part 4 opteron review

[HurleyBird]

Look at that! The opteron is an unstopable gamming machine! Get some 400 or 533MHz DDR in their (from the aces review we know that 400MHz can probably be done) and you've got stellar proformance! as well

From NVIDIA website:

Workstations offering NVIDIA nForce3 Professional platform processors will be available from:

@Xi Computer Corporation
Alienware® Corporation
Angstrom Microsystems
BOXX Technologies Inc.
Colfax International
Concordia Graphics
Core Microsystems
DigitalScape Corporation
Einux
Hypersonic PC
Max Black
MaxVision Corporation
NTSI
Polywell Computers
Si Computer Italy
Sys Technology, Inc.

It looks like alienware will be using Nforce 3 Pro... this can only mean one thing! opteron systems geared for gaming, albeit at a HUGE price premium.

The thought of owning one of those monsters makes me want to drool...

 

[Adul]

i want one, mainly because it can double as a gaming server, dev sever, and over all bad ass machine

 

[HurleyBird]

I wonder how well it would overclock...

With the no FSB pretty much any memeroy speed should work with it... even if its not "supported"

With A64 And opteron i think we'll see a lot of new memory standards appearing

 

[Viper5]

WE all must remember that unless they change the core for Athlon 64, its going to be slower Mhz to Mhz to that of the Opteron. Say goodbye to 1Mhz L2 cache and Dual DDR memory, and say hello to 256K L2 Cache and single DDR memory!

That said, im not sure how competitive the Athlon 64 will be against the new P4 prescott unless they released it at some considerable Mhz. I'm also quite concerned at the DVD encoding and 3D rendering performance. It's not too flash at all on the Opteron, so i cant see it being any better on the Athlon 64!

 

[mechBgon]

Originally posted by: Viper5
WE all must remember that unless they change the core for Athlon 64, its going to be slower Mhz to Mhz to that of the Opteron. Say goodbye to 1Mhz L2 cache and Dual DDR memory, and say hello to 256K L2 Cache and single DDR memory!

That said, im not sure how competitive the Athlon 64 will be against the new P4 prescott unless they released it at some considerable Mhz. I'm also quite concerned at the DVD encoding and 3D rendering performance. It's not too flash at all on the Opteron, so i cant see it being any better on the Athlon 64!

Recall that the Athlon 64 will be available in 1Mb versions too. Presumably there will be higher clock speeds for the A64, but I still envision some enthusiasts going for an Opteron/nF3Pro or dual-Opteron setup. I sure wish nF3 Pro had a 64-bit PCI bus, that would be sweet.

 

[apoppin]

Taking into account that most of our tests weren't overly sensitive to cache size increases and aren't incredibly memory bandwidth intensive, these performance figures suggest some very strong potential for the Athlon 64. If AMD can launch the Athlon 64, even at 2GHz, it could be competitive with a Pentium 4 3.0C across the board; we'll have to wait and see what will be necessary for AMD to remain competitive once Prescott hits later this year.

Lots to look forward to . . .

 

[Whizzy]

i'm personally not to sure about Athlon 64 gaming performance.. looking at part 4 of the coverage and taking into account that most of the Athlon 64 chips are deliverd with a 64 bit bus to the memory and a smaller cache ( I wonder how many people will actually buy one with 1M cache instead of less (due to costs) )
I think at 2Ghz it can just out perform a 3000+ Barton... imho the Athlon 64 should be released at least at > 2Ghz (this certainly counts for the smaller cache equipped models) in order not to fall back on the performance indicator so that they start at 3000+

Hopefully a AGP nforce3 board will increase performance even more, as some people have said before: if not then the Prescott will be the gaming cpu to get..

These are just all guesses so you may or may not take it seriously

 

[draggoon01]

not only will athlon64 have less cache compared to those results, but the top consumer level pentium will have more cache and updated hyper threading. clock speed differences will probably be the same.

 

[JeremiahTheGreat]

i'm quite sure that the A64 will come in 2 versions. The highend will have 1mb L2 cache, while the "Duron 64" will have 256kb L2.

Hell, if P4 proves to be too much for it, i'm quite certain AMD will put in that second mem controller - they still have 6months of so to do it.

 

[draggoon01]

Originally posted by: JeremiahTheGreat
i'm quite sure that the A64 will come in 2 versions. The highend will have 1mb L2 cache, while the "Duron 64" will have 256kb L2.

