Hi all,

I'm looking to upgrade my desktop computer, and to overclock it.

I understand that getting a Barton made before week 39 (unlocked) is pretty hard, but that all Mobile Bartons are unlocked.

Will a Mobile Barton (let's say 2500+) work on an Abit NF7 v2.0 mobo (i.e. will this mobile CPU work in a desktop environment) ?

Also, shoud I get high quality DDR400 or medium quality DDR466 or DDR500 ?

Thank you all in advance,

Razvan

yeah...i got the same question...help fast!...ordering this afternoon!

Mobile Bartons work on NF7. If I were you I'd get the mobile 2600+. The mobile Bartons are specially selected regular Bartons which runs the rated speed at much lower Vcore. The one rated mobile 2600+ will have a little bit more headroom in Vcore than the mobile 2500+ for overclocking.
As for RAM, get qulity DDR400 with tight timings, sigle channel or dual channel if you board supports it. DDR466 and DDR500 are not neccessary because most of the mobos support Barton will top around 220 FSB.

Hi all,

Thanks for the replies.

I'll try to find a Mobile Barton (although it's very hard to find around here - Bucharest, Romania).

And I'll go for quality DDR400 (anyway, to my knowledge, quality DDR400 - like OCZ for instance - will go up to DDR466 anyway if cooled right).

How about GeIL DDR ? Is it good ? They have good prices.

And between Corsair XMS, Kingston HyperX and OCZ DDR400 CL2, which should I choose ?

As I can't seem to put my hands on a pre-week 39 Barton around here, any thoughts on how a Super Locked Barton could get Super Unlocked ? :)

Thanks again.

All the rams you mentioned are good, making good DDR400 is an easy feat these day. Get the one with the tightest timings in your price range.
You cannot unlock Bartonsl, the mobile Bartons are factory unlocked.
If you cannot find one in your country, try newegg.com to see if they ship overseas.

Make sure you get some thermal paste AS5, a dependable PSU and good HSF combo too if you plan on overclocking.

____________________________
The greatest risk is not taking one

Report Offensive Follow Up For Removal

Right now I have a noname 400W PSU which handles pretty well, but I'm planning on getting either an Antec TruePower or a ThermalTake Pure Power over 420W.

The HSF I'm planning for is either a Thermaltake Extreme Volcano 12 or a Polo 735, or a Thermalright SLK-series (which is pretty hard to come by around here).

I will get the Arctic Silver as well.

Oh.. and newegg doesn't ship overseas :(

"high quality DDR400 or medium quality DDR466 or DDR500?"

medium 466. With "high quality" you get tight timings which are not as important as mhz. But if your mobo cannot exceed 210 or 215 you might get higher priced 400 so you can use the tighter timings. In general though, memory timings are way overpriced considering the minimal throughput impact.

"The one rated mobile 2600+ will have a little bit more headroom in Vcore than the mobile 2500+ for overclocking"

Each chip is rated at 1.45v so I do not understand this point.

As we all know CPUs are made of thin film of semiconductors and dielectrics. You need apply bias on the film to make the structure work. But film thickness varies from site to site on a wafer, especially from center to edge. Chips made of the film from these site on the wafer will need different bias to reach the same performance. That is reflected on different Vcore in regular and mobile bartons. Simply put, mobile Bartons have thinner film in them than regular Bartons. Therefore, mobile Bartons require lower power or Vcore to reach the same speed as regular Bartons. This is the case in different rated mobile Bartons also. Ideally, mobile 2600+ needs less Vcore to reach a specific speed than mobile 2400+ or 2500+. Say a mobile 2600+ can be oced stable to 2.6GHz at 1.8V Vcore, a mobile 2500+ may need 1.85V to do that. Or put it another way, at certain bias voltage chips with thinner film can be rated at a higher speed. That's the headroom I was talking about.
This brings me to another point I want to make, people always worry about heat generated by increased Vcore and think if cooling is appropriate to dessipate the heat and keep the temp low, Vcore can be raised to whatever they want. That's wrong, the thin film in the chips has a breakdown electric field strength. You are going to destroy the film in there if your Vcore is too high. So keep your Vcore in a reasonable range.

Thank you all for your detailed responses ;)

Right now, the big battle around here is to get an unlocked Barton (nearly impossible) or a mobile Barton (nearly impossible as well).

With the DDR, it's another piece of pie. There's a decent choice, but most of the market is invaded by OEM DDR.

For approximately the same price I could get something like GeIL or Wintec AMP-X memory rated at DDR466 up to DDR533 with CL2.5 or CL3 (in a medium to high price range) or OCZ, Corsair XMS, Kingston HyperX, Mushkin DDR400 but with latencies like 2-2-2-6.

Considering the fact that this upgrade will take a lot out of my pocket, I was thinking of trying something like Wintec or GeIL DDR466 or more and in the end if it's not stable enough, drop them to DDR400. Like this I get the headroom.

Plus the fact that I have an nForce2 based mobo (Abit NF7 v2.0 - I'm thinking of upgrading to an NF7-S v2.0 or the newer AN7) and in my SoftMenu III I get to set a divider in order to get different DDR mhz related to my FSB.

