Windoze boxes

[LoadWB]

I have to agree with the statement about Intel stepping up.  AMD long outperformed Intel CPUs.  With the Core2Duo, Intel pulled ahead of AMD for desktop performance, though I still see better server performance from AMD CPUs.

Intel was beating a dead architecture to death with its x86 CPUs.  Even EM64T was originally grafted on top of the 25 year-old x86 architecture.  Something has changed recently with the newer CPUs, and I suspect that they have revamped their x86 emulation core as part of the adoption of 64-bit computing.

Frankly, your question should have been about x64 machines, not x86 machines.  I have been running XP x64 for quite a while now.  It runs very smoothly, and I have been able to find drivers for every device I have, even some more obscure ones.  It does take a little more work, mind you.  In many cases Windows 2003 x64 drivers work, in others the standard XP x86 drivers have x64 compatibility, and in rare situations I have used Vista 64 drivers (in particular for the later case, the modem in my Dell D430 runs the Vista 64 driver.)  I have only had one program not work in XP x64, and I just run it in a VirtualPC instance with XP.

I have toyed with Vista 64 and Windows 7 beta 64-bit on a machine.  The 64-bit operating systems flow much better than their 32-bit counterparts.

Back to the point, the biggest thing Intel has going for it is its vertical integration.  They make the CPUs, motherboards, and even the chassis for servers.  Even though I have preferred AMD for a long time, I just can't beat that for self-built systems.  Sure, I can get an AMD-based server system from Sun or HP, but in most cases off-the-shelf hardware will not integrate.  So Intel takes the cake every time.

For a short while there I was happy building systems with an nForce 4-based motherboard, both workstations and servers.  The price came out right, and the performance and compatibility was awesome.  Then that particular board was discontinued (within months, actually) and replaced with an nForce 5 chipset, which was horrible.  That left me out in the cold.  So, as much as I did not like to do so, I went back to Intel.

I have had some good experiences with ASUS, MSI, and Shuttle motherboards for AMD workstations.  Low cost, top performers, and usually with the exact peripherals I want and need.  I am running an Intel Core2Duo system with 8GB RAM right now since I won the motherboard at a vendor's raffle, and it has been very nice.

I still would like to stay with AMD.  Even though the Intel will be more pricey, as I plan a replacement for my main server running Solaris 8 on an AMD Athlon 1900+ on a DFI motherboard, a quad-core Intel CPU system under Solaris 10 will most likely be my choice for a personally-built server.

But back even more to the point, unless you are a gamer, a workstation is pretty much a workstation regardless of what you pop under the hood.  AMD, Intel, it really does not matter so long as the system meets your specifications and budget.  Either system can be hampered by poor motherboard capabilities, particularly Intel integrated graphic chipsets -- Intel does NOT know how to do graphics and should stay the hell away from graphic chipsets.

Another short-coming on Intel motherboards is the Intel Matrix RAID.  It handles RAID 0 and RAID 1 very well.  Some boards handle RAID 5 at a cost: you cannot manage a RAID 5 array in BIOS, you MUST have a working operating system and use the Matrix management software.  RAID 1 works well for desktop systems, anyway, so that is pretty much a moot point.  For servers I prefer 3Ware RAID controllers.

[LoadWB]

Trev wrote:

@LoadWB

A friend of my wife used to work in graphics driver development at Intel here near Sacramento (Intel has a campus in Folsom, just outside the city--worked there myself about 13 years ago). So, I'd disagree with you a bit on the software side. Intel's target market (general purpose graphics) is quite a bit different than AMD's and NVIDIA's, though.

The integrated Intel graphics chipset performance has been historically abysmal at best, from practical exposure as well as numerous.  Some of the newer stuff, read as "more expensive," is not so bad, but running full-screen video on the newer 945-based desktop boards still produces an extraordinary amount of heat, as well as pauses and what seems like the drive and video subsystems fighting for bus bandwidth.

I never understood why the 815 and 915 chipsets were so horrible at full-screen video, even under 1024x768, when a PCI or AGP card from ATI could handle it just fine.  The 915 chipset was so horrible that an otherwise capable AGP video card was so starved for bandwidth that it could not handle full-screen video, either.  Which could have been the general problem, not so much the video portion of the chipset.

Heck, even integrated video driven by ATI performed better than Intel's integrated video.

I would be very interested in what insights your driver developer friend could offer on Intel graphics *ahem* accelerators.

Something interesting I discovered on my D430 is the integrated chipset, the Mobile 945 Express, shares memory with system RAM, but it seems to be dynamic.  That is an interesting concept.  Full-screen performance on this system is adequate although the system fan spins up like a hurricane.

I suppose, then that the target video audience for Intel could be compared to that of Matrox.  Matrox was absolute pants for 2D or productivity video.  I keep several Matrox PCI and AGP cards around for building cheap systems, especially for Solaris 8 stuff.  (Never really got into Solaris 9, and 10 has been a pain for compatibility on older equipment.)