all share the same Hybrid SLI features through an integrated GPU, but only the top dog has mult-SLI support, the mid grade supports SLI and the entry level has only one PCI-E slot for Hybrid SLI only. To put it another way; the three chipsets differ in how many PCI express lanes each has and how many graphics cards you can stick in it (though even the lowest end PCI 2.0 is still has more than enough bandwidth for any single core graphics card out there). Of course, motherboard manufactures may also add more features to the higher end boards than a lower end but this is to be expected.

This of course is glossing over the details of the various Nvidia chipsets, so to help you decide which chipset is right for you here is a more detailed breakdown of their capabilities:


730a

HyperTransport 3.0 Support
NO SLI technology support
PCI Express 2.0 support
1 x16, 3 x1
Up to 5 PCI Slots
Up to 6 Serial ATA 3.0 Gb/s drives
Up to 2 PATA drives (Ultra DMA 133/100/66/33)
RAID configurations 0, 1, 0+1, 5
12 USB Ports


750a

HyperTransport 3.0 Support
SLI technology support with two identical SLI Ready graphics cards
PCI Express 2.0 support
19 lanes, 4 links: 2 x8, 2 x1 or 1 x16, 3 x1
Up to 5 PCI Slots
Up to 6 Serial ATA 3.0 Gb/s drives
Up to 2 PATA drives (Ultra DMA 133/100/66/33)
RAID configurations 0, 1, 0+1, 5
12 USB Ports


780a

The 780a series by Nvidia is for enthusiasts who maninly want the biggest and the best when it comes to graphics (mulit SLI) and performance. For many years Nvidia was the only real choice for AMD enthusiasts and as such this chipset has a long and rich pedigree behind it. However, as it requires an additional chipset (namely the nForce 200) and a crossbar link to get all the features and it is getting a little long in the tooth in comparison to some of the new single chip solutions (especially in the heat and power consumption arena). However, if SLI is your main deciding factor it is the chip to own! Depending on the motherboard you chose your overclocking may only be limited by your CPU (or ram in some cases) and not due to this chipset. Are there reports of some glitches with this chipset and overclocking? Yes, but its many positive features do make a very strong case for it.

Basic Specs:

HyperTransport 3.0 Support
Quad SLI technology support with two dual-GPU NVIDIA graphics cards
3-way SLI technology support with three single core GeForce graphics cards
SLI technology support with two identical SLI Ready graphics cards
PCI Express 2.0 support
35 lanes, 7 links: 1 x16, 2 x8, 3 x1
Up to 5 PCI Slots
Up to 6 Serial ATA 3.0 Gb/s drives
Up to 2 PATA drives (Ultra DMA 133/100/66/33)
RAID configurations 0, 1, 0+1, 5
12 USB Ports


8200 / 8300

​

While technically the 8200/8300s are geared towards mainstream consumers who normally want more than integrated graphics offer but at the same time don't need the horsepower of a high end GPU either. This means that alot of HTPC enthusiasts are starting to take a closer look at this solution.

Even though it is a low to mid range chipset it still does pack a lot of features which we have come to know and love from Nvidia's AMD oriented chipsets. The combination of power and the ability the SLI with just one standalone board (aka Hybrid SLI) does make it a very good choice for consumers who are not interested in monster overclocks, high end gaming, etc. but are interested in a stable, all round good choice which leave them enough money in their pockets for other things (like more hard drives, better CPU, etc). For anyone who wants to know why we have lumped the 8200/8300 together it is because the only real difference is the speed/ power of the integrated GPU. In many ways you can consider the 8300 the successor to the 8200 or a higher clocked version of it at least!

Basic Specs:

HyperTransport 3.0 Support
Built in Video (DVI, HDMI, VGA)
Hybrid SLI (onboard + graphics card)
PCI Express 2.0 support
1 x16, 3 x1
Up to 5 PCI Slots
Up to 6 Serial ATA 3.0 Gb/s drives
Up to 2 PATA drives (Ultra DMA 133/100/66/33)
RAID configurations 0, 1, 0+1, 5
12 USB Ports


And Fighting in the Red Corner...

