Xeon64 or Opteron

[ALIEN3001]

LOL, no compatibility, no benchmarks

?Software IOTLB ? Intel® EM64T does not support an IOMMU in hardware while AMD64 processors do. This means that physical addresses above 4GB (32 bits) cannot reliably be the source or destination of DMA operations. Therefore, the Red Hat Enterprise Linux 3 Update 2 kernel "bounces" all DMA operations to or from physical addresses above 4GB to buffers that the kernel pre-allocated below 4GB at boot time. This is likely to result in lower performance for IO-intensive workloads for Intel® EM64T as compared to AMD64 processors.?

Important: Although Windows Server 2003 for 64-Bit Extended Systems supports both AMD Opteron processors and Intel Xeon with Intel Extended Memory 64 Technology (EM64T) processors, this evaluation version predates Xeon processors and supports only AMD Opteron processors. Future beta versions will be released that support both AMD Opteron and Intel Xeon with Intel Extended Memory 64 Technology (EM64T).
In my words -> Intel didn't make a complete copy of AMD64, that's why Windows XP for 64-bit extended systems DOES NOT support Nocona YET.

 

[ALIEN3001]

>>32bit apps I think will stay in comparison to the P4 Prescott and Opteron comparisons but on the 64 bit level I think Intel may have something good.

So, Nemesis, your answer?!

 

[zerodeefex]

blah blah blah, bs bs bs, look at me.

AMD has been around long enough and isn't going to die out anytime soon. Sanders drove the company into the ground with his delusions of grandeur after he went off the deep end. From building "the white house" to 100 million dollar parties with people like faith hill performing during losing years, AMD still sits bloated with the remnants of his reign, around 75 VPs, more than 80% of whom don't really do anything. They have quarterly layoffs, but oddly, management is almost never touched... Great idea, let's leave a bunch of sychophants to manage with no employees...

And still they keep going. The 286 may have been a flop and the 386 didn't make a lot more headway, but they had a hit with their 486 chips. The k5s and k6s brought a lot of criticism but there were enough budget rigs running k6-2s at one point to keep them afloat. They have had it rough during the time of this athlon series but the resurgence in the enthusiast market as well as the improvement in their brand image that the Opteron/A64 has brought along has proven that despite them being such a small company, they will not die out despite whatever crap people speculate. Ruiz has effected a lot of change like he promised and the company is sure to have quite a few surprises in store.

But on the flip side, Intel was in the right place at the right time and ended up the one on top. They have a huge amount of resources and such a huge lead in the market that it is not as simple as AMD having good technologies. Obviously they recognize this and seem to be focusing on bringing about change in the way people percieve their products. With so many people realizing the potential of the A64 and opterons, things do seem to be looking good for us as both of them compete to make better products which is definitely in our best interest!

 

[IntelUser2000]

WHO CARES? Are any of you guys gonna get Nocona or Opteron? I doubt most of you will anyway. If you wanna argue which processor is better at what and how next processor from this company will beat the next processor from that company, you can argue forever, never ending. Okay I'll add another one. Sure K9 is dual-core and better Hypertransport and everything and might own the current Itaniums. But the 90nm Montecito core Itanium will have 667MHz bus 256-bit and 4x the performance. Now we see Montecito owning, and by a lot!!! If you go to www.x86-secret.com, you'll see a roadmap about Intel next gen processor, code named Cedarmill, which will go to dual core later, but initially single core. It seems the first dual-core processor might as well be Itanium Montecito, since Itanium Montecito is supposed to come mid-2005, while desktop/notebook dual-cores will come Q4 2005. We'll see whether the performance both Intel and AMD claim will come true anyway!!! By the way, according to rumors, Nocona has 16KB Trace Cache compared to 12KB in Prescott!!! That will make a big difference if its true, since that cache decreases penalties from 31 stage pipeline. And according to Prescott's workstation benchies, Prescott is well suited to workstation, even with the current 12KB trace cache!!! I mean its faster than Pentium 4 EE!!! Which is Xeon MP basically with 800MHz bus and dual-ddr 400!!!
Link: http://www.anandtech.com/cpu/showdoc.html?i=1965&p=5

 

[Mloot]

Originally posted by: IntelUser2000
WHO CARES? Are any of you guys gonna get Nocona or Opteron? I doubt most of you will anyway.

Actually, I already have (see rig in sig, if interested). I am very interested in how the Nocona's performance compares to Opteron, and if (personally, I don't think it will) Nocona puts Opteron to shame, then I will probably switch platforms. So, for me at least, these discussions are very relevant.

