Click here to Skip to main content
15,891,976 members
Home / Discussions / Hardware & Devices
   

Hardware & Devices

 
AnswerRe: Print server Q Pin
Mike Dimmick12-Oct-07 11:26
Mike Dimmick12-Oct-07 11:26 
NewsIntel or AMD [modified] Pin
Benjamin Dodd30-Sep-07 5:58
Benjamin Dodd30-Sep-07 5:58 
AnswerRe: Intel or AMD Pin
User 171649230-Sep-07 14:29
professionalUser 171649230-Sep-07 14:29 
GeneralRe: Intel or AMD Pin
Benjamin Dodd30-Sep-07 21:31
Benjamin Dodd30-Sep-07 21:31 
GeneralRe: Intel or AMD [modified] Pin
User 17164921-Oct-07 1:14
professionalUser 17164921-Oct-07 1:14 
GeneralRe: Intel or AMD Pin
Benjamin Dodd1-Oct-07 5:51
Benjamin Dodd1-Oct-07 5:51 
GeneralRe: Intel or AMD Pin
Dan Neely1-Oct-07 2:18
Dan Neely1-Oct-07 2:18 
GeneralRe: Intel or AMD Pin
Mike Dimmick12-Oct-07 12:28
Mike Dimmick12-Oct-07 12:28 
I've always preferred Intel. In the Pentium II and III days Intel typically had a clear lead, but were usually more expensive. AMD started to pick up a performance lead towards the end of this era. However, I tended to observe that the computers using VIA and SIS chipsets were less reliable than Intel chipsets. When all the processors used Socket 7 (Pentium and AMD K5/K6) you could match an AMD processor to an Intel chipset; once Intel moved to Slot 1 for the Pentium II and the AMD clock rates started climbing, the old Intel chipsets couldn't cope any more leaving VIA and SIS as the only (flaky) choices.

It could be that the behaviour of some of the owners of the computers was actually the cause of the flakiness - generally they'd swap components over regularly and install and uninstall a lot of (often slightly dodgy) software. Still, people with Intel systems did the same and they were generally more reliable.

Pentium 4 - NetBurst architecture - was an attempt by Intel to massively ramp the clockspeed and it was designed to make best use of fast clocks, by having a long pipeline in which a little progress was made in each stage. If you're not familiar with electronics, it takes a certain amount of time for the signal at the output of a series of logic gates to stabilise at the correct result, or in the case of latch circuits, to latch at the correct result. If the result is sampled too soon (with a faster clock), the incorrect result can be latched and you get incorrect answers.

To get the most benefit from this, though, the pipeline needed to be kept filled, and instruction dependencies, memory latency, and branch mispredictions tended to mean that it couldn't be kept full, meaning cycles were wasted. In the case of a dependency error or branch misprediction the whole pipeline has to be thrown away and restarted to recompute the correct operations.

The trouble is, the design was intended to ramp to 10GHz and as we know, we never got there - 'NetBurst' Xeons topped out at 3.8GHz or so. The problem was simply heat - too much current was leaking, causing the circuit to consume more power and emit it as heat rather than do useful work. AMD, on the other hand, kept a shorter pipeline consistent with its older models, with larger amounts of work done per cycle. In the end the Athlons were able to clock high enough to surpass the P4s at much lower clock rates.

At this point the choice of chipsets for an AMD system appeared to be VIA, nVidia or ATI. nVidia and ATI can't get a video board right (I use whichever the computer manufacturer chooses because they're equally bad) so I hold out no hope they can get a chipset right, even if the memory controller is now part of the processor package.

Meanwhile Intel were effectively still developing the Pentium III for laptops - as the Pentium M - as the NetBurst architecture was utterly wrong: the power characteristics would only be suitable for 'desktop replacement' laptops that were only moved from place to place, as the P4 would flatten virtually any battery in minutes. Towards the end of the P4's life it was clear that the Pentium M was actually faster, not just clock-for-clock but overall, and you saw a few desktop boards. Result, a massive about-face, and the new desktop processors eventually christened 'Core 2' were a derivative of the Pentium M (Virtual PC 2007 still identifies my Core 2 Duo T7200 laptop as an 'Intel Pentium III class'). It's reasonably clear that Intel have taken back the performance lead.