On the red (well, green too) side you also have another 700 series chipsets which are Phenom capable which allow you to Crossfire to your hearts content (770, 790X, 790FX for example). Once again the differences between the chipsets comes down to how many PCI Express lanes they have. 790FX is for the enthusiast who want to have loads of GPUs in their system, the 790x is more mainstream in its approach for people who may experiment with Crossfire but are probably only going to run 1 of the 2 GPUs allowed, and the 770 is you entry where Crossfire is not supported. The new kid on the block is the 790GX with the SB750 chipset which is the stand-in until the SB800 chipset is released.

Once again to help you decide which chipset is right for you here is a more detailed breakdown of the AMD chips capabilities:


790FX

​

The 790FX can be considered the premier ATI based chipset for the Phenoms. It would be very surprising to see any motherboard with this chipset not be Phenom 125W TDP (or even 140W) compatible as it is the heart and soul of the "Spider" platform. The 790FX really is the chipset to own if you are an enthusiast looking for the latest and greatest AMD has to offer.

With all this being said (and when compared to Nvidias line up), the 790FX really is a great low power, low heat choice which has unfortunately been paired with the older 600 series southbridge. Because of this uneven paring the 970FX does suffer from lack of RAID options and even lower USB capabilities. With the release of the SB750 this situation will shortly be fixed, so you may want to pay close attention in the coming months to what the manufactures are going to be offering.

The higher end 790FX based systems you have a lot more freedom, as the DFI we use is very good, and so is the Gigabyte DQ6 and the various ASUS options you have.

Basic Specs
HyperTransport 3.0 Support
PCI Express 2.0 support
Maximum four physical PCI-E x16 slots and discrete PCI-E x4 slot,
the chipset provides a total of 42 lanes solely in the Northbridge
1 xUltraDMA 133/100/66
Up to 6 Serial ATA 3.0 Gb/s drives
RAID configurations 0, 1, 10
10 USB Ports


790X

​

Unlike it bigger brother the 790X is more of a middle of the road / consumer grade chipset. It is still a great choice if you are not concerned with quad Crossfire; it does support Crossfire but its capabilities (i.e. the number of lanes it has) is severely curtailed. In fact the only issue we have with the 790X is earlier motherboards (and a some of new ones) can not support the 140W TDP Phenoms, nor even the 125s.

Based on our experience 790s boards, compatibility is hit or miss, but I personally would try the Gigabyte DS4 790x if I was forced to use (another) 790x + 125W TDP Phenom combination. As we mentioned earlier this is a situation AMD is aware of and we are sure they will get it sorted out with the manufactures...we hope.

Basic Specs:
HyperTransport 3.0 Support
PCI Express 2.0 support
CrossFire Support
Two physical PCI-E x16 slots
1 xUltraDMA 133/100/66
Up to 6 Serial ATA 3.0 Gb/s drives
RAID configurations 0, 1, 10
10 USB Ports


770

As with the 730a/750a from Nvidia we would be hesitant to recommend any motherboard as even higher end boards have troubles with 125TDP Phenoms. Also like the 730/750 the basic specs have been included for information purposes only.

Basic Specs:
HyperTransport 3.0 Support
PCI Express 2.0 support
NO CrossFire Support
One physical PCI-E x16 slot
1 xUltraDMA 133/100/66
Up to 6 Serial ATA 3.0 Gb/s drives
RAID configurations 0, 1, 10
10 USB Ports


780G

​

The 780G is a consumer grade chipset aimed directly at the same people the 8200 / 8300 chipset by Nvidia geared towards. It is one of the most popular chipsets for AMD's motherboard partners to use as an SFF (Small Form Factor) product. In the last few weeks we have seen everything from mATX to ITX boards using this chipset to great effect. Since it offers onboard graphics, Hybrid CrossfireX support, decent HD playback capabilities and an optional HDMI connector, many HTPC enthusiasts have been gravitating towards this chipset. To put it another way, not everyone is interested in multi-GPU setups and just want a good motherboard which has great feature yet doesn't break the bank and that describes the 780G perfectly.

Unfortunately, the 780G chipset is an older chipset and as such does have a lot of issues with the higher TDP Phenoms. Reports of high end Phenoms not working properly with the 780G was so prevalent AMD actually came out and stated the 780G was not intended for use with them! Here is the now (in)famous quote by Jack Whitman:

"What people have done, mistakenly, is paired a 780G (chipset-based) motherboard with the higher frequency Phenom--the 125-watt Phenom....They've taken an enthusiast-class quad-core part and paired it with a mainstream motherboard . . . And not all motherboard manufacturers have tweaked their boards to support a 125-watt TDP....We've never made claims that 780G motherboards are enthusiast-class motherboards."

