Athlon Socket AM2 Review

NerdMaster writes "Hardware Secrets has just posted an Athlon 64 X2 5000+ review, one of the first AMD CPUs to support the new socket AM2. It runs at 2.4 GHz, has two 512 KB L2 memory caches (one for each core) and supports DDR2 memories." However, many are still predicting an end to AMD's dominance in the market thanks to Intel's Conroe.

Comments

[offtopic] 'Print' version is split into 10 pages.

[Whiney Mac Fanboy]

Hardware secrets - you suck.

10 pages not saying very much is (irritating, but) acceptable, but when you split the print article into 10 pages, you've crossed line from greediness to stupidity.

(fires up IE). Oh nice, and there's advertisments on each of the print pages too. How is that supposed to be printable?

Uneven Benchmark

[Anonymous Coward]

According to the last page of the review, they gave the Socket 939 platform 2GB of ram, and the AM2 platform 1GB.

Re:Uneven Benchmark

[kjs3]

Apparently, saving the allowance his Mom gives him hasn't added up to the $75 or so needed for another 1GB of DDR2.

Upgrade?

[Enderandrew]

The 939 NForce 4 Ultra/SLI line is showing age so on one hand, I'm glad to see the new tech (and DDR2 support) finally.

However, I don't know that I can convince my wife to let me spend the money on such a large overhaul again. I'm fairly happy with my AMD 64 system at the moment.

Honestly, I just hope AMD maintains their lead long enough for people to start taking notice (like Dell using AMD in the server line).

Re:Upgrade?

[Ihlosi]

The 939 NForce 4 Ultra/SLI line is showing age

Does that mean it has finally reached some degree of maturity ?

And sorry to disappoint you - since the chipset-cpu interface remains the same, the "old" chipsets can be used for AM2 processors, as long as the mainboard has an AM2- and DDR2-sockets.

Re:Upgrade?

[Enderandrew]

The chipset itself is showing age, not necessarily your PC.

Early SLI offerings leave much to be desired in efficency and design. Furthermore, the NForce 4 series was innovative in features, but since so many of these features were new, they were buggy or weren't implemented horribly well.

We're due for a more stable NForce 5 series.

Ignore these benchmarks

[Enderandrew]

They actually tested the 939 rigs with 2 gigs of memory, and the new rig with 1 gig of memory. They said they didn't have 2 sticks of DDR2 for the AM2 rig, but then they should have only used 1 stick in the 939 rig.

When benchmarking, you should try to keep all test systems as comparable as possible. I really am disappointed by what I consider a glaring oversight.

Seriously, for shame.

Re:Ignore these benchmarks

[Barny]

The testing of one rig with 2 sticks of 1G ddr ram (2x64bit bus, accessed in dual channel mode so 128bit) is VERY much inacurate against a rig running one stick of 1G (1x 64bit bus, can only be accessed in single channel, 64bit, mode), and I believe that is a serius error.

However, AMD is going to lag a little behind in the short term, although the memory latancy is still allmost half of what intels current parts are capable of (check out theinquirer.org for some interesting insights into thise issue).

Re:Ignore these benchmarks

[Enderandrew]

The thing that really got me was the conclusion page. For many people, that is all they read. In said conclusion they said the DDR2 support wasn't necessary that great because the FX-60 beat out the 5000+ in several test.

In all the CPU specific test, the 5000+ came out on top. In tests that benchmarked total system power, the FX-60 rig with 2 gigs running in dual channel mode came out on top.

If these guys don't know how to put together proper testing rigs, let alone extrapolate results, then AMD shouldn't be wasting test equipment on them. They can send processors to me, and I'll put up some serious benchmarks, including detailed overclocking, burn-in testing, temperature readings, etc.

Re:Ignore these benchmarks

[MadRocketScientist]

From TFA, Page 3: Memory: Two Corsair CM2X512-8500 modules with 512 MB each, installed on DDR2 Dual Channel configuration (using 4-4-4-12 timings) for socket AM2 CPUs.

