The Official PC building thread -3rd Edition

my PCs have uber cable management, ask sam

i sell them because i have no need for them.

 

Sam,
The later 65nm chips perform better and have more memory bandwidth than the older 90nm chips had in socket AM2, in spite of having less L2 cache. As far as the "until recently" goes, when I bought the 4800+ for Oxi, it ran better than I expected it to. Eventually I will replace the AM2 motherboard with an AM2+ 780GX, and unleash about a 15-17% improvement in performance. Of course I have to rewire the front panel, but that's fairly easy to do.

As far as genuine affection for a computer goes, The time I've spent getting things straightened out so that mine runs the way it does now turned out to be a lesson in "what can go wrong" for me! Having multiple problems with different hardware turned out to be a challenge, but I hung in there and have been rewarded with a very nice machine, that I've developed a genuine affection for, for what it can do! The only thing I see that runs slower than my E6750 did is Win RAR, and that's really meaningless as once you start archiving something, you can go do something else. It's not like you have to sit there and wait for it to finish! LOL!! The only thing I don't understand is I can run it as high as 3.6GHz with no problem. Run Orthos, OCCT or Linpack continuously without failure for hours on end, even overnight, yet it will randomly reset itself for no apparent reason I can find. Sometimes it will go for a couple of days with no problem and then just suddenly reset itself. I'm completely baffled as to why. Voltages and Heat are non-issues! There's no readily apparent cause, and it doesn't have to be doing the same thing every time it happens. It seems to do the same thing whether it's running at 3.3GHz or 3.6GHz. It's rock solid at 3.2GHz, so that's where I'm gonna leave it. By letting the EPP control the memory, I get better timings at lower voltages. From what I can gather with the limited time I've spent fooling around with a Phenom II, on the same model motherboard, I should be looking forward to a machine that is incredibly smooth to use, only a lot faster than what I have now! I'm looking forward to it!

Russ

 

I thought i'd just interject and post this

http://answers.yahoo.com/question/i....1byrOhr7hR.;_ylv=3?qid=20090526072123AAolDO9

Just thought it was rather funny, don't think she knows the definition of the word "virtual".

 

the generation I'm referring to is the one between CPUs like the 4800+ and 5200+ and the Phenom IIs. i.e. the 6000+ and original Phenom generation. The original AM2 chips went up against E6 series CPUs, which was a semi-reasonable fight. The latter stuff had the 45nm Core 2s to live up to though, so was flattened. The Phenom II is the first new promising tech from AMD. I don't really consider CPUs like the X2 7750 and 7850 in this, they're worthy contenders because they're crazy cheap, but they have to be, they aren't worth much more.

 

As sad as it is to say you are right. The 7750BE is really just a way for AMD to dump loads of old Phenom chips for cheap. It does make it a great deal though and IMO it is competitive as an entry level product. However cheap they may be. They aren't near as fast as the original Core 2s even, but they do offer a reasonably fast and decent OCing dual core for those on a budget.

Interesting note. At 3.2GHz my 7750 BE has a SuperPi of 25 seconds while my 940 BE at stock gets 20 seconds. I thought it was multi-threaded...

The first Phenoms had OC powerhouses like the E6750 to deal with and they really weren't very good at all. Mediocre performance, buggy, and practically zero OC capability. The Phenom IIs are good because they offer usable performance for a reasonable price and can actually OC to a decent level of power. Unlike the Phenom I they are viable for higher-end machines and can actually push fast graphics cards effectively.

We all know AMDs are not up to snuff with Intels yet but they are quickly closing the gap. The Phenom II is a major leap forward for their technology and it's not long before we'll be seeing them on top again. They've been trading places for years remember

 

Estuansis,
My 7750BE at 3.2GHz does SuperPi in 24.512 seconds, and I find it to be a very competent chip. About the only thing I do use that's slow is Win RAR, and I can do other things while it's zipping things up, so the problem becomes minimal! I rarely zip up anything real large to begin with! The E6750, even at a higher clock speed, can't match the 7750BE at encoding with DVDRB/CCE though. My personal standard, "Lost in Space" encodes an incredible 11 minutes faster with the 7750BE! I actually ran it twice because I touught I had made a mistake somewhere, or had accidentally turned 2 pass to one pass! LOL!! The 7750BE is also a Barcelona Core, not a Brisbane. Unlike the original Phenoms, there's no fix needed for the TLB thing, as the B3 stepping eliminates that without the speed penalty of the original "work around" AMD used, that severely limited the chips cache performance.