Ouch! Though the good news is some motherboard manufactures have stepped up to the plate and delievered rock solid 780G motherboards which DO support the Phenom 9750! A great example of this is the M3A78-EM by ASUS.

Basic Specs:
PCI Express 2.0 support
One physical PCI-E x16 slot
Side-port memory as local frame buffer (DDR2 or DDR3)
1 x UltraDMA 133/100/66
Up to 6 Serial ATA 3.0 Gb/s drives
RAID configurations 0, 1, 10
12 USB Ports


790GX

​

The motherboards based on the new AMD 790GX / SB750 chipset are mostly geared towards the enthusiast crowd who want a full-featured heart to their new computer while having a modest onboard graphics processor. While this may be the new kid on the block, features like Hybrid CrossfireX (which uses a discrete GPU in combination with the integrated GPU to further increase 3D performance), standard CrossfireX and HDMI-out have made people stand up and take notice. These boards are AMD's answer to the Nvidia-based products which support SLI with AMD processors.

Basic Specs:
HyperTransport 3.0 Support
Hybrid Crossfire
PCI Express 2.0 support
Two physical PCI-E x16 slots at x8 bandwidth
Side-port memory as local frame buffer (DDR2 or DDR3)
1 xUltraDMA
Up to 6 Serial ATA 3.0 Gb/s drives
RAID configurations 0, 1, 0+1, 5
12 USB Ports


Unfortunately, this is not the end of the discussion when it comes to motherboards and Phenom’s. It seems that there was a communications breakdown between AMD and the various motherboard manufactures as not all motherboards (regardless of chipset used) are able to handle the loads the 125W TDP Phenoms would place on them. In the past it was fairly easy to pick the proper motherboard for your needs, as you just needed to figure out if you need the “basic”, the “mainstream” or the “enthusiast” class chipset and then picked the motherboard with the right one (which also fits your budget). Well, those days are long gone when it comes to AMD systems, as not all motherboards are created equal, and even some higher end 790X based motherboards are not rated to handle the high end Phenoms.

 

[AkG]

R-E-S-P-E-C-T

R-E-S-P-E-C-T

​

To put this powerhouse through its paces we selected MSI’s 790X - K9A2 CF motherboard, this motherboard is a 4 phase (really a 3+1) which uses the older 4pin 12v aux connector. Even thought it is rated for 125W TDP CPUs our 9750 popped two of the 4 MOSFETs like they were Chiclets after only a few minutes or so of use. Believe it or not, Windows was not even fully installed, it got to the first reboot portion and upon reboot there was a pop, at which point we KNEW what that sound meant and jumped up to turn off the PSU. Unfortunately, before the switch could be flipped, a second pop occurred and a 4 inch flame erupted from one of the MOSFETs. Luckily this system was in an open bench so putting out the fire was relatively easy. Lesson of the day: “When Thou Art Dealing With Phenom’s Thou Shall Not Skimp On Thy Motherboard!”

After this occurrence we rethought our AMD AM2+ test bed setup and decided to go with a more heavyweight board. We took a long hard look at the various motherboards that would fit our wants/needs/requirements and the deciding factor came down to BIOS options. While there are many great companies making excellent AM2+ motherboards (including MSI) one stands above the rest when it comes to shear number of possibilities available to an enthusiast in its BIOS. We of course are alluding to none other than DFI; DFI has a well deserved reputation for offering the most complete BIOS options of any company.

​

To be more specific on which of their mainboards we went, with we went for the full meal deal version of the their 790FX chipset based motherboard line. The DFI LANPARTY UT 790FX-M2R is a wonderful board with a great chipset and we were very honored to have its horsepower made available to us. It is too bad that our test sample 9750 was not a Black Edition (i.e. it has a locked multiplier) as this DFI board would more than likely have been able to eek out a higher OC if we were able to do more than adjust the HTT and voltages.

​

For most tales of woe this would be the happy ending; as this is reality, and Reality/ Karma /Mr. Murphy has a fond love of kicking you when your down this was just the end of the beginning. For anyone who has read our recent review of the OCZ Vendetta 2 CPU cooling solution already knows, we had selected this cooler as a our AM2+ test bed cooler. It is a great cooler in its own right and when paired with a Noctua NF-P12-1300 fan it becomes a truly legendary performer…of Intel Quad 775 CPUs.