And anyway, I'm thinking that if the cheaper DDR has CL 2.5 at 466, it will do CL 2 at 400. Right ?

That was the general idea.

So, johnoh (respect flyes out to U for Ur experience), what do you advise me to do ? Should I do it the cheaper way and if it's crappy drop to DDR400 or get some bada$$ DDR400 to start with and hope they'll go up the mhz scale ?

Thank you again, all.

Razvan

That is an excellent summary of wafer-location dependant quality but it applies more toward gpu and historical cpu yields than it does today. For the last 18 months or so the best overclocking amd cpus have been the 1.5v 1700+ tbred b and then the 1.65v 2500+ barton and now the 1.45v 2500+ barton. Of course the higher rated cpus (at same vcores) are as good, but I contend they are not better. This is due to the pass/fail method of chip testing used at the plant and the narrowing of the gap between outer-wafer and inner-wafer chips and the fact that the 2500+ lot sizes are so large compared to higher models which means a lot of them are center-wafer themselves. I'm just saying that your point is right on in theory but not economically material in practice, in regard to current cpus. We still see it in gpus though. (But if I could buy a 2600+ for the same price as a 2500+, I would take the 2600+ for exactly the reasons you outlined.)

"So, johnoh (respect flyes out to U for Ur experience), what do you advise me to do ? Should I do it the cheaper way and if it's crappy drop to DDR400 or get some bada$$ DDR400 to start with and hope they'll go up the mhz scale?"

I'm not sure there is a "right" answer to your question but here's my opinion. When it comes to computer parts brand name differentiation is rarely worth the premium you pay for it. Almost every computer component made works perfectly. Yes with brand names you may get better support or a sense of comfort, but if you go out and buy brand x memory and brand x mobo etc (Psus are a different animal) they will all work perfectly together. As will maxtor or seagate or samsung or hitachi in hard drives, or amd and intel in cpus, or ati and nvidia in vid cards. So to pay extra for a certain brand name of memory is irrational. Further, memory timings are overrated just as cooling is overrated - perhaps there are so many oc'ing enthusiasts in the world that their zeal in these areas have inspired an overestimation of their economic return among the public in general, which is why joe user ends up throwing money at them. Yes there is a difference, just as above I said I'd take a $80 2600+ over a $80 2500+, but the point is the difference in practice is nowhere near the difference in price.

A rational view of computer performance says do not throw money at timings, but do put it toward fsb speed. Do put it toward hard drive speed (like raid0 7200rpm with 8mb cache). Do put it toward pci-express and hyperthreading and broadband and an audio processor and hypertransport and your favorite keyboard and mouse. Those things will make you work or play faster.

Imagine a world where 99.5% of clothing would never wear out, an would look on day 3000 just like it did on day 1. Buying clothes would cease to be about reliability and be all about style. Computer parts have moved at least somewhat in that direction.

I'd say I agree with you on the narrowing gap between chips made in one wafer. But wafer to wafer and batch to batch differences are still there. They have better control now but they cannot totally eliminate it. That's why AMD offers mobile Barton 2400+, 2500+ and 2600+ (by the way, I remember I read somewhere there was mobilt Barton XP3000+ but forgot where). Of course it's also possible AMD makes thinner film specifically for their mobile Bartons. The mobile 2500+ is popular because it's the only one available a while back. Now you see people play with all three and generally speaking, 2400+ is inferior to both 2500+ and 2600+ in overclocking Vcore-wise. My own experience with 2600+ makes me believe 2600+ is a little better than 2500+ in Vcore. On same board with same components, both chips can be oced to 2.62 GHz (12x218) which is not the limit yet, but for them to be stable, 2500+ needs a little more juice at 1.85V vs 1.8V for 2600+. But I also must admit I didn't put 1.85V on 2600+ and see how much higher I could go. I think AMD draws lines somewhere for differentiating chips into these three speed groups. Will we see huge difference in OC potential between 2500+ and 2600+? Maybe not. Certain chips may be designated 2500+ or 2600+. It's the chips at both ends of the spectrum that shows the biggest difference, but number of these chips are small given the distribution. The difference between the means of these two groups are not that obvious anyway, but it's there. Anyway, there is a $9 difference in price between mobile 2500+ and 2600+ at NewEgg.com, I'd certainly get the 2600+.
As for RAMs, it's true at ultra high FSB, timings do not make huge difference as extra MHz's do. But if we are talking about nForce2 boards such as NF7 series, the timings count because these boards tops around 220 FSB. 2-2-2-5 definitely is better than 2.5-3-3-7 in this frequency range. So you want DDR400 or DDR433 for these boards with tight timings. And I believe nowadays almost all DDR400 can go up to 210~215 with very tight timings. Of course top of the line DDR400 from name brand such as Corsair and OCZ can go even higher. But since DDR400 is almost universally good, brand name only may not justify the extra cost.

Thanks for the wise advice and the detailed explanations.

This somewhat reassures me.