Re:[offtopic] 'Print' version is split into 10 pag

[Shadows]

Summary: Blah blah, blah, blah blah blah... and finally something interesting:

[from page 10]

We didn't see any performance improvement by the use of DDR2 memories instead of DDR. In fact, Athlon 64 FX-60 was faster in several situations, showing that at least for the software we used having a bigger L2 memory cache is better than having DDR2 memory instead of regular DDR.

But, on the other hand, we have to consider the future and what is behind AMD's strategy in going to AM2 socket and DDR2 support. In our opinion, what is important isn't the release of Athlon 64 5000+, but the possibility of increasing the memory bus clock rate in the near future. For socket 939 platform this was not possible since the top DDR official standard was DDR400/PC3200. With AM2 AMD can finally go beyond that, as DDR2 official standards include at least three speed grades above that: DDR2-533, DDR2-667 and DDR2-800. That is promising.

Welcome, Intel

[eebra82]

I for one welcome Intel to the top performance game. I used to be all happy about AMD gradually taking over the market ever since their first Athlon slot CPU:s. This has also caused the pricing war significantly since AMD already knew they kicked Intel in most parts of all reviews.

Now that Intel is back, we can finally see some heavy competition between the two. The Core Duo is a superb processor and I am eagerly awaiting my MacBook to arrive and I can't wait to see the second release of the Core Duo.

Remember what it was like a few years ago? I used to follow the price charts of CPU:s for drops and they were a lot more frequent than they are today. So now it's easy to say that we should get the same competition all over again and I am quite sure that Hector Ruiz at AMD has a backup plan ready to be enrolled this year.

So once again, welcome, Intel!

Re:Welcome, Intel

[chrismcdirty]

You feel like you're overpaying for Intel, but you didn't ever feel like you were overpaying for PowerPC in the past? I have a feeling that if Apple had used AMD chips, you'd still be paying the same price, but more of it would go directly into Apple's pockets.

Re:Welcome, Intel

[eebra82]

There are quite a few of them out there. Here's one for CPU:

http://www.sharkyextreme.com/guides/WCPG/article.p hp/3607116

And here's one for RAM:

http://www.sharkyextreme.com/guides/WMPG/article.p hp/3607991

Re:Welcome, Intel

[NutscrapeSucks]

You feel like you're overpaying for Intel, but you didn't ever feel like you were overpaying for PowerPC in the past?

Speculation was that Apple paid under $50 for a G4 CPU.

For a large OEM like Dell, I doubt there's any huge difference between Intel and AMD pricing. Apple however is pretty much only using the luxury Core Duo parts, so they are probably spending a lot more money on CPUs and saving it elsewhere by using Intel chipsets and integrated video. If component costs were really a huge concern for Apple, they'd be shipping boxes with Celerons and Pentium-Ds. But with 20-30% profit margins, who cares?

Re:Welcome, Intel

[spectrumCoder]

I'm aware of price watches, which give the week-on-week price changes for components, but what I believe would be really handy is a graph of the price of any given component over time (with a time range of anything from three months to three years).

One could then analyse the steepness of the curve and come to some conclusions, such as the best time to buy a graphics card is nine or twelve months after release, LCD panel prices are unlikely to experience the same rate of decrease in the coming twelve months than in the previous twelve months, etc. (these statements are just examples of the possible sorts of conclusions)

I'm guessing that CPUs have a reasonably shallow curve (since the product lifespan is longer than the typical motherboard or graphics card), but I'm not sure.

Re:Welcome, Intel

[default luser]

but what I believe would be really handy is a graph of the price of any given component over time (with a time range of anything from three months to three years).


Happy Birthday!

Anandtech's Real-Time Pricing Engine gives you real-time prices, plus a graph of the last 6 months of price changes.

I'm amazed at the fact that they log 6 months of history, and offer it to the public at no charge. Most accounting firms charge through the nose for this kind of pricing trend data.

I'm guessing that CPUs have a reasonably shallow curve (since the product lifespan is longer than the typical motherboard or graphics card), but I'm not sure

In recent years, you'd be correct. The x86 CPU market has come up against a wall, because you can only parallelize scalar x86 code so much on-the-fly. Improvements in CPU architectures like Conroe now take MUCH longer to design, for less and less performace gains than in previous years. Also, the percentage megahertz gain from process improvements is also hitting a wall, which is another reason CPU prices remain stagnant.

HotHardware's Coverage Of AM2, Much More Detail

[Anonymous Coward]

The folks over at HotHardware.com also have a very detailed performance evaluation of AM2 Athlon 64, right here

Problems with memory dividers

[Delph1]

What many seem to have neglected is that there is an evident problem with odd CPU multipliers. AMD has no support for "half" memory multipliers (4.5/5.5/6.5/etc), which means that you will actually not be able to run memories at their full potential when using processors with odd multipliers (7/9/11/etc):

http://www.nordichardware.com/Reviews/?page=3&skri velse=481

Buying PCs isn't as exciting as it used to be

[Anonymous Coward]

My 1.66GHz Athlon XP 2000+ does the job well and is still very often limited by memory (1GB) and harddisk bottlenecks. That's a 3 year old system, and CPU+Board cost only $140 back then. Am I really expected to pay that price several times over to replace a reliably working system and get less than a 2x performance increase?

Re:Buying PCs isn't as exciting as it used to be

[ArcherB]

My 1.66GHz Athlon XP 2000+ does the job well and is still very often limited by memory (1GB) and harddisk bottlenecks. That's a 3 year old system, and CPU+Board cost only $140 back then. Am I really expected to pay that price several times over to replace a reliably working system and get less than a 2x performance increase?

Where do you get 2x performance? Are you just looking at the clock speed? I ask because my A643000+ is a good clip faster than my AXP3200, both at the same clock speed.

HardOCP review, link to print version

[Shivetya]

http://enthusiast.hardocp.com/articleprint.html?ar t=MTA2NQ==

Much better than the posted story which is nothing more than an advertisement for Hardware secrets.

When will /. editors review entries to prevent this abuse? If anything, when new hardware is released we all know multiple sites will cover the release. That means, put links to the more popular review sites into the story instead of helping one person get his ad hits.

Much More benchmarks are available on the web

[Anonymous Coward]

Small selection:
- HardOCP
- X-Bit Labs
- HotHardware

up to date list here

am2 seems to be doing fine

[mAriuZ]

Here is another review of the new socket/ddr2 amd part
It's great to look at fx-62 results - it looks like only that processor (or if you overclock it) can
use the available bandwidth
        "Frankly speaking, it's the main competitor who must be bustling now. AMD is doing great anyway. At least in terms of CPU performance. Durability of the K8 core and its capacity to adapt to new market realia is admirable: having lived without major modifications through two process technologies, dual cores, and now a new memory controller, this core meticulously responds to each improvement with performance gains. We were very skeptic about future chances of the new AMD platform against the new processor core from Intel (Intel designed the new core nearly from scratch, while AMD K8 is rather old), but our tests warmed up our interest. The situation may turn out not that simple"

AMD Catches Up in Technology and Shoots Out in Performance
http://www.digit-life.com/articles2/cpu/amd-athlon -64-fx-62.html/

My review

[Anonymous Coward]

This architecture is a pretty big disappointment.

Sure, Socket 939 was amazing when it came out. Nine hundred and thirty-nine pins -- quite an amazing figure.

But I'm sure I wasn't the only one who was less than thrilled at the release of Socket 940. I mean, just one more lousy pin? That's only a tenth of a percent improvement on Socket 939! One wonders what AMD was thinking.

And now here we are, with this farce of an architecture. Yes, Socket AM2 has only 940 pins as well! All that AMD has done in all this time is to shift the pins around on the CPU, much as the occupants of the Titanic spent their time rearranging their deck chairs -- even while that one guy kept on yelling "Iceberg ahead!"