With this combination at stock speeds and voltages our 9750 idled in the mid 30s and topped out around 65°C +/- a few degrees. This while hot, was still acceptable as we of course had disabled Cool n Quiet and the Phenom’s four CPUs were running full speed all the time and most importantly ambient temperature was 25°C. Then we began our first round of Overclocking and things took a real turn for the worse.

At anything above 2.5GHz the system would run fine and then during stability testing shutdown (no blue screen, no errors, just a hard reboot). At first we just thought this chip was a terrible overclocker and increased the juice slightly. However, instead of making the system more stable this gave the appearance of making the system LESS stable as it would crash even faster. This was a truly “Eureka” moment and we went back into the BIOS and changed the CPU Overheating protection from 75°C to disabled.

​

One of the biggest problem with the Phenoms and overclocking is that unlike Intel C2D’s, Phenom’s just give an approximation of their temperature; thus Core Temp only shows one temperature and in our case it was undervaluing this temperature (this is a known issue with Phenoms so while we expected it to be off we were expecting it to be at least close!).

While we were in the process of tweaking the BIOS and removing the overheating safety we also decided to beef up the cooler as the Vendetta 2 + Noctua was clearly not able to handle the thermal loads. We knew the Vendetta 2 had the potential so instead of swapping it out we rummaged around in the Ye Olde Parts Bin and found a 120x38mm Panaflo 103CFM fan which works fine except it has a blown PWM and thus is either ON or OFF. We also took the precaution of lowering ambient temp back to 20°C. This potent new combination lowered stock temperatures to the mid 20s and low 50s for idle and load temperatures respectively. Armed with our new and improved Vendetta 2 we restarted our OC’ing phase with renewed hope and vigor.

You can read how our OC’ing went in the Overclocking Performance Results section. Before we continue we would like to point out that these Phenoms are truly heavy weight performers which unfortunately gives us flash backs to the days of certain 930 Ds with it’s fireball predisposition. Quite literally the thermal load was so high during initial overclocking that the top of the heatpipes became discolored and mottled in appearance. This tells us that these heatpipes were pushed to their limits and maybe even a little past their limits!

 

[AkG]

Testing Methodology

Testing Methodology

When it comes to picking a new CPU, especially when said CPU requires a new architecture to run it properly, there are a few variables that we like to use before making our final decision. These variables roughly translate into the following categories: Performance, Game Performance, Energy Efficiency, Thermal output, Overclocking. As the level of importance of each of these variable will depend greatly on what you intend to use your computer for (e.g. game system, folding system, HTPC, etc.) we have broken these variable up into their own sections so you can give more or less weight to each of them as you see fit.

For the overclocking section we disabled spread spectrum, disabled cool and quiet and we set the maximum vCore of 1.45volts. This is very conservative but it is what we use for C2Ds and we want to make the playing field as even as possible. NB core voltage & HT voltage was also upped as necessary but kept as low as possible as we were not interested in doing any “suicide runs”.

For testing a given overclock had to be stable for 2hr of Prime95 “large fft” to be considered successful.

Since we will be comparing multiple Intel and AMD CPUs, we have taken steps to minimize external variables that could adversely effect the results. The first and biggest step is only 1 AM2+ and 1 Intel 775 motherboard will be used. For Intel CPUs we will be using a Gigabyte P35-DS4 motherboard. For all AMD tests a DFI LANPARTY UT 790FX-M2R will be used. The same RAM, Hard Drive, Video Card, CPU cooler, open rack (case), Power Supply and even DVD player will be used. All CPUs will have new virgin XP SP3 OS installed on them and then virgin state Vista Ultimate 32bit SP1 OS installed on them. All tests will be run 4 times with only best results used.

Complete Test System:

Processor:
Intel:
Q6600 & E4600
AMD:
Phenom 9750 & X2 5000 Black Edition

Motherboard:
Intel:
Gigabyte p35 DS4
AMD:
DFI LANPARTY UT 790FX-M2R

Memory: 4GB Mushkin HP2 PC6400 (5-5-5-15 2T @ 2.0v)
Graphics card: ASUS 8800GT TOP Edition
Hard Drive: 1x Western Digital Se16 500GB
Power Supply: Seasonic S12 600W


Please note: The Q9450 used in some of these benchmarks is for illustrative purposes only. This review is a “battle of the mid range quad cores” and it is only included to show what a higher end Intel CPU is capable of.