"That's why AMD offers mobile Barton 2400+, 2500+ and 2600+"

AMD offers these models because the public wants variety. If their manufacturing techniques produced nothing but identical chips, they would still offer this variety. A chip that does not pass the 2600+ test is tossed out, it is not retested as a 2400+ and if it passes is labelled such. That near-random labelling is a big reason why the chips end up so close together, and why overclocking enthusiasts tend to buy the lowest rated chip at a given stock vcore, assuming an unlocked multi.

As I look through overclocking databases I see no correlation between ghz and vcore and model number for cpus of the same stock vcore and manufacturing week. This has been true since the tbred b. It was not true before that. If post #8 was talking about the palomino or tbird I'd agree 100%, but things have changed now and the chips (as evidenced by results) are now so close as to be called equal.

Of course at ultra high fsb and ultra low ghz you get wasted fsb. A previous post (if this site had any sort of search engine I'd point to the post) showed that a multiplier of 3 is when this starts to happen. Up until then (like multiplier of 5 or higher) you cannot get enough cpu utlilization to get maximium throughput since the fsb saturates while the cpu is sporadically idle. Or both are sporadically idle depending on the rest of the system. I use multiplier here in the non-traditional way, as the ghz divided by fsb rate, not ghz divided by system clock.

I appear to be down on tight vs loose timings and the difference between a 2500 and 2600 not because there is not some slight difference between them, but because they are overrated and therefore overpaid for. Which is to say that if a 2500 costs $80 and a 2600 costs $89, I believe it is better to buy the 2500 and then spend $9 on pizza then to buy the 2600, as the 2500 and pizza is likely to make a greater net positive impact on your life than the 2600.

"A chip that does not pass the 2600+ test is tossed out, it is not retested as a 2400+..."
I disagress. Either AMD specifically deposit film for their different speed rated chips (in which case 2600+ should be better in Vcore headroom because it's disigned so) then discard chips not passing the speed test, or AMD has these chips coming out as a mixed bag then tests and divided them into different speed groups, in this case it's hard to imagine a chip does not pass 2600+ will get tossed out without being tested for 2500+ or 2400+, that'll be huge waste after thausands of manufacturing steps.
My point was that 2600+ should collectively better in Vcore headroom than 2400+ or 2500+. How much better? It's up for debate. Is extra $9 worth it for getting a 2600+ instead of a 2500+? It's up to you and your luck. For me, absolutely worth it. At the very least I have the bragging right on 0.5V lower Vcore to my poor friend to whom I sold the 2500+. If it makes one feel even better, save $13 more for an extra large pizza and get the mobile 2400+.
If AMD thinks people want varieties and packs same chips in different lables, that's an insult on people's intelligence, especially when AMD doesn't lock multis on these mobile Bartons.

Oh and one more:

If in the end I get a locked Barton 2500+, is there any chance of getting it at 233x11 (to have it in sync with DDR466) on an Abit NF7-S 2.0 (or an Abit AN7) with an Extreme Volcano 12 ?

"If AMD thinks people want varieties and packs same chips in different lables, that's an insult on people's intelligence"

Totally. As is charging 2x for beer that is advertised by c-cupped twins. You are a minority - thinking and analyzing and being offended when your intelligence is insulted, but society's economic engine is driven by the majority - who follow their feelings and herd behavior and ask not to think or act responsibly but to have others (advertisers jump on this opportunity) to tell them what to do.

Here are two advertisements:

"amd 2500+ cpu"

"amd 2600+ cpu"

Amd makes their chips the same way then sorts them by wafer location and tests them according to that week's orders. I agree the 2600+ 1.45W comes from the middle, after the 1.35W cherries are picked. Its not about insulting people, its about cost.

If all their chips were identical they would still offer us a variety of labels to choose from.

"it's hard to imagine a chip does not pass 2600+ will get tossed out without being tested for 2500+ or 2400+"

But that is what happens. A chip that does not pass the 2600 test but is retested and passes as a 2400 is likely to be returned under warranty as a failed 2400, something amd wants to avoid. Since 30% of chips fail their test, it makes no sense to restest all of those and find the few that will pass the test of the model rating two notches below that, especially since they may then fail in the field anyway. Better to toss out 30% than go to the time and expense of tossing out 29%.

"My point was that 2600+ should collectively better in Vcore headroom than 2400+ or 2500+."

Well said. But since almost all consumers are uninformed on the points we are discussing and actually believe that a 2600 is materially different from a 2500, amd is able to garner more dollars from 2600 purchasers than 2500 purchasers then is actually due to the ability of the chip. $8 of those $9 are due to the (advertising) label, and maybe $1 is due to the chip's ability.

So I guess I should not have said $9 worth of pizza, I should have said $8.

I believe in promoting the path of economic efficiency which means paying for value but not advertising, and those opn's with the 2600 on them are charging $8 for advertising while delivering $1 of value.

When I win the lottery I'll change my tune.

Beating the system means enjoying the twins while drinking generic beer. And buying the lowest rated cpu with the lowest rated vcore within a given manufacturing process. That's why excaliberpc's 1.35w mobile bartons sold out so quickly the other day.