Originally posted by: dherman76
Dear Anandtecher's,
I'm an assistant to my professor at Skidmore College in Saratoga Springs, NY and its my turn to lead a discussion on Hardware in Monday's Management Information System Class. I'm very proficent in hardware - understand how everything works, etc, but have been out-of-the-loop so to speak for about 2 years about the fastest this and the fastest that. I basically have been out-of-the-loop since the 2GHZ PIV chip and the 7200RPM HD.
What I'm asking to you is:
What would you think people/students would like to know about today's computing (especially hardware)? What is the fastest processor? The fastest MOBO chipset? RAM? HD? How has it evolved the past few years/months? Is Intel and AMD still head to head? What about the chips in the new tablet pc's?
Another question I pose...Where is computing going? Where will it be in 2-3 years? 10 years? I'm interested to hear your responses, and trust me, you'll be fully credited in my speach to class.
Thank you for reading!
Darren Herman
Skidmore College
Class of 2004
Asst. Professor MIS
d_herman@skidmore.edu
The fastest processor is essentially a tie between the Pentium 4 3.06 GHz and the Athlon XP 2800+ Tbred (besides the upcoming Barton 3000+ XP -which is just an XP with 512K of L2 cache).
I'll explain the above (and it's very pertinent to your discussion).
The P4 3.06 is the first Hyperthreading CPU out in the market, and intel is going to make their entire P4 top-to-bottom HT enabled. Hyperthreading (if you don't know) is used by the processors to enable multithreading on a single CPU. Specifically, Windows (2000/XP) views the P4 3.06 GHz as two separate CPU's that can each do one thread at a time, so the degree of paralellism in even single-CPU solutions is increasing. Hyperthreading right now gives a 0-40% speed improvement (depending on application), and makes running several programs/processes much more efficient (it reduces the amount of CPU stalls, etc). This runs on, of course, intel's 0.13um Northwood core with 512K L2 cache (full speed, of course) and Intel is rumoured to be moving to a 1MB L2 cache very soon.
AMD has had a very sucessful transition to .13um also, with the second revision of their Thoroughbred core (Thoroughbred is the code name for the .13um Athlon XP). The first revision (Tbred A) wasn't very scalable, and seemed to top out at ~2 GHz, so they went back to the drawing board and added another layer to the CPU and some other minor additions (too technical for me at the moment - check some reviews of the Thoroughbred B architecture). Anyways, the Tbred B right now overclocks to ~2.4-2.5 GHz on air, and AMD sells up to 2.25 GHz (which is rated at 2800+).
The two architectures (AMD's XP and Intel's P4) are very different, yet in the end result in similar performance. Intel went for speed with their 20 stage pipeline and weak Floating Point performance (which is supplemented with SSE2, a highly parallel floating point set of operations, like MMX for the FPU). AMD has a core that only has a 14 stage pipeline I believe, and is based on slightly older technology, with only SSE (version 1), and therefore less FPU specific optimizations. However, as a result of it's shorter pipeline, and naturally stronger floating point performance, the Athlon XP's IPC (instructions per clock) are much higher than the P4 Northwood as the same frequency.
As for Motherboards/platforms, both Intel and AMD are switching to dual-channel DDR SDRAM motherboards for the highest bandwith and lowest latency. AMD has the nForce 2 motherboards from Nvidia (fastest platform for AMD chips right now, with dual channel DDR SDRAM) and Intel has the upcoming Granite Bay (from Intel) which is also a dual channel DDR platform. If you remember, RDRAM had a ton of bandwidth (PC1066 has 4.2 GB/s of bandwidth in a dual channel setup I believe), but very high latency. DDR SDRAM offers the best of both worlds - high bandwidth (with dual channel DDR333/400) and lower latency.
You can also talk about the difference between RDRAM and DDR SDRAM - my friend in Computer Engineering was talking about how he learned in class that RDRAM was actually a much smarter design (in theory) with it's high bandwidth
serial design. DDR SDRAM is an extension of classic
parallel DRAM. Remember, RDRAM has serious flaws in practise however - it's higher latency compared to SDRAM was a big killer. Plus DDR333 (also called PC2700 SDRAM) offers 5.4 GB/s of bandwith in a dual channel setup (2.7 GHz *2), so the bandwidth edge of RDRAM is no longer a factor anymore. RDRAM's design of "ultra high frequency" RAM has failed since the frequecy hasn't scaled up fast enough in relation to DDR (Which has doubled in the last couple of years).
HD's just keep getting bigger all the time, platter sizes are unreal; 80GB/platter is the new standard and I belive 100 or 120 GB platters are coming out soon. More GB/platter means faster disk transfers since more data tracks fit into the same area. 7200 rpm is still the max on IDE drives; SCSI is, and has been at 10,000 and 15,000 rpm for awhile, but it's far to expensive to implement in consumer HD's.
How are modern home PC's evolving? More RAM all the time (of course), more on-chip L2 cache sizes, more transistors per CPU, smaller and smaller manufacturing processes (0.09um coming very soon).
There has been a shift of power in the video card market - ATI is now top dog and nVidia is playing catchup (read more in Anand's review of the GeForce FX on this site).
AMD was lagging before but have pulled out 2 aces in the hole in a row now: first Tbred B (which scales much higher than originally expected of the XP core) and now Barton (an .13um XP with 512K of L2 which further increases performance and IPC).
Tablet PC's - no idea. As usual though, laptops keep getting slower versions of mainstream CPU's, so there are XP 1500+ laptops and above now.
Where will comping be in 2-3 years? Fast! 5 GHz seems possible in 2 years, as does 1-2 GB of RAM in 2-3 years. I believe more than 4GB of RAM requires 64-bit addressing, which is coming in AMD Hammer in a year anyways.
One very abrupt transition is the easy acceptance of flat panel LCD's. They are dropping in price at an astounding rate (you can check this out online). Apparently, by 2005-2006 or so (I think that's when it was), flat panels should be comparable to CRT's in price (so says my materials prof
).
That's all I can think of off the top of my head. Best of luck to you!
Facebook
Twitter