 

[AkG]

SuperPi / wPrime

SuperPi

SuperPi calculates the number of digits of PI in a pure 2D benchmark and can do so upto 32M places. For the purposes of this review the seemingly standard 1M and 32M test runs will be performed. For many years now this program has the measuring stick for many an enthusiast. It may be getting long in the tooth but it is still a good way to judge how efficient your system is as many factors from CPU speed & efficiency to memory speed and timings can have a direct impact on the results. As with all tests our 4GB (2x2GB) Mushkin HP2 PC6400 Ram was in its default 5-5-5-15 2T for all systems.

​

AMD processors are notorious for needing tight as possible timmings to get the best performance out of them and the default of 2T timing really does hurt their performance and doesn’t do much to show off their integrated memory controller capabilities. Though loose timings or no we were expecting the numbers to be a lot closer than this. If we were running either of these AMD processors we certainly would take the time to tighten up the memory timings as that would certainly pay dividens. That being said AMD should know that most of their customers neither have the will nor the inclination to do this and should really have taken the time optimize their new memory controller for the general public.

wPrime

Unlike SuperPi, wPrime is a truly multi-threaded benchmark. The aim of the authors was to max out all you cores, regardless of the number by first determining how many cores your system has via CPU-Z and then using that number for the number of threads it will run. For our purposes both the 32M and 1024M tests were run on each of our tests CPUs.

​

Now these scores are more like it! The Phenom 9750 easily beats a Q6600 and can even give a q9450 a run for its money. When it comes to midrange quads and wPrime AMD Phenoms are the clear winner.

 

[AkG]

Cinebench 10 / Lame MP3 encoding

Cinebench R10

Another benchmarking community favorite, Cinebench renders an intense 2D scene relying on all the processing power it can. Cinebench R10 is likely the most efficient program tested today at utilizing all cores of a processor. We will be running both the single threaded and multi-threaded benches here today.

​

The Intel Core 2 Duo architecture is a very refined and mature line and its superb performance reflects this fact as it easily trumps the new K10 AMD quad core in both single and multi-threaded rendering. In time as AMD enhances the Phenom core this may change; but as things stand right now AMD needs to work on improve its efficiency of the individual cores and their ability to interact with each other.

LAME MT

Lame MT was developed to be a multi-threaded version of the venerable LAME audio encoder to show off the power of Intel’s HTT (Hyper Threading Technology). Of course, it shows off the power of multi-cores processors even better! Unfortunately, it really is only capable of using 2 cores thus making it like many modern programs that can’t really take full advantage of a quad core CPU. Even with this limitation it still does show the real world performance of a given CPU and how they relate to other CPUs. For this benchmark a 100MB .WAV file was converted into an .MP3. The full command line was as follows "lame -q 0 -V 1 test.wav test.mp3".

​

While 26 seconds is not exactly terrible….for 2005. Of course that is a HUGE exaggeration but by today’s standards the Phenom’s audio conversion performance is just…well LAME! Once again you do have to take it with a big grain of salt as the Phenoms are a brand new architecture and in time they will mature to the point where they will be true contenders.

 

[AkG]

WinRAR / 7-Zip

WinRAR

The popular compression program winRAR includes an internal benchmark program that measures the performance of its decompression and compression algorithms. It does this by using an intensive worst-case scenario set of data. This benchmark is great for comparing processors of different size, speed and even cache sizes as it does not scale linearly with number of transistors or cache that a given chip has. WinRAR 3.8 beta 3 was used and it was allowed to run for 5 minutes before results were recorded.

​

Ouch. As is becoming a reoccurring theme the Phenom and AMD processors are just being left in Intel’s dust. Even taking the q9450 out of the picture 1616 vs 1222 is a down right monstrously huge performance gap. One can only hope that AMDs upcoming die shrink will include a processor efficiency tune up.

7 ZIP

As with WinRAR 7-ZIP comes with a n internal benchmark program that measures the performance of its decompression and compression algorithms. All settings were left at default except for number of passes which was set to run for 8 full passes. Only after 8 passes were the results recorded.

​

Just when things were starting to look bleak for AMD 7-Zip comes to the rescue; or at least partially to the rescue. The 9750 Phenom’s decompression ability manages to beat the q6600. In time its lagging compression score will also come improve in a revision or three!