 

[bobbyk]

Um, not qute. The first dual cores are already out. Unless my memory just failed, IBM has the power4+ and sun has the UltraSparc IV. Those are both dual core

 

[IntelUser2000]

Want Intel innovation? How about Trace Cache, which will effect Nocona a lot if Intel gets extra 4KB Trace Cache on the Nocona(16KB vs 12KB) compared to Prescott?. That's a new technology that's been never used as far as I know(not new in terms of how long it has been introduced but). How about raised transistors(also called 3D transistors or tri-gate), that's Intel technology that will come with 45nm generation. How about Matrix Storage RAID which allows RAID 0+1 with 2 drives on the Alderwood/Grantsdale chipsets? Another first. How about the wireless implementation that will be enabled with the PCI card Intel wireless 2225BG(Go see i925/i915 at Intel site which has that link at the bottom right side) which allows your PC to act as access point or router in software mode?

AMD didn't patent hyperthreading(or SMT), they took tidbits from Intel's and some others, the article said they were considering to put it(but since you would need close to 20 stage pipeline to take advantage of SMT, I don't think AMD wold go soon, or with much performance increase as Netburst architecture). Some of you may remember that article, it was from www.theinquirer.com I think.

 

[bobbyk]

Yes, it is true those are innovations and good ones at that. The only note is compare revenue. Intel 20+ Billion. Amd ~4 or 5 Billion. With 4x the revenue shouldn't intel be able to do a lot more than that?

 

[ALIEN3001]

What's worse - no iAMD64 or lousy iAMD64 implementation!?
http://www.aceshardware.com/forum?read=115087541

 

[Vee]

Originally posted by: IntelUser2000
WHO CARES? Are any of you guys gonna get Nocona or Opteron? I doubt most of you will anyway. If you wanna argue which processor is better at what and how next processor from this company will beat the next processor from that company, you can argue forever, never ending.

Exactly. I think even Intel themselves have said not to expect performance from 64-bit operation in the Prescott64/Nocona. It's intended as a 32-bit Xeon server, with "64-bit capability" tacked on as the means for Dell to comply with specs presented by corporative buyers.
Otherwise this is mainly the usual brawl between Intel and AMD fans. That is all good fun of course. But the Prescott is not Intel's response to AMD64 in anything but a very temporary context.

Okay I'll add another one. Sure K9 is dual-core and better Hypertransport and everything and might own the current Itaniums. But the 90nm Montecito core Itanium will have 667MHz bus 256-bit and 4x the performance. Now we see Montecito owning, and by a lot!!!

I disagree. With a projected SPECint_base_rate2K of 65 from 1.7 billion transistors against 38 from 205 million transistors by the dual-core Opteron, it looks like a megalooser. Itaniums track record of projected vs. actual performance has been utterly abyssmal so far, BTW. I don't think Montecito is any threat at all against AMD. IA64 is dead. Performance advantage is marginal or nonexistent. Chip area and hardware costs are many times greater, software costs are many, many times greater, and there is also a huge loss in time, software flexibility and software delays. Only the real prize corporative idiots will be buying Itanium! They are around, of course... Intel insists on IA64. And they have made progress in compiler technology, reducing costs, converging Itanium and Xeon hardware structure. But I fail to see Itanium being relevant beside IBM Power and AMD64.

If you go to www.x86-secret.com, you'll see a roadmap about Intel next gen processor, code named Cedarmill, which will go to dual core later, but initially single core.

x86-secret.com might be poorly informed on this one. Most of the technologies leading towards Cedarmill have beeen cancelled. I think Cedarmill was to be a dual core Tejas, migrated to .065 micron. But it was looking like a 150W-160W CPU anyway. Don't worry, I think Intel have a better option up their sleave.

It seems the first dual-core processor might as well be Itanium Montecito, since Itanium Montecito is supposed to come mid-2005, while desktop/notebook dual-cores will come Q4 2005. We'll see whether the performance both Intel and AMD claim will come true anyway!!! By the way, according to rumors, Nocona has 16KB Trace Cache compared to 12KB in Prescott!!! That will make a big difference if its true, since that cache decreases penalties from 31 stage pipeline. And according to Prescott's workstation benchies, Prescott is well suited to workstation, even with the current 12KB trace cache!!! I mean its faster than Pentium 4 EE!!! Which is Xeon MP basically with 800MHz bus and dual-ddr 400!!!
Link: http://www.anandtech.com/cpu/showdoc.html?i=1965&p=5

The P4EE is not fast, so that was hardly worth the 3 excamation marks I expect Prescott Xeons to scale as poorly as Northwood Xeons. In real terms, there is no competition to Opteron. But there doesn't have to be. A good part of the market is prepared to buy Intel no matter what. As long as they can provide parts complying to buyer specs, id est compatible with Windows64. That's Prescott64's sole mission. Don't expect performance.

