Intel Out of Touch with 64-Bit Reality?

<A HREF="http://slashdot.org/comments.pl?sid=54835&cid=5371889" target="_new">http://slashdot.org/comments.pl?sid=54835&cid=5371889</A>

This guy from Epic Games thinks so...


I want to move to space, so I can overclock processors cooled to absolute zero.

Although we all see 64-bit as unneeded in the current home user world, it seems professional game programmers are seeing it as actually bringing performance increases.

--
This post is brought to you by Eden, on a Via Eden, in the garden of Eden.

I don't see how going to 64-bits will help much. I have under 1GB of RAM, and I don't see a pressing need for more when I'm playing a game or using an app? If anything, a better vid card and larger hard drive would be better for me, than 10GB of RAM.

Also, the OS for a 64-bit consumer based processor isn't available, so I can only use Linux or some other pure 64-bit OS for the total 64-bit experience, or just throw Windows on, but miss out on fully utilizing the 64-bit capability.

64-bit CPUs might be needed, someday, for consumers, but I don't see it happening this year. It's like HT, it's nice, but nothing uses it in my everyday apps/games, so why do I need it?

What?<P ID="edit"><FONT SIZE=-1><EM>Edited by nmdante on 02/27/03 06:47 PM.</EM></FONT></P>

We got 2 benchmarl

1 sap about the same
2 Spec 2/3 of the speed of Itanium 2 (will be disband when opteron be in the market)

No CPU can rival or in the future will rival for a long time with IA-64 family.

Just next to the lab and the bunker you will find the marketing departement.

The pre-release Opteron actually fared rather better in the SAP benchmark. I don't remember if it was running at its planned release speed, but I seem to recall that it was somewhere below that.

Itanium 2 did about 10-15% better than pre-release Opteron in SPECfp and fell behind by about 30% in SPECint.

<i>I can love my fellow man...but I'm damned if I'll love yours.</i>

That was sap benchmark and was a projection Compare to the one of the slowest Itanium 2.Dont get too fast nothing can beat Itanium performance wise.I just eager to see the so call ""Four fold power" of Power 5 SMT.

Just next to the lab and the bunker you will find the marketing departement.

This game have work on UT some a pro AMD

Just next to the lab and the bunker you will find the marketing departement.<P ID="edit"><FONT SIZE=-1><EM>Edited by juin on 03/01/03 12:24 PM.</EM></FONT></P>

All x86 processors seem to perform disproportionately well in SpecInt. The P4 3.06 got an excess of 1000 and is currently the best performer in SpecInt out of any processor (Power4, Itanium2, etc etc.). However, when you look at actual transaction scores, the Xeons are not on top, not in SAP, not in SQL, not in DB2. So I wouldn't go so far as to say being able to score better in SpecInt gives an advantage to any integer-intensive server programs.

Last I heard, AMD's next die shrink wasn't till 2004. With so long to get Opteron out, how are you sure they will "actually have some process refinement" within 2 months or so? Madison will have both a 50% boost in clockrate (to 1.5 GHz) and an added 6 MB of L3 cache (compared to the current 3 MB on McKinley). Its .13 micron copper interconnect process will probably allow it to scale a bit higher (look what it did for the P4). However, I think for a while, 1.5 GHz is the number.

Last I recall, they'll be released at 1.4, 1.6 and 1.8 GHz. Haven't heard anything about the 2.0 GHz model. BTW, the SpecInt and FP score were of a 2.0 GHz model, or so says AMD.

"We are Microsoft, resistance is futile." - Bill Gates, 2015.

While this may be true (far be it from me to comment), I fail to see the impact this will have from a performance standpoint. Does AMD plan on releasing 2 or 3 new speedgrades right after launch? If they launch at 1.8 GHz. Would they release a 2.0 and 2.2 GHz within 2 months of launching? That'd completely destroy their marketing and pricing models.

The P4's seem to be getting a lot more out of their .13 micron copper design. The limitation being power and thermal restrictions (taking in the overclocking results vs the Willamette, which had similar power and thermal restrictions). Also take in mind the yields on the P3 Tualatins, which seem to be 1.6+ GHz looking at the overclocking results vs a flat 1 GHz (pretty much stuck at that) for the CuMines (which ironically used aluminum interconnects).

"We are Microsoft, resistance is futile." - Bill Gates, 2015.

I doubt this. AMD has said themselves that Opteron will not be positioned in a place to challenge the high-end Itanium 2 line. They're probably aiming for something the likes of Deerfield, which I will agree, does not neccessarily hold a performance advantage to Opteron even if it was released at 1.8 or 2.0 GHz.

The majority of overclocks done with the Willamette, even with extreme cooling, as I recall, hit a 2.2-2.3 GHz barrier. Compare that with people who got the Northwood to 3+ GHz with extreme cooling upon release and are not going past 4+ GHz with extreme cooling. Near a 90% clockspeed improvement.

It'd be interesting. Wonder if Intel releases any samples that allow frequency adjustement. I do wonder why Madison would offer much more heat than McKinley though. Much of the added die is just cache and that doesn't add much heat at all seeing how most of it is idle most of the time. And a 50% increase in clockspeed shouldn't offset completely the transition to .13 micron and copper interconnect. Come to think of it, I wonder why the Itanium produces so much heat to begin with. Its core size can't be more than 15ish million transistors and most of it is just cache.

"We are Microsoft, resistance is futile." - Bill Gates, 2015.