There's a bit of back-and-forth over different Single-Instruction-Multiple-Data instruction sets (MMX, 3DNow!, SSE, SSE2, SSE3, SSE4 I think is in the works) but on the whole, and despite much encouragement from Intel, I think most code still uses the antique x87 FPU instructions for floating-point operations. The NX (No eXecute) bit added by AMD in the Athlon 64/Opteron has been adopted by Intel (as eXecute Disable, XD); both are now offering incompatible virtualization support, but Virtual Server 2005 R2 SP1 supports both types and I think VMWare supports both as well. AMD invented 64-bit extensions to x86 and after a bit of arguing, Intel adopted them as well (with a few minor omissions and differences, but there's enough common ground to write an OS to).

Intel's own attempt at a next-generation 64-bit architecture is much more elegant (you don't get much less elegant than bodging 64-bit extensions onto a 32-bit architecture that was itself already a bodge of a bodged 16-bit architecture - 8086 was crap from day one) but bizarrely has been stuck at low clock speeds - while it might be much more efficient clock-for-clock than x86, Core 2 (Xeon 51xx, 53xx) and Opteron are faster overall because they have faster clocks. They're also a heck of a lot cheaper. It's no wonder Itanium didn't take off (it didn't help that the built-in x86 emulation was woefully slow).


DoEvents: Generating unexpected recursion since 1991

QuestionAMD Turion x2 Mobile Technology TL-56 (64bit) Processor Pin
Civic0628-Sep-07 5:00
Civic0628-Sep-07 5:00 
AnswerRe: AMD Turion x2 Mobile Technology TL-56 (64bit) Processor Pin
Dan Neely28-Sep-07 5:43
Dan Neely28-Sep-07 5:43 
AnswerRe: AMD Turion x2 Mobile Technology TL-56 (64bit) Processor Pin
#realJSOP30-Sep-07 3:41
mve#realJSOP30-Sep-07 3:41 
QuestionHow to get memory's manufacturer? Pin
simonchen.net27-Sep-07 20:24
simonchen.net27-Sep-07 20:24 
AnswerRe: How to get memory's manufacturer? Pin
Dave Kreskowiak28-Sep-07 1:05
mveDave Kreskowiak28-Sep-07 1:05 
GeneralRe: How to get memory's manufacturer? Pin
simonchen.net12-Mar-08 5:31
simonchen.net12-Mar-08 5:31 
GeneralRe: How to get memory's manufacturer? Pin
Dave Kreskowiak12-Mar-08 10:58
mveDave Kreskowiak12-Mar-08 10:58 
AnswerRe: How to get memory's manufacturer? Pin
#realJSOP30-Sep-07 3:42
mve#realJSOP30-Sep-07 3:42 
GeneralRe: How to get memory's manufacturer? Pin
Luc Pattyn30-Sep-07 4:34
sitebuilderLuc Pattyn30-Sep-07 4:34 
QuestionLPT Listener Pin
TPN26-Sep-07 21:19
TPN26-Sep-07 21:19 
AnswerRe: LPT Listener Pin
Luc Pattyn27-Sep-07 1:16
sitebuilderLuc Pattyn27-Sep-07 1:16 
GeneralRe: LPT Listener Pin
TPN27-Sep-07 20:21
TPN27-Sep-07 20:21 
GeneralRe: LPT Listener Pin
Luc Pattyn28-Sep-07 0:45
sitebuilderLuc Pattyn28-Sep-07 0:45 
GeneralRe: LPT Listener Pin
LittleYellowBird28-Sep-07 1:07
LittleYellowBird28-Sep-07 1:07 
QuestionAnalog meter display on PC Pin
Pluto724-Sep-07 23:50
Pluto724-Sep-07 23:50 
AnswerRe: Analog meter display on PC Pin
Dave Kreskowiak25-Sep-07 4:17
mveDave Kreskowiak25-Sep-07 4:17 
GeneralRe: Analog meter display on PC Pin
Pluto727-Sep-07 21:03
Pluto727-Sep-07 21:03 

General General    News News    Suggestion Suggestion    Question Question    Bug Bug    Answer Answer    Joke Joke    Praise Praise    Rant Rant    Admin Admin   

Use Ctrl+Left/Right to switch messages, Ctrl+Up/Down to switch threads, Ctrl+Shift+Left/Right to switch pages.