 

[bobbyk]

The problem is not intel's innovation. The problem is their mfg. process. Strained Silicon creates heat. The layers are thinner so the electrons switch paths easier which creates heat. Period. Intel is having problems because they invested billions in the technology that works wonders. (strained silicon) It just has heat issues. Intel couldn't do 3.6 GHz without it most likely. they need to start over, another several billion dollars.

 

[Nemesis2038]

I expected something more from Nonoca. Like it to work. This is still the sorriest product launch ever. Apple can Mock Intel for this. Even Cyrix/VIA can possibly mock this launch. This was very dissapointing.

I do own an Athlon 64 3500 overclocked and Nonoca was a surprise to me it was released this soon. I honestly didnt expect the chip so soon. It made my recent purchase worrysome I may have jumped the gun. So I expected Intel rushing to get their chip out because it was so good and they wanted to capture the lead again. Totally not true. I am relieved to find my Athlon 64 is still top of the line computing.

I think this is a serious mistake on Intel's part to release a CPU that cant complete a Windows 64 install to complete benchmarks. But I suspect there is an NDA which denies showing the Intel chip in any bad light. Thus there is nothing to show about the chip because performance is that lack luster to non working.

So Has anyone done even 32 bit benchmarks on this chip or is that broken too?

 

[sandorski]

Can Intel survive a sucky Nocona? Sure they can. However, if Nocona turns out to be crap(still can't conclude this) with only a Marketable, yet practically useless, 64bit feature, I think in the longterm Intel's credibilty is going to take a hit.

Itanium was not their first attempt to play with the big boys(IBM, Sun, and other Large highend server manufacturers). Their first bid failed in the late 80's/early 90's and it looks like Itanium may never catch on. The Xeon line is what gives Intel the credibility to even attempt to release an Itanium class processor, if the Xeon line becomes undermined with a lacklustre attempt to compete with the Opteron and fails, not only will Intel lose marketshare in the coveted Server market, but their reputation is going to take a hit. So if the Itanium needs to be cancelled, will the Xeon be there to soften the blow?

 

[Nemesis2038]

I think Nocona in the end will help AMD get recognition.

Both AMD and Intel benefit from AMD gaining some market share.
AMD needs the dollars to survive.
Intel needs AMD so they arent a monopoly.

Although I cant recommend anyone buying a Nocona because of what I hear and see on Linux Performance and Windows Incompatibility. Granted windows will get fixed but who will buy a CPU you cant get an OS installed properly on? This is not good marketing because it will cause people to hate Intel chips. I dont think this is what Intel is intending to do. What intel would be telling people is that Intel makes bad chips. But most people will proably blame Microsoft OS instead of Intel. If anything Microsoft may and should consider not supporting Intel i86x64 unless its fully compatible with AMD64. Otehrwise Microsoft might lise thier reputation.

I would suspect Nocona is merely Prescott with 64 bit emulation turned on.

I also suspect its a chip that will get scrapped with a much improved x86 64 bit core leaving Nocona a risky purchase because it may be a cpu with a very short life span.

 

[ALIEN3001]

Intel needs AMD so they arent a monopoly.

Yeah, and Microsoft needs Linux so they aren't charged with monopoly.
If Intel could have a better CPU, then they would put it out to masses, it's not like they are making crappy CPUs just to not be in a monopoly position

 

[bobbyk]

Good point. I don't think intel made nocona suck on purpose, until the itanic iss till on the table. hm... maybe.

Just one question, why all of a sudden is the nocona a bad thing. Intel has almost always pulled a marketing blitz and lowered prices way down so that amd can't compete.

I think amd knew all along. Intel really didn't want to do 64 bit. I don't think that has changed. (i mean by xeon standards)

 

[cybergen7]

Talking about opteron would you say that the opteron 146 is a good cpu when compared to the FX or3400+ I am really distraught on what to get. What do you think?

 

[bobbyk]

what does that have to do with this thread. This is about opteron and nocona.

