Xbitlabs:Opteron 144/ N3 article.

[Remedy]

link.

 

[dullard]

I skimmed the benchmarks. Since the Athlon 64 won't be much different than this, this link will be disappointing to many AMD fans. The Opteron was far behind in almost all tests except for games. The gaming performance was quite good though. Sure a boost to 2.0 GHz will help the Athlon 64, but I don't see the initial chips having a chance against a 3.4 GHz Prescott unless something major changes.

 

[AmdInside]

AMD Athlon64 will be faster on normal everyday tasks with the same chipset because it will do away with registered/buffered memory requirements.

AMD Forever

 

[dullard]

Originally posted by: AmdInside
AMD Athlon64 will be faster on normal everyday tasks with the same chipset because it will do away with registered/buffered memory requirements.

AMD Forever

I'm not saying you are wrong, but can you provide one link of actual benchmarks that getting rid of ECC will help performance. I heard that over and over again in rumors that were years old based on years old computers - but never one link of data and of course nothing with any current technology. So I tested a memory intense program on my work computer (dual 1.7 GHz Dell Precision 530) and saw absolutely no difference in speed (testing PC800 RDRAM with PC800 ECC RDRAM). But that is just one test on one program on one computer. So is there any current data that backs up your statement?

 

[Dug]

Now if I could buy it for $50 and oc it like their 1700+ then it would be something to get excited about.

 

[Mikewarrior2]

i think hte o/c story will be very interesting... seeing as how those crazy ln2 overclockers from japan got a mere 400mhz out of an opteron 1.8ghz, and that included a crazy, crazy volt mod.


mike

 

[Electrode]

This may have been explained at one point, but I'm not entirely sure of what the differences are between the Athlon 64 and Opteron 1xx, other than the 754-pin package used in the former and the 940-pin package used in the latter.

BTW, I've been using dual Opteron 240's for about 2 weeks now, and I'm very impressed. I'm seeing 160-200 FPS in Quake 3 during fairly heavy combat at 1024x768x32 with all eyecandy maxed out, and I've gotten the kernel to compile in just over 2 minutes. That's in 32-bit mode, I haven't really done much with 64-bit mode yet since my distributed computing project of choice doesn't like it.

 

[Remedy]

Originally posted by: dullard

Originally posted by: AmdInside
AMD Athlon64 will be faster on normal everyday tasks with the same chipset because it will do away with registered/buffered memory requirements.

AMD Forever

I'm not saying you are wrong, but can you provide one link of actual benchmarks that getting rid of ECC will help performance. I heard that over and over again in rumors that were years old based on years old computers - but never one link of data and of course nothing with any current technology. So I tested a memory intense program on my work computer (dual 1.7 GHz Dell Precision 530) and saw absolutely no difference in speed (testing PC800 RDRAM with PC800 ECC RDRAM). But that is just one test on one program on one computer. So is there any current data that backs up your statement?

Your test were just about as right as the rest of us who use Registered memory. (760MPX with Gigs of Registered ECC here) I can agree with you that the whole Unbufffered/vs R/ECC speculation is way out of hand. Unbuffered memory doesn't progress linearly in desktop vs R/ECC. It only effects the benchmarks because the test are alot of more sensitive to memory signal control, CAS, and clock freq. For ie; There are some Unbuffered boards that come in with memory clocks of 133.69 vs 132.5 vs 134mhz. From brand x,y,z motherboard manufacs. In benchmarks, all of this will reflect in the B/M tables. But in the real world, it does not. People blow it out of proportion.

 

[PrinceXizor]

This whole opteron = athlon64 seems to be a bit of an assumption. They are similar, but not exact. They have different pin-outs (quite substantially), and if the Athlon64 was SO alike to the Opteron...why isn't it available at the same time?

In all honesty, while reading that review, I just didn't feel like a whole lot of care was put into this test and a LOT of assumptions were being made. Of course, that's just my very subjective opinion.

P-X

 

[dullard]

Originally posted by: PrinceXizor
This whole opteron = athlon64 seems to be a bit of an assumption. They are similar, but not exact. They have different pin-outs (quite substantially), and if the Athlon64 was SO alike to the Opteron...why isn't it available at the same time?

According to the latest rumors, the initial Athlon 64 will have the same pin-out as the Opteron. Then it will change from 940 pins to 939 pins and also come out with a 754 pin layout.

According to AMD, the Athlon 64 is basically ready but AMD is waiting for a consumer version of Windows with 64-bit capability.

You of course could show proof denying both of those statements above - but until then I'm believing them. Thus I agree that the INITIAL Athlon 64 and the 100 series Opterons are very similar.

 

[jonnyGURU]

Originally posted by: dullard

Originally posted by: AmdInside
AMD Athlon64 will be faster on normal everyday tasks with the same chipset because it will do away with registered/buffered memory requirements.

AMD Forever

I'm not saying you are wrong, but can you provide one link of actual benchmarks that getting rid of ECC will help performance. I heard that over and over again in rumors that were years old based on years old computers - but never one link of data and of course nothing with any current technology. So I tested a memory intense program on my work computer (dual 1.7 GHz Dell Precision 530) and saw absolutely no difference in speed (testing PC800 RDRAM with PC800 ECC RDRAM). But that is just one test on one program on one computer. So is there any current data that backs up your statement?

He didn't say ECC versus non-ECC. He said Registered/Buffered, as in Registered versus Unbuffered.

There shouldn't be any kind of speed difference with ECC because the Error Correcting Chip doesn't kick in unless there's a problem (which pretty much means THERE'S A PROBLEM!) What registered does is delay data by one clock cycle, so by it's very NATURE it's going to slow down a PC.

The problem I find today is that most registered RAM I find, is also ECC and most ECC, unless it's RAMBUS, is also registered. This didn't used to be such a common practice, but I guess the memory manufacturers are wanting less part numbers and figure "if the customer wants fault tolerance, he must not care about speed" and go ahead and go with the beaucoup fault tolerance solution.

The Athlon-64 won't require either ECC or registered. Also, the Athlon-64 won't require 6-layer MLB's either. This should make Athlon-64 motherboards and complete systems cheaper. Sure, the built in memory controller will make the chip cost more, but I think in the long run we'll see a good value and a pretty quick system compared to the Pentium 4 (clock for clock, of course.)

 

[alexruiz]

Let's wait and see...... My piggy bank is ready!

 

[Xenon14]

OK, However - sicne the Athlon64/Opteron have longer pipelines than the XP models (not as long as P4) it will have a much higher increase in performance with each additional mhz than the XP models. Plus it will be able to scale relatively higher than the XP's. Don't judge the Athlon64 so harshly... as the younger P4's once were.

 

[thorin]

Those tests are almost completely pointless they aren't running a 64bit OS or 64bit software.

Not to mention that they can't decide whether it's a 144 or 244 that they're testing...having called it by both model #s throughout the article (I mainly noticed this between pages 15 and 16 but I'm sure it isn't a limited occurance....)

Thorin

 

[jjyiz28]

Originally posted by: Xenon14
OK, However - sicne the Athlon64/Opteron have longer pipelines than the XP models (not as long as P4) it will have a much higher increase in performance with each additional mhz than the XP models. Plus it will be able to scale relatively higher than the XP's. Don't judge the Athlon64 so harshly... as the younger P4's once were.

actually you have that mixed up. a longer pipeline most of the time has a lower increase in performance per an additional mhz. the exception to this would be the banias processor, longer pipeline but higher ipc compared to the p3.

my opinion, but athlon64 will survive only if more and more programs are being optimized for it. a a64 windows version should help as well.