----- Original Message ----- From: "Benjamin Scott" <[EMAIL PROTECTED]> To: "Greater NH Linux Users' Group" <[EMAIL PROTECTED]> Sent: Tuesday, April 23, 2002 1:58 AM Subject: Re: AMD vs Intel (was: Hardware Pointers)
> On the other hand, I've used high-quality boards with VIA chipsets that > have never given me a second of trouble. The difference is in the quality > of the board, not the chipset. > The fact that VIA's implementation of the AGP specification is flawed or their AC97 sound codec is worthless has nothing to do with the board manufacturer. > > Bugs is one thing. How fast the workarounds are created is another. > > I'm really not sure what you are trying to say there. If you mean the > speed of the fix... well, if I remember correctly, Intel had to face major > media pressure before they even admitted the FDIV bug existed... ? > > > True, it is a perception issue, but it's still important. > > Eh? Are we talking technical merits here, or who has the bigger marketing > budget? :) If marketing muscle is an "important" factor, then I guess I > should stop using this Linux thing... ;-) > I'm saying that Intel is still the "reference standard" like it or not. If a software program works on Intel, and not on AMD, they will first blame AMD before blaming themselves. And yes it IS all who has the bigger marketing budget and the largest market presence. As far as your Linux comment goes, that is irrelevant. When two corporations are fighting for the consumer dollar, then yes, marketing muscle is important. Linux is not "fighting" for any market share, and the large corporations with the big marketing budgets have nothing to fight against. Where Linux is concerned, marketing is irrelevant. That's really what the corporate types don't understand. > > The errata is masked by the fact that most people use Windows. Server > > admins mostly use Intel. > > I really don't get what you are trying to say here. How are processor > bugs masked by the fact that most people use MS Windows? And what does the > fact that server admins mostly use Intel have to do with that? > System halts are common enough with Windows. Reboots are frequent and accepted by users. Any processor bugs are more than masked by failures in the OS and the software that runs on those processors. (Why the HELL do you need to reboot an NT 4 machine to clear the PRINTER QUEUE???? But you do....) Servers are expected to run continuously, 24/7, under anything from idle conditions to heavy loads. ECC RAM, SCSI RAID, redundant power supplies, UPS, Intel processors and Linux or BSD all go towards the reliability and continued functionality of these machines. > > There is also the latency issue, which I have yet to have confirmed or > denied to my satisfaction. > There IS higher latency with RAMBUS, which is finally becoming evident now that DDR is approaching RAMBUS's memory bandwidth. We are seeing boards that utilize DDR400 (still an unratified spec) beating boards using the 850 chipset and RAMBUS. However, RAMBUS is serial and capable of much higher speeds, which can overcome the latency issue. We also see latency within the CPU core. This is being addressed in various ways, including speed, pipelining, and branch prediction. Just because a system has higher latency doesn't necessarily mean it's a bad design. > > But the fact that you CAN overclock an Intel processor much more than an > > AMD processor ... > > Would you mind backing that up with some references? I am not an > overclocker, but I do read about that stuff from time to time, and > everything I've encountered has indicated a strong preference for AMD over > Intel in the OC world. My understanding is that Intel makes it much harder > to overclock their chips. A quick check of the CPU database at overclockers.com yielded these results: AVERAGE overclocking results from an Athlon XP 1900+ (1600 MHz): 1800 MHz or 12.5% AVERAGE overclocking results from an Intel P4 1.6A (1600 MHz): 2351 MHz or 46.9% The MOST overclockable AMD I found in a quick perusal of the database was a Morgan 1GHz processor that averaged 1245 MHz or 24.5%. Please let me know if you find higher ratios. I didn't go through the whole thing, and I don't know which AMDs are the "most overclockable." > > There is also the issue of the 1.13 GHz Pentium III, which was *already* > overclocked, by Intel, beyond stable limits. That's not what I would call > "much more", but rather, "much less". :-) > As with both companies, there are CPUs that overclock well and those that don't. Obviously, the 1.13 GHz PIII you mention (I asssume you mean the Coppermine not the Tualatin) is already at the top (or past the top?) of it's frequency limit. The "most overclockable" CPUs are the ones that are at the bottom end of the frequency range for a particular core design. Thus the 1.6 Northwood is much more overclockable (46.9% avg increase) than the 2.2 Northwood (31.4% avg increase.) When perusing the AMD database I looked at the lowest speed of each core design. While 24.5% was the highest I found, the average was more like 12 to 20%. > > ... says a lot for the "overhead" that these processors have in terms of > > stability and reliability. > > You could look at it that way. You could also say Intel is selling you a > product that has been deliberately de-tuned. If the chip is capable of > running stable at 1200 MHz, I would like the manufacturer to sell it as > such, and not lock it at 800 MHz just to keep prices high. Of course. That's exactly why Intel locks the multipliers on its chips. But now the motherboard manufacturers are responding by allowing the FSB to be changed without affecting the rest of the hardware. AMD locks its multipliers too, it's just that the way they do it is not impervious to a determined overclocker with a pencil. :) > > By all accounts, Intel is killing off the PIII line not because its time > had gone, but because the price/performance ratio beats out the P4 by a > considerable margin. > More likely it is because the PIII crosses the gap a little too well between SMP-able Xeons and non-SMP-able workstation processors. Rich Cloutier President, C*O SYSTEM SUPPORT SERVICES www.sysupport.com ***************************************************************** To unsubscribe from this list, send mail to [EMAIL PROTECTED] with the text 'unsubscribe gnhlug' in the message body. *****************************************************************
