- Oct 9, 1999
- 72,636
- 47
- 91
The fly in Opteron's ointment? Itanium to get software update for improved 32-bit performance.
- Thread starter NFS4
- Start date
I doubt that will have much effect on the Opteron's success, or lack of.
It's been said before, and now that we know the prices of the Opterons, it's been confirmed IMO, Opteron competes with Xeon DP/MP not Itanium.
It's been said before, and now that we know the prices of the Opterons, it's been confirmed IMO, Opteron competes with Xeon DP/MP not Itanium.
Originally posted by: JC
lol what a band-aid
Too bad that Intel's spin doctors will make sure that hardly anyone ever knows how embarrassing it is that the hardware implementation is so pitiful that a software emulation beats the pants off of it!
Yea the hardware sucks so bad. I'm sure you failed to see the Itanium II 1Ghz 3MB beat a Opteron 1.8Ghz 1MB in POV-Ray by more than 3X in IA-64 mode.
Originally posted by: dexvx
Originally posted by: JC
lol what a band-aid
Too bad that Intel's spin doctors will make sure that hardly anyone ever knows how embarrassing it is that the hardware implementation is so pitiful that a software emulation beats the pants off of it!
Yea the hardware sucks so bad. I'm sure you failed to see the Itanium II 1Ghz 3MB beat a Opteron 1.8Ghz 1MB in POV-Ray by more than 3X in IA-64 mode.
That particular result is kinda questinable IMO, considdering someone managed to put up a result that beat the I2.
That result was a 100 MHz something running FreeBSB.
Not saying the Itanium II isn't a good CPU, or that the POVRay bench is fake, just saying I don't trust that particular bench as much as a SPEC bench for example.
Originally posted by: Sunner
Originally posted by: dexvx
Originally posted by: JC
lol what a band-aid
Too bad that Intel's spin doctors will make sure that hardly anyone ever knows how embarrassing it is that the hardware implementation is so pitiful that a software emulation beats the pants off of it!
Yea the hardware sucks so bad. I'm sure you failed to see the Itanium II 1Ghz 3MB beat a Opteron 1.8Ghz 1MB in POV-Ray by more than 3X in IA-64 mode.
That particular result is kinda questinable IMO, considdering someone managed to put up a result that beat the I2.
That result was a 100 MHz something running FreeBSB.
Not saying the Itanium II isn't a good CPU, or that the POVRay bench is fake, just saying I don't trust that particular bench as much as a SPEC bench for example.
Actually if you do some proprietary (well whos the judge of whats proprietary and not ?) coding, the I2 is incredibly fast. Using SpecFP, the I2 has a peak that is almost double a 3.06HT P4's peak. It was actually running at 1Ghz with 3MB L2 cache.
Yeah, like I said, I don't deny that the I2 is very fast, just saying that one particular bench doesn't mean much to me.
Originally posted by: dexvx
Originally posted by: JC
lol what a band-aid
Too bad that Intel's spin doctors will make sure that hardly anyone ever knows how embarrassing it is that the hardware implementation is so pitiful that a software emulation beats the pants off of it!
Yea the hardware sucks so bad. I'm sure you failed to see the Itanium II 1Ghz 3MB beat a Opteron 1.8Ghz 1MB in POV-Ray by more than 3X in IA-64 mode.
He obviously meant the x86 hardware on the Itanium 2. It contains a x86 decoder frontend for 32-bit compatibility. But it seems that a pure software emulation will smoke the hardware implementation in performance.
Originally posted by: dexvx
Originally posted by: Sunner
Originally posted by: dexvx
Originally posted by: JC
lol what a band-aid
Too bad that Intel's spin doctors will make sure that hardly anyone ever knows how embarrassing it is that the hardware implementation is so pitiful that a software emulation beats the pants off of it!
Yea the hardware sucks so bad. I'm sure you failed to see the Itanium II 1Ghz 3MB beat a Opteron 1.8Ghz 1MB in POV-Ray by more than 3X in IA-64 mode.
That particular result is kinda questinable IMO, considdering someone managed to put up a result that beat the I2.
That result was a 100 MHz something running FreeBSB.
Not saying the Itanium II isn't a good CPU, or that the POVRay bench is fake, just saying I don't trust that particular bench as much as a SPEC bench for example.
Actually if you do some proprietary (well whos the judge of whats proprietary and not ?) coding, the I2 is incredibly fast. Using SpecFP, the I2 has a peak that is almost double a 3.06HT P4's peak. It was actually running at 1Ghz with 3MB L2 cache.
SPECfp_base2k
P4 3.06GHz: 1077
I2 1GHz: 1431
The difference is 33% in favor of the I2 and if you examine the sub-benches the P4 wins several of them (and loses the rest)
dullard
Elite Member
- May 21, 2001
- 26,066
- 4,712
- 126
I still don't really see the need for this. Not many high-end servers have users trying to run their personal software on them. As for high-end workstation uses, again I don't see the need. When I do my simulations (which are often on HP/Sun/SGI servers) I just send the data over to a desktop machine to do the data analysis/report writing there. That frees up the expensive computer for more simulations while I'm working away on the cheap computers. In fact everyone I know in several university/industry settings does that too. Thus you are continuously doing your 64- bit work on the 64-bit computer, then you send the data over to a 32-bit computer for the rest of the work. I can just see the faces of the other researchers if I told them they cannot do their research since I want to use the server to do something like type a report. Its not like a company will only have an Itanium server and no other computer...