Hell, if P4 proves to be too much for it, i'm quite certain AMD will put in that second mem controller - they still have 6months of so to do it.

has amd actually given indications that they will do something along those lines?

 

[mechBgon]

Yes, that's the official plan (both 256kb and 1Mb Athlon 64's).

 

[KenAF]

Unfortunately, as other web sites are beginning to note, it looks like Anandtech may have made a serious error in its Opteron desktop performance evaluation.

As Johan from Aceshardware pointed out, it looks like Anandtech used the PCI-X slot for their testing. The PCI-X slot on Opterons enjoys a huge latency advantage over the standard PCI implementations in the nForce and Canterwood. If the nVidia GF4 Mx420 board supported 66MHz PCI (as some older nVidia boards have done), then the Opteron would not only have had the benefit of PCI-X's much lower latency, but also twice the bandwidth.

Past benchmarks have shown that AGP GF4 Mx420 boards significantly outperformed their PCI versions on the Athlon and P4 platforms. And several of Anand's tests clearly show that the reported frame rates are not cpu limited. For the Jedi Knight II results right here, the Athlon XP 3000+, 2200+ (Barton), and 2200+ (Tbred) all scored exactly 39.7 fps. In contrast, the P4 3.0 scored 65.8 fps, while the Opteron scored 74.2. These results are consistent with the differences in PCI implementations that some members have reported for the nForce and Canterwood.

In effect, Anand is not testing the gaming performance of the Opteron. He is testing the performance of the PCI implementations on these platforms. I noted the above in the official Anandtech part 4 discussion thread, which was subsequently deleted.

 

[KenAF]

As we see right here in another Anandtech review, the P4 2.4GHz results with the Geforce4 Mx440 AGP trash those of the Opteron with the Geforce4 Mx440 PCI.

In Jedi Knight2 at 640x480, the Opteron scores 74.2 fps and the P4 3.00 scores 39.7 fps with the Geforce4 Mx440 PCI. In the same Jedi Knight2 test at 1024x768, the P4 2.4 GHz scores 124.9 fps with the Geforce4 Mx440 AGP.

Here again we see, the Anandtech is simply testing the PCI implementations on these platforms, rather than of the processor themselves. How else would you explain a P4 2.4GHz with GF4 Mx440 AGP scoring 124.9 fps at 1024x768, but the P4 3.00GHz scoring only 39.7 fps at 640x480 with the PCI card. The Opteron simply has an unfair advantage when the PCI-X slot is used; the PCI implementations in Canterwood and nForce2 were never meant to be used with 3D graphics.

 

[kuk]

How else would you explain a P4 2.4GHz with GF4 Mx440 AGP scoring 124.9 fps at 1024x768, but the P4 3.00GHz scoring only 39.7 fps at 640x480 with the PCI card.

That's pretty understandable ... the results are completely limited to the video card (AGP removes the bottleneck present in PCI). I see how that this may pose as a problem in the Opteron review (reviewing a CPU were the video card is your bottleneck, even in 640x480). The benchmarks you have posted, however, do not demonstrate that they are testing the PCI implementations (not this is may not be true ... just that I don't see it as proof).

Anyone have a link comproving that the Geforce4 Mx440 PCI can take advantage of a 66mhz bus?

 

[KenAF]

kuk,

As I noted above, the PCI-X slot of the Opteron enjoys a significant advantage in latency over the PCI in Canterwood and nForce. So even if the nVidia board does not support 66MHz PCI, the Opteron still has an unfair advantage over the P4 and Athlon XP in the tests.

The benchmarks you have posted, however, do not demonstrate that they are testing the PCI implementations (not this is may not be true ... just that I don't see it as proof).

How can you not see that? If a P4 3.0 can score 200+ fps with a Mx440 AGP at 640x480, why shouldn't it be able to to match the Opteron's 74.9 fps with the PCI card (instead of scoring what it did), if the PCI implementations were comparable?

 

[human2k]

So how exactly is AMD going to convince your average consumer to purchase a 1.8GHZ cpu over a 3ghz cpu which prolly will cost the same?

 

[LeeTJ]

Originally posted by: human2k
So how exactly is AMD going to convince your average consumer to purchase a 1.8GHZ cpu over a 3ghz cpu which prolly will cost the same?

hmmm, let's do some thinking here.

Intels Itanium2 is running what 1.0 ghz?? wonder if anyone is seriously considering the P4 3.06 over the Itanium.

It's called PRODUCT LINE DIFFERENTIATION.

