Opteron video encoding benches

[Hulk]

http://www.anandtech.com/cpu/showdoc.html?i=1818

Some of these benches are very promising. I think that this more accurate preview of what we can expect from Athlon 64.

DivX encoding 14% more efficient than Athlon XP, WM9 encoding 19% faster. And remember that this is comparing relatively unoptimized server board for Opteron to a highly optimized nforce 2 board for the Athlon XP.

If the Athlon 64 is 10% more efficient in MS Pro than the Athlon XP then it will be quite a competitor.

The XP is already about 30% more efficient than the P4 in MS Pro.

If launched at 2GHz, this chip would perform like a 3.33GHz P4, assuming no benefit for HT.

If AMD can scale this thing to 2.4GHz, Intel will have to counter with a 4GHz P4, assuming no improvements in the Prescott core.

Sure, I know there are lots of "ifs" here, but I'm just making some educated guesses. If things go as I am predicting, the P4 and Athlon will continue to be neck and neck in MS Pro.

-Mark

[fberger]

A very interesting article indeed.

Encoding is certainly a good way to compare platforms and CPUs, but personnally I don't mind letting my system encode a 4Gb file overnight. What matters to me is responsiveness while using the video editor interactively, such as MSPro. Is RT performance directly linked to encoding performance? If yes, then Athlon 64 will be competitive against a P4 3.0c when it arrives.

P4 platforms based on Canterwood or Springdale should be available next month, whereas we'll have to wait until September for the Athlon64 and nForce3. Intel is still ahead for now!

Fred

[Brian Ellis]

The big IF is whether MSP will be coded for 64 bit processing.

Remember when we asked about Linux? They replied that if it became mainstream, they would, even though it would already give a big performance boost compared with Windows. By extension, you are assuming Athlon64 will become mainstream. This is not certain. Despite Iridium, Intel are not planning any major 64 bit processing, except for servers (which is also what Opteron aims for). Win64 is not ready either and its management of 64 bits is far from ideal, still with many 32 bit features.

AMD are putting themselves on a limb by planning to go 64 bit. Remembering that corporate computing is the biggest market and generally would not need 64 bit for anything other than serving, and that intel has the lion's share of the corporate market, AMD may relegate themselves back into a niche position - or may go bankrupt.

Have a look at the recent AMD and Intel financial results and start hedging your bets. If 64 bits is slow to take off, then AMD will be in dire trouble, so the success or not of the company is entirely up to the truths, half-truths and downright lies that the AMD and Intel marketing guys inflict upon us.

And the winner is ...

[Hulk]

fberger,

There is some correlation between encoding and MS Pro overall performance. The important benches above are Athlon XP to Operton, and Opteron bests XP in almost all of these, per clock cycle. This is a good indication of predicted Athlon 64 performance in MS Pro and other video editors.


Brian,

I'm not sure if AMD will live or die on a major platform switch to 64bit computing, in the short term anyway. I am convinced that the success of Athlon 64 is going to be based on how it performs on current 32bit OS's and software. If A64 can remain competitive with Prescott for the next year, then the 64 bit question will have to be answered because I think it will be quite a while before we see a significantly revised core from AMD. Intel on the other hand, will probably continue to scale the P4. This is when AMD will have to answer Intels hardware with software, i.e. a move to 64 bits.

Many things can happen here:

A64 can succeed in 32bit performance AND significant numbers of 64 bit apps can appear in the next year. This would be best case scenerio.

A64 can be crushed by P4 in 32 bit performance AND no 64 bit apps will appear in the next year. This is the worst case for AMD.

Reality will most likely be somewhere in the middle of these two predictions.

From what I have been reading regarding Opteron, I think AMD has a winner in the server market if, and this is a big if, they can convince die-hard Intel servers buyers to make the switch. Loyalty can be a tough nut to crack, but then again, money is a good nut cracker and the AMD solutions do appear less expensive than the Intel solutions for equal performance. But then there's that server validation thing...

Just too many unknowns here!

-Mark

[Brian Ellis]

Mark

If an Opteron cost $200 less than an Iridium, will this be really significant in the largest sector of the server market, with an average price of ~$8,500?

If Intel play their cards right, they will undercut the Opteron when Iridium 3 comes out, if only to knock the AMD feet out from under them. If Intel is cheaper, then the corporate market is assured.

Where the Opteron may otherwise win is in the SME server sector ($2,000 - 5,000), which is actually big in numbers but less so in prestige.

[Hulk]

Brian,

Good points, but as you know, there is more to the server market than just the cpu. Motherboards and other components are also a big factor. And, with the integrated memory controller in the Opteron, motherboard pricing for the AMD system will have a decided advantage. I'm suspecting that AMD will respond aggressively through pricing if things don't go as they plan.

Time will tell. I guess I'm kind of rooting for AMD because I would hate to see them go under. We'll be back to the days of $1000 cpu launches, prices that don't come down that much, and insignificant clock speed increases once a year. aka Pentium 60>66>90>100>120>133>150>166>200>233 ahhhh! Please not that again.