AMD is scheduled to bring out a bunch of new chips between now and the end of 09. I get the distinct feeling that all Future x2 AMDs will be Barcelona or Deneb cores, as the Brisbane on 65nm and the few remaining Windsor 90nm cores can't keep up with the 7750/7850 Barcelonas, in spite of their larger L2 cache.

So far the real sleeper of the Denebs is the Phenom II X3 720 2.8GHz chip. It's very quickly become AMD's best triple core by a wide margin. Hey if the price comes down a bit or I catch a sale, I may even buy one and sit back and wait for the AM2/AM3 Phenom IIs to come down in price, before I buy the Quad!

Best Regards,
Russ

 

When you talk about your superPi times how many digits are you calculating?

TIA

Red Maw

 

LOL!!
I'm sorry. Oman7 and I have discussed this before so I didn't think of mentioning, it's for 1M. I know it sounds high compared to an Intel, but it caused quite a sensation on another Forum when I mentioned a Dual Core AMD that would do under 25 sec for 1M. I just wish that Xtremesystems would get off their dead asses and fix the validation so that it works, since they own SuperPi now! It hasn't worked for months! You enter the information, double and triple check it and it returns an Invalid Checksum every time! Sorry for the mental lapse on my part! LOL!!

Russ

 

Thanks Russ. Should have guessed it was 1M since that seems to be the standard.

Know that I think about it though 25s is a lot more then I would have expected from a 3.2ghz dual core. I knew there was a gap between intel and AMD but never knew it was that large. Going to have to rethink the AMD build now lol.

Side note: right before posting vista decided to BSOD (again) and was very surprised to see FF saved my post. Also managed to finish superPi (1M) at 19.250s before it crashed

 

On the 'bad' PC in my room and wow, just thought i'd try it out for the hell of it to see how poor my E6600 (default 2.40GHz) was, but it managed 21 seconds for 1M... I feel sorry for AMD

 

Yeah but AMDs have always sucked at SuperPi. So 25 seconds is actually quite solid for a Phenom I Dual Core.

 

Estuansis: I wouldn't go as far as to say the 7750s aren't as fast as the original Core 2s, as they would beat everything up to the E6600, and the 7850 would probably beat the E6750 as well. There are loads of Wolfdale chips that would serve the 7750 and 7850, but there are also ots of considerably more expensive Wolfdale chips too. Stock for Stock, the bang for buck of the 7750 and 7850 is better than the E5200 and E5300. Of course, add overclocking to the equation and the AMDs get obliterated, even CPUs like the E2140 and E2160 flatten the AMDs, let alone Conroes or Wolfdales.
SuperPi is single threaded, which is why Phenoms score poorly at it, as AMDs were never very good at SuperPi, not that it really matters, SuperPi is about as synthetic as you get.

 

Shokz,
Than hang onto your hat then, because my E6750 would get in the 13.80 second range at 3.55GHz. Before I switched to 1066 memory and tightened up the sub-timings, the best I could do was 14.236 or so! The 13's are right at E8400 country!

Russ

 

Heh, I think my Q9550 managed about 11s at 3.9Ghz.

 

I certainly hope so with 4 cores! LOL!! You probably remember when I started shortening the memory Sub-Timings after I got the 2GB of Dominator 1066 memory for my P35-DS3R! It's very tedious work, but I haven't heard of a lot of people breaking into the 13's in SuperPi with E6750s. Not below 3.6 or 3.7GHz, anyway! I think had I kept it considering how well the replacement E6750 overclocked, I would guess