It obviously requires some education, but it won't take much.

 

[Nemesis77]

Originally posted by: Viper5
WE all must remember that unless they change the core for Athlon 64, its going to be slower Mhz to Mhz to that of the Opteron. Say goodbye to 1Mhz L2 cache and Dual DDR memory, and say hello to 256K L2 Cache and single DDR memory!

Release of Athlon 64 is still months away. During that time, they can revise the mem-controller (driving down the latencies). Also, we must remember that Athlon 64 will have higher MHz than Opteron does. And the MoBo's/Chipsets will be more geared towards desktop-systems. As to the cache.... Athlon 64 will come in 256KB and 1meg configurations. And we must not forget that Athlon64 can use DDR400, whereas Opteron has only been tested with DDR333

Basically, there is quite alot they can do to boost Hammers performance on the desktop. What we have now seen benchmarked is the server-version.

 

[Nemesis77]

Originally posted by: human2k
So how exactly is AMD going to convince your average consumer to purchase a 1.8GHZ cpu over a 3ghz cpu which prolly will cost the same?

Because MHz is meaningless? You can use it to compare otherwise identical CPU's, but you can NOT use it to compare CPU's from different CPU-families.

 

[Nemesis77]

Originally posted by: Whizzy
I think at 2Ghz it can just out perform a 3000+ Barton

Huh???? Barton 3000+ runs at 2167MHz. So you are saying that Athlon 64, with SSE2, super-fast FSB, integrated low-latency mem-controller, improved branch-prediction, more L2-bandwidth, more TLB's and other improvements, would barely beat Barton running a mere 167Mhz faster? Add to that list of improvements the potentially bigger L2-cache and x86-64...

 

[Duvie]

Originally posted by: Nemesis77

Originally posted by: Whizzy
I think at 2Ghz it can just out perform a 3000+ Barton

Huh???? Barton 3000+ runs at 2167MHz. So you are saying that Athlon 64, with SSE2, super-fast FSB, integrated low-latency mem-controller, improved branch-prediction, more L2-bandwidth, more TLB's and other improvements, would barely beat Barton running a mere 167Mhz faster? Add to that list of improvements the potentially bigger L2-cache and x86-64...

Why not!! If he is interpreting numbers from the PREVIEW done at xbits then he may not be too far off...there the 1.8ghz athlon64 barely beat the 2800+ barton (which we know took a drop in mhz from the Xp line which it didn't beat in testing recently) and in many apps lost by a large margin!!! Granted you are right and there will be some improvements, but 2ghz+ is also IMO a must...They can't just come out and be a notch better then the barton they need to make some distance from what is widely known to be their budget line when the athlon64 comes out.....

Don't so easily discount the fears and the pointing out by many that 1.8ghz at this already delayed time is a major concern. They are obviously having issues with the SOI process as is widely being reported throughout the industry. They have brought in IBM to help them out. All of this leads to concerns and the possible reality that 2ghz may be the best it hits at debut...Remember it will be released in 6months but it needs to be already be fabbed in some large numbers by then so imagine it needs to be ready in 4 months or so.




The anandtech testing thing seems interesting...I wonder if that is why I can't get the main page to load right now.....


Also looking at the opteron scores (even if they may be skewed right now) are not a fair representation of the athlon64 for the reason of the daul memory controller. Nemesis if you think they will add the cache and the dual mem controller you are wrong. There is no way they will due that as it wont continue the differentation it wants between it more expensive server line chips and the desktop chip. I think they want this separation...

 

[dullard]

They can't just come out and be a notch better then the barton they need to make some distance from what is widely known to be their budget line when the athlon64 comes out...

I've been arguing for a year or so that that senerio won't happen. The Barton is at 3000+, possibly could be 3200+ by September if AMD wants to. The TOP Athlon 64 was said repeatedly will be at 3400+ speeds. Notice there is virtually no performance difference (assuming AMDs PR ratings are done well). But that means the lower clocked Athlon 64s will be slower than the Bartons...

 

[Nemesis77]

Originally posted by: Duvie
Why not!! If he is interpreting numbers from the PREVIEW done at xbits then he may not be too far off

I rank xbit slightly above The Inquirer and The Register.

there the 1.8ghz athlon64 barely beat the 2800+ barton

Athlon 64 was a preproduction CPU running on preproduction MoBo using preproduction Via chipset. Chipset-drivers were propably still beta or pre-beta. Not to mention the BIOS was propably still in beta as well. In short: Not a valid comparison.