But even if someone wanted to have just an Itanium server and nothing else, it still likely isn't a problem. The 32-bit uses are most likely pretty mundane (ie not CPU taxing) - and slow performance won't be noticed.
Referring to the article though, it is nice to see some places finally admit that Itanium can do IA-32 in hardware (even if it is crappy performance).
But even if someone wanted to have just an Itanium server and nothing else, it still likely isn't a problem. The 32-bit uses are most likely pretty mundane (ie not CPU taxing) - and slow performance won't be noticed.
Referring to the article though, it is nice to see some places finally admit that Itanium can do IA-32 in hardware (even if it is crappy performance).
It's called being smart. In the grand scheme of things, this means nothing as the 2 products are not competing for anything, but this will still give Intel a boost in the perception ratings. No one who was planning to buy an Opteron system is going to switch to Itanium 2 because of this announcement, nor is the reverse true because of the Opteron release.
Bingo! This is pure marketing, the Itanium II and the Opteron aren't in the same market anyway. They might have a small overlap in the 8-way SMP configuration, but that is really somewhere 64-bit software will be used.
will programs be emulated line by line or jit compiled?
also, what does the boot process on an itanium machine look like? does it use openfirmware or bios?
also, what does the boot process on an itanium machine look like? does it use openfirmware or bios?
Not speaking for Intel, but since this is on a project that I have worked... it's been in the works for a long time, and it's actually only recently that all of the pieces fell into place. Whether or not the actual announcement was convienently timed or not, the fact is that the timing couldn't have been much different regardless of what other companies in the industry were doing.
The performance of the Itanium 2 on legacy code is purely a function of what target consumers were anticipated to be looking for. It's not like Intel doesn't know how to design high-performance IA32 microprocessors. But on the design of the Itanium 2 attempted to maximize 64-bit performance and to do this sacrificed 32-bit performance. This was a design choice.
As far as the performance of the Itanium 2 in 64-bit applications, there are plenty of benchmarks and real world examples and, in my opinion, the benchmarks speak for themselves far more eloquently that I could. And this is on a product that was released almost a year ago.
And, as someone who uses POVRay occassionally, I know that waiting around for a scene to render just so that you can tweak the code and re-render again is frustrating. It is a particularly useful benchmark if you use POVRay for rendering - as plenty do.
First
Lifer
- Jun 3, 2002
- 10,518
- 271
- 136
And it doesn't hurt that IPF is extremely fast in POVRay, as shown over at Aceshardware.
Even if that were true, IBM is not by any means a small win for AMD. If I had my pick of a tier one vendor for Opteron I'd pick IBM in a second. Maybe HP Opteron HPCs (which AMD probably won't get) would have been almost as good as IBM HPCs.
Btw, Gateway, IBM and Fujitsu Siemens plan on building Opteron boxes.
I presume that was why Aces threw it into the benchmark mix, since that was the only Itanium 2 benchmark that they mentioned.
My point in talking about POVRay was not to talk about how great the Itanium 2 is because it renders so quickly, although that's nice, but to respond to the comment by Sunner that it's not a particularly useful benchmark. It is useful if you render frequently in POVRay - which plenty do since it's a pretty decent program, especially for the price. Although, I do agree with Sunner. I trust SPEC CPU2000 more than anything else too as an overall estimate of likely cross-platform performance.
I seem to recall a little over a year ago that a site, Aces IIRC, put up a review of the Pentium 4 and showed how badly it did at the time in POVRay compared to the Athlon. And this was used as an example to show how poorly the FPU on the Pentium 4 performed. And then a couple of people here on Anandtech recompiled POVRay using the latest version of MS's Visual C with Pentium III optimizations turned on and the score improved dramatically - like an 80% increase - such that it was actually faster at rendering in POVRay than the Athlon after the recompilation. There's a thread on it somewhere in the archives here in the forums.
I remember that too, but it was compiled with the Intel C++ compiler (v5.01) and it made the P4 and K7 perform about the same (the fastest non-overclocked at the time)
Which makes me wonder about this Itanium 2 score in POV-Ray. What compiler was used for the x86 results for the Athlon and P4? And what compiler were used with the Itanium2 ? IIRC POV-Ray used an old Watcom compiler as default for x86 which was not shown to be very fast (especially for the P4 but also the Athlon which also gained from the Intel compiler). I'm not sure if the AT Forum recompile ever made it outside of here. IMHO the IA64 version most likely used the Intel Compiler (v7.0) on a Windows platform.
So if it turns out that the comparison was between the Watcom compiler on P4 and K7 vs the Intel Compiler on the Itanium 2...
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 23K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter