Someone asked mw what would be a good time to buy for a major upgrade and when to buy it.
This question is much harder nowadays than it was in the past. There’s a lot of good options out there, and, depending on your priorities in other areas, an argument can be made that any of the following is “best.”
$50 Wonders I’m getting some squawking that I’ve been terribly unfair to the “J” stepping TBredBs. A few people have reported levels of performance approaching “A” TBredB levels.
While I still think the “A” TBredBs are a better bet, all this may mean in your eyes is that I have less respect for the extra $50 an “A” will cost over a “J” than you do. Fine by me, it’s your money.
If you think this is a great way to update a system with an older mobo, or a cheap fix while awaiting a cheap Barton solution, though, you really ought to make absolutely sure your mobo supports it. If it doesn’t support a 2400+; odds are it’s not going to support a 1700 or 1800 “J,” either.
The $100 Safe Bet The 2100+ “A” TBredBs have been consistently getting in the range of 2400MHz, maybe a little less, maybe a little more, with not too much pushing. If money is an important but not overwhelming factor, this is a reasonably cheap, reasonably safe choice.
How Good Is Little Bart?
It looks like we’ll see Barton in a couple weeks. Now we’re moving into unknown territory. There will be three Bartons. Only the one at the bottom of the totem pole is somewhat reasonably priced.
We don’t know yet if the 2500+ will do as well MHz wise as the $100 TBredB. They really do need to match them to make this a viable overclocker choice.
Even then, is an extra 5-7% worth an extra $75? Depends on how much you value your money. Depends on how you value time, too. That $75 not only buys you an extra 5-7% of performance, it buys it for you for some length of time. Odds are the 2500+ won’t get down to $100 until the fall. That alone isn’t compelling, especially if you’re fat and happy with what you have, but it’s something to think about.
I wouldn’t expect any major increases in MHz in the TBredB/Barton line so long as it remains a .13 micron chip (and whether there will ever be a .09 micron TBred is unknown at this time).
nForce boards seem more than a little quirky, enough people are having problems with changing and savings multipliers no matter what brand they have to blame it on human error, but others seem to do fine.
Unfortunately, Via doesn’t provide much competition at the moment, so it’s really a matter of watching and waiting to see if BIOS changes clear up these irritants.
Intel: The Mobo Is The Message
Which will be the fastest overclocking machine in 2003? It’s pretty safe to say it will be an Intel-based machine; the only question is when.
At this point, I wouldn’t buy a new Intel core system. It’s too close to the introduction of the first true mass market dual DDR mobos. We’ll probably see them arrive in April.
There are two big reasons why these mobos are worth waiting for.
1) Dual DDR should add about 7% to overall performance
2) These mobos for sure will be compatible with at least the intial versions of Prescott, the next generation of PIV. In fact, Prescott was initially planned to come out
at the same time as these mobos.
Faster But When?
The CPUs are more of a problem. There are still plenty of B0 stepping chips still in the pipeline. Cheap C1 stepping chips often exceed 3GHz, but often fall a bit short.
If you read between the lines in some of Intel’s statements, it’s fairly clear they think there’s more than a bit more life left in the Northwood. A recent Inquirer article noted that Prescott will debut at 3.8GHz. Since new generations of Intel processor normally start where the old generation left off, we’ll probably see at least a 3.6GHz Northwood at some point in time.
Though not many have shelled out the money to do so, those who have overclocked a 3.06 have found that they perform significantly better than their junior partners.
So I think it’s pretty likely people will be able to get to 3.5GHz or better without too much effort with cheap Northwood PIVs at some point in time, but when? Might be March, more likely will be closer to November.
April should also bring somewhat reasonably priced hyperthreaded PIVs. As things stand now, this is going to be a rather perverse CPU, great for Joe Sixpack, terrible for overclockers.
Hyperthreading is a somewhat dubious technology for those out to get one thing running very, very fast, but it certainly will make the computing experience for the average Joe running a couple not-too-demanding apps smoother.
The real problem for overclockers, though, is the 200MHz FSB all these processors will run on. Unless mobos start allowing you to run FSB faster than memory rather than the other way around, there’s just not much overclocking room on these chips.
A 2.4GHz 200MHz FSB CPU may be inherent capable of, say, 3.6GHz, but to get to that speed, you’d have to run it at 300MHz FSB. The average overclocker is not going to get DDR to run at 300MHz; that’s DDR-II territory.
So unless BIOS setting change dramatically, your choice will be to overclock a lot with a 133MHz FSB chip, or get hyperthreading. Not both.
Price Advantage: AMD
Intel solutions will likely end up somewhat faster, but that little extra will likely cost you a good deal more. It’s hard to see a PIV/Canterwood/dual high-speed DDR core system costing less than $500, and $600 is more like it.
Equivalent socket A core replacments systems will probably cost a couple hundred dollars less, and many with existing socket A systems will certainly upgrade less than that.
The Dark Horse: Clawhammer
It looks like AMD has changed its plans for desktop Hammers. The initial Athlon 64 looks like it will have a 1Mb cache, while a cheap version with just 256Kb cache will debut in the fall, no doubt to be or become the future Durons in AMD’s lineup.
Desktop Hammer solutions (at least from the overclocker’s perspective) look like they’ll be even more expensive than the Intel options. For this reason, it is unlikely that they’ll make many people give up their socket A systems.
The only way that will change will be if AMD offers awfully cheap low-speed 1Mb Hammers (say, $150 or less) that are hellacious overclockers. I don’t think that’s going to happen with the 1Mb version, but if it does, it probably won’t be until the summer/fall.
What is far more likely to happen is that AMD will offer cheap 256K Hammers at that time, but I don’t think they’re going to be too appealing to the socket Aers. It’s going to take a .09 micron chip to get them to change platforms.
There’s No One Solution
I could reasonably argue for any of these choices, then turn right around and just as reasonably argue against them. There’s no one “reasonable” choice for all here; what you decide on depends more on you than the hardware.
There’s no clear single “best” choice here. It’s more like a sliding slope. You’ll basically get what you pay for, the choice really depends on how much is “enough” for you and how much you’re willing to pay for it.