Don't so easily discount the fears and the pointing out by many that 1.8ghz at this already delayed time is a major concern. They are obviously having issues with the SOI process as is widely being reported throughout the industry.

They HAD problems with SOI. AMD paid IBM quite a sum to fix their problems with SOI. IBM sent engineers to their fab and fixed the problems.

They have brought in IBM to help them out.

And IBM helped and fixed the issues. So what's your point?

[All of this leads to concerns and the possible reality that 2ghz may be the best it hits at debut

Based on what? On the fact that AMD HAD problems with SOI and on the fact that they still have several months 'till the launch? There's plenty of time between fixing of the SOI-problems and launch of the chip. My guesstimate is that Athlon 64 is released at 1.8Ghz and up from there (propably 'till 2.2GHz).

Also looking at the opteron scores (even if they may be skewed right now) are not a fair representation of the athlon64 for the reason of the daul memory controller. Nemesis if you think they will add the cache and the dual mem controller you are wrong.

Where exactly did I say that they will add the second mem-channel? And it is widely reported that there will be two version of Athlon 64: one with 256KB of L2, other with 1MB of L2.

There is no way they will due that as it wont continue the differentation it wants between it more expensive server line chips and the desktop chip. I think they want this separation...

But there already is the separation they need! Server-chips have 2x64bit mem-controller and they are SMP-capable. Athlon 64 has 1x64bit mem-controller and they are not SMP-capable. Although I think that Athlon 64 will get 2x64bit mem-channel in the future. Propably in 2004.

 

[ElFenix]

ugh, the main site is getting /.'d or something

 

[Duvie]

Originally posted by: Nemesis77

Originally posted by: Duvie
Why not!! If he is interpreting numbers from the PREVIEW done at xbits then he may not be too far off

I rank xbit slightly above The Inquirer and The Register.

there the 1.8ghz athlon64 barely beat the 2800+ barton

Athlon 64 was a preproduction CPU running on preproduction MoBo using preproduction Via chipset. Chipset-drivers were propably still beta or pre-beta. Not to mention the BIOS was propably still in beta as well. In short: Not a valid comparison.

Don't so easily discount the fears and the pointing out by many that 1.8ghz at this already delayed time is a major concern. They are obviously having issues with the SOI process as is widely being reported throughout the industry.

They HAD problems with SOI. AMD paid IBM quite a sum to fix their problems with SOI. IBM sent engineers to their fab and fixed the problems.

They have brought in IBM to help them out.

And IBM helped and fixed the issues. So what's your point?

[All of this leads to concerns and the possible reality that 2ghz may be the best it hits at debut

Based on what? On the fact that AMD HAD problems with SOI and on the fact that they still have several months 'till the launch? There's plenty of time between fixing of the SOI-problems and launch of the chip. My guesstimate is that Athlon 64 is released at 1.8Ghz and up from there (propably 'till 2.2GHz).

Also looking at the opteron scores (even if they may be skewed right now) are not a fair representation of the athlon64 for the reason of the daul memory controller. Nemesis if you think they will add the cache and the dual mem controller you are wrong.

Where exactly did I say that they will add the second mem-channel? And it is widely reported that there will be two version of Athlon 64: one with 256KB of L2, other with 1MB of L2.

There is no way they will due that as it wont continue the differentation it wants between it more expensive server line chips and the desktop chip. I think they want this separation...

But there already is the separation they need! Server-chips have 2x64bit mem-controller and they are SMP-capable. Athlon 64 has 1x64bit mem-controller and they are not SMP-capable. Although I think that Athlon 64 will get 2x64bit mem-channel in the future. Propably in 2004.

Ohhhh...I am sorry!! I didn't know I was talking to an AMD or IBM rep.....What is your source that refutes what is still being said around the industry??? Do you have a link???


I am getting tired of talking to you about this as you refuse to look at the numbers at all and garner some sort of conclusion even if they are seen as a progress report. NO one is saying this is the final and the athlon64 will blow because of this....BUt you just don't get it....

Hey one last comment....

OPteron is out and on what should be a solid platform (I hope since companies will be buying them and hoping they work up to the par they should be), yet they still performed quite horribly in the desktop testing. Athlon64 aside but thisis an official released product on a board that is also released. where are the improvements??? As bad as I think the opteron did on some of those desktop test the athlon64 should be even more neutered with the initial 256kb cache and single channel memory controller....