-Mark

[Ray Austin]

Originally posted by Brian Ellis
Mark

If an Opteron cost $200 less than an Iridium, will this be really significant in the largest sector of the server market, with an average price of ~$8,500?

If Intel play their cards right, they will undercut the Opteron when Iridium 3 comes out, if only to knock the AMD feet out from under them. If Intel is cheaper, then the corporate market is assured.

Where the Opteron may otherwise win is in the SME server sector ($2,000 - 5,000), which is actually big in numbers but less so in prestige.

Brian, I think you mean Itanium , Iridium is the satellite phone that Motorola shelled out 5Billion dollars to create only to sell it later for 25million or so.

Also, Amd did the 64bit right by allowing 32bit functionality as well... we will move to 64 bit eventually time is tale telling force.

Bigger changes are looming tho when Intel is proposed to redesign the PC doing away with AGP, & PCI with the new bus termed presently PCI Xpress.

So there are many changes on the horizon, Amd has definately got their work cut out for themselves, I would say profitability is their biggest problem...but if demand for 64bit is their the problem may shift to Intel...

It will be interesting to see... but with IBM signing on as an OEM for Opteron, leads me to believe Amd will be A-ok...

we'll see....


Sorry, Mark I corrected it

[Hulk]

Ray,

You have quoted Brian's response to me, those are not my words.
BTW, I agree with you!

Mark

[Brian Ellis]

Sorry, When you are old as I, you tend to get confused -- or should that be defused?

[Ray Austin]

Originally posted by Brian Ellis
Sorry, When you are old as I, you tend to get confused -- or should that be defused?

no problem, I was surprised to say the least. There is so much going on in tech that it's pretty easy to get things confused.

Glad we cleared that up.... I didn't think anyone would ever remember Iridium let alone think it was a processor...hehe

[Technoid]

When the pentium debuted we were on 16bit OS don't forget that

[Hulk]

Brian,

Current best price for an Itanium on Price Watch is $2758!!!

- Mark

[Brian Ellis]

Technoid

That is as may be, but it was the original Pentium, which was a 32 bit processor, which inspired a true 32 bit OS (which was W2k in the Windows series). Now just consider the time between Pentium and W2k and extrapolate that to A64 and a true 64 bit OS... W95, W98 and ME were essentially 16 bit systems with a few 32 bit bits and pieces tacked on, the same as W64 will be essentially W2k (NOT XP) with some 64 bit bits and pieces tacked on. I don't believe MS will take the risk of developing an all-64 bit OS, which will require hundreds of man-years of work, until it is clear that there is a future in 64 bit processors. Would you?

In any case, the only thing in common between the original Pentium and the P4 is the name and the corporate image of the manufacturer!

Mark

I agree that the current price of Itanium is prohibitive, except for top-level servers, but they are currently made in very small quantities on pre-production lines. As the current models are quite slow (I think 1 GHz, if I remember correctly), there is no technical reason why they could not be mass produced on some of Intel's older lines at a price comparable with the Pentium. In any case, the main cost in a CPU is amortisation of the production equipment, which has a very short lifetime, marketing and profit. The actual material and production cost is probably well under 10 bucks, whether it is an 8086 or an Itanium.

[Hulk]

Brian,

Good points, and I appreciate the lively discussion. Yes, the migration from 16 to 32 bits for the PC did take 5 or 6 years, some would say it's still going on actually. But, that doesn't mean it can't be accomplished much faster this time since the market has a successful model for how to do it.

I do believe Windows 95 was 32 bit, except for some key parts of the kernal which were hand-coded 16 bit assembly. This is why 16 bit applications always ran faster on 95/98 than on NT/2000/XP.

Yes, the raw materials of a cpu are very cheap, but that really doesn't matter when pricing a product. Intellectual property has value and many hours of labor behind this that must be amortised. If the demand for full mass production of Itanium is not there, then THAT is a good reason for them not to be mass produced, cost prohibitive for the possible sales.

I do believe Intel is quite the juggernaut and can vigorously respond to any AMD threat, I just hope AMD can continue to fight on!

- Mark

[Ray Austin]

Originally posted by Brian Ellis
the same as W64 will be essentially W2k (NOT XP) with some 64 bit bits and pieces tacked on. I don't believe MS will take the risk of developing an all-64 bit OS, which will require hundreds of man-years of work, until it is clear that there is a future in 64 bit processors. Would you?

I don't agree with this, for one the processor is capable of performing 32 bit functions as well as 64 bit, so naturally this will make any OS a little difficult to implement. I am sure M$FT has been working on this for some time. They have $40 billion in cash the last time I heard so they have no reason to rest on their laurels.

I doubt W64 was built from the ground up however, I think many parts of the XP kernel were re-written to accomodate the A64 architecture. Just how much has changed from XP to W64 remains to be seen, I personally hope to be wowed!

M$FT has no reason to be afraid of risk. So I think it's unlikely they haven't spent or aren't already spending R&D dollars on 64bit software. It's where the market is moving... if it's not then Amd may not be here in another year......