 

[ALIEN3001]

Hehe, Intel should improve AMD64

Offtopic: If you take Opteron 2.2GHz out of your system (let's say you have it) and put in a Athlon64 FX-51 (2.2GHz, socket940), they would perform THE SAME, just Opteron hasn't got un unlocked multiplier, but the FX does. If you put in an Opteron x46 (2.0GHz) it would perform the same as a Athlon64 FX-49 (if it would exist).

 

[IntelUser2000]

"Itaniums track record of projected vs. actual performance has been utterly abyssmal so far, BTW"

Really? Take a look at this.

Itanium Merced 800MHz ~400 specint ~701 specfp

Next gen Itanium was said to be 2x faster

Itanium Mckinley 1GHz ~800 specint ~1356 specfp

The 0.13 micron Itanium was supposed to be 30-50% faster

Itanium Madison 1.5GHz ~1300 specint ~2161

So you saying track record of projected vs. actual performance sucks? Actually no, it has been better than projected.

Look at spec_rate for 2 processors
Itanium Merced 800MHz results not shown at www.spec.org
Itanium Mckinley 1GHz specint rate ~18.7 specfp rate ~27.6
Itanium Madison 1.5GHz specint rate ~30.5 specfp rate ~42.4

Considering that specint rate is same for Madison as Opteron 250, nearly doubling the score with 62 vs 38 is quite an improvement.

 

[Showtime]

Intels last 2 releases were crap IMO. The 1st batch of p4's ran slower than the existing PIII's and now the E chips are, clock for clock, slower than the C/EE P4's. I know that it all changes at around 3.6ghz for the new E chips, but why release a new slower chip? Why not wait and release a 3.6E that is faster than comparable C chips?

My last 2 or 3 rigs have been intels and they were great. I currently run AMD cuz of the price to performance ratio. Eventually I may build a P4e system, but I don't trust anything intel intially brings to the market. The new Intel cpu/mobo/ddr2/pci xpress rigs are disappointing so far. I still feel sorry for the people who bought Celerons over other products because they were to that Ghz was the end all of computing.

-show

 

[ALIEN3001]

IntelUser - just consider the price and how ofter are new processors released. The price of 1.5GHz 9MB Intaium2 is WOOOOOOW, for a lower SPECint score and higher SPECfp score, but when more CPUs are added (2 or 4 CPU systems) Opteron gains even more becouse of better scaling. Just consider the price of hardware and software -> new OS, new apps (really expensive, have to be ported to IA-64), new PSU, new case, new motherboard (really expensive motherboards), new CPUs,... Just one Itanium2 vs Opteron is expensive, when you take all of this into account - too much $$$ for the same performance, but less availlable applications.

 

[Vee]

Originally posted by: IntelUser2000
So you saying track record of projected vs. actual performance sucks? Actually no, it has been better than projected.

Look at spec_rate for 2 processors
Itanium Merced 800MHz results not shown at www.spec.org
Itanium Mckinley 1GHz specint rate ~18.7 specfp rate ~27.6
Itanium Madison 1.5GHz specint rate ~30.5 specfp rate ~42.4

Considering that specint rate is same for Madison as Opteron 250, nearly doubling the score with 62 vs 38 is quite an improvement.

Yes, it would be, if it were actual. but it isn't.

And actual specint peak rate for IBM 2XOpteron250 is 32.9, and 30.3 for HP 2XItanium2 1.5GHz.
Your argument all depends on what you think is *projected* and *actual*. Basically, let's get back to this when actual dualcore Opterons and Montecito systems starts posting specints.

But I have some additional comments for now. The first is that Itanium performance is extremely dependent upon compiler optimization, and that SPEC2000 is something Intel compiler engineers have come to understand very well. That means it may not be exactly "actual" performance. Look at Itanium scores for 'base' and 'peak'. Now why is that?
Another comment is that, originally DEC Alpha servers, and now eServers and clusters from IBM, with Opterons, cpu for cpu, equal or even outperform Itanium/Itanium2 servers and Itanium2 clusters. Is that consistent with Intel's "projected" performance from Itanium/Itanium2?

My final comment has nothing directly to do with this Montecito's projected performance: Is it even reasonable to compare Itanium2 systems to Opteron systems on a CPU to CPU basis? What about costs?

 

[bobbyk]

SO we are now what, on the fourth day. There is barely any mention of the xeon em64t. Did they all disappear or something?

Intel itsself even said though, that at first they will be availabl ein "limited" quantities. Dell seems to have them though .....