Of course, AMD has tried to hide their laziness with the snazzy marketing name AM2. And yeah, I have got to admit that the name sounds pretty damn good. But in the end, isn't the socket itself more important than a mere name?

Re:Intel's roadmap to the Cornroe and beyond

[Enderandrew]

It really isn't fair to compare the next-gen Intel offerings (just now coming out) against the three year old K8 core which is what people are doing. AMD's next gen processors are rolling out this summer. Let's try to compare apples to apples.

When we compare apples to apples (the P4 line to the K8 line) Intel actually uses more power, generates more heat, runs slower, and is more expensive.

I'm glad Intel stepped up and made a good processor. The Core Duo is a good processor, don't get me wrong. But for all the marketing buzz about Intel right now, AMD does in fact have a response.

AMD is not only rolling out Turion X2 series processors, but they've got the AM2 lineup such as the 5000+ X2, the quad core series, etc.

Re:Socket consideration

[GiMP]

Don't forget the new memory. Usually they tend to change the memory when they change to socket as well, and the dual-memory boards are usually from PC Chips or other quality manufacturers, like ECS. Right now, my basement is full of EDO, SDRAM, DDR, DDR2, etc. Some ECC, some not, some motherboards don't work with ECC, some don't (officially) work without ECC. I have some older and more exotic memory as well.

Am I complaining? Not much, I understand it is necessary to improve the architecture. Nn the other hand it isn't fair to say that a processor upgrade is only a processor and motherboard. It is often the case that new RAM is required as well.

Re:Intel's roadmap to the Cornroe and beyond

[powerlord]

Exactly. The other thing that most people miss is the strengths for each player.

Even if the Core Duo turns out to be better than the comparable AMD chip, Intel still has ground to catch up onin the multi-chip arena due to memory bandwidth. For most people at home this might not be an issue, but for servers it can be.

Between that and the quad-core chip, it would be very interesting to see Intel manage to recapture some desktop space while AMD gets into the Server Room. :)

Better Reviews

[Google85]

There is better and more balanced reviews here and here

Re:[offtopic] 'Print' version is split into 10 pag

[myurr]

Are they really? Conroe has been shown to be quicker in limited tests overseen by Intel. Note that, as far as I am aware, none of the 'independent' testers so far have been responsible for setting up any of the test machines. And while I am sure Intel did not do anything untoward, can you honestly guarantee that absolutely everything is fair and equal in that test?

Take a look at this article: http://www.dailytech.com/article.aspx?newsid=2487

It's showing the new woodcrest chips to be somewhere between 5 - 15% faster, as opposed to Intels claimed 33%.

Finally, have you considered that AMD may just be working on something new. For a start this is their first AM2 chip, they will surely start ramping the core clock as well as the memory clock soon. And they have kept very quiet about their other future developments, not to mention any process changes (ie. catching up with Intels 65nm lead).

So while Intel may indeed recapture the top performance slot for a while it's likely to be transient until AMD release their next big thing. Which will capture the performance crown for a while until Intel release their next big thing. And so on.

Re:[offtopic] 'Print' version is split into 10 pag

[jiushao]

Of course I forgot to mention that Conroe is 32bit

Well, that's true, except it's not.

What about Pacifica?

[jimcooncat]

Is Pacifica included with this? Can we finally run unmodified Windows under Xen?

Ah...

[Christopher Rogers]

...Memories of DDR2...

Re:Intel's roadmap to the Cornroe and beyond

[westyx]

See, I like comparisons which compare products that are actually out, because comparing product a with (hypothetical) product b inevitably ends up with "product b rocks! shame it isn't out yet. can't be bought. benchmarks will be x. probably. if everything works out"

It's totally fair. You could say that the upcoming amd chips *might* be better than what intel is rolling out now, but to say it isn't 'fair' smacks of fanboiism.

Re:Socket consideration

[scharkalvin]

The existing socket 939 dual core cpu designs will probably
be available for a while yet. So you will be able to
upgrade. Consider that you can't use DDR2 memory in DDR sockets,
so you would STILL need a new MB even if AMD kept the same
socket for the new cpu.

Concerning nVidia chipsets:

[Ayanami Rei]

The nForce 2200/2050 MCP is a much better part than the nForce 4/SLI. Its my favorite system baseline for new servers. I love being able to have dedicated PCIe 4x lanes on each HT link and such. In that configuration there aren't many drivers to speak of that you need. forcedeth works fine for networking, it even supports much of the TCP offload capabilities...

An even better mix is the AMD 8131 + nForce 2050. That gives you PCI-X AND PCIe, dedicated. It's my favorite server platform.

The Once and Never Champion

[Catbeller]

AMD has beaten Intel's offerings for, what, three years now, and STILL they can't get a break. Apple won't use them. Dell just this month tentatively offers AMD chips for their server line.

The chips are cheaper and are faster than Intel's. You couldn't tell from the press!

No matter what AMD does, the next line in this type of story is "But Intel's next CPU, expected in the year 2121, is expected to outperform AMD's best. Is AMD doomed? "

Re:wishful thinking

[MBGMorden]

"Whitebox" not on par quality-wise versus OEM? I know you must be joking (or trolling). OEMs are popular for one main reason at large corporations: when something breaks you get the convenience of a single point of contact for warranty service. Other than that, they are generally built with whatever lowest cost crap they manufacturer could get their hands on (ESPECIALLY Dell, which in recent years has just gone to crap. It's a pathetic thing when I can pickup a desktop system unit and by lightly twisting on each ends with my hands see the system FLEX back and forth).

You also have to deal with their horrendous OS loads which nearly cripple the system and the fact that 99% of the time they don't include an actual OS install CD to start from scratch - just a "restore disk" that will start off with a fresh system just as messed up as when you started. Heck even when they do include an install disk it's often missing stuff from the full version.

Yes, a whitebox system can be crappy too, because you have the option of buying crap components. However, a custom/whitebox system built with good memory, a reputable mainboard/graphics card/etc, will beat an OEM machine in quality and stability every time. The only caveat is that you have to be preparted to do some legwork regarding the warranties if something breaks (which in my experience, isn't often. only things I've ever had actually break were hard drives, and most of those were from that horrdendous IBM 60GXP line).

WRONG LINK

[default luser]

I could have sworn I linked correctly.

Here is the correct link:

Anandtech's Real-Time Pricing Engine

Re:[offtopic] 'Print' version is split into 10 pag

[ak3ldama]

Wow. I didn't realize what is going on until now. This is just the resurgence of the Intel fanboy. Which is coinciding and colliding furiously with Apple fanboys. Intel has sucked until Pentium M, and now doesn't suck so much with the Intel Core Duo. The price/performance/wattage of machines is becoming equal whether AMD or Intel.

The funny part is that it just hasn't happened yet, but the Intel fans have been waiting for so long that they've just gotta get up and sing anyways. AMD has only a 90nm process, and is still competing with the futuristic Conroe. AMD is still the best option on the desktop. AMD is still the only real option in server land. Want 4 sockets (or even just 2)? Want a reasonable cooling/power bill? AMD is the only option. Intel lucked out by keeping the Pentium 3 architecture around thus managing to keep the laptop/mobile market. Now they're putting those benefits into their desktop/server platforms and by doing being competitive again.

But Intel has not won the battle yet, is the Conroe out yet? Where is the massive proliferation of desktop motherboards for building Core Duo machines? They're not here yet.

Once DDR2 moves beyond 400mhz, AMD will have some serious bandwidth. Later this year AMD will have 65nm, this will bring huge results for their power/watt numbers. So both sides of the story can spout off about what they will have 'eventually.' People need to sit back and re-evaluate things. All we have right now is a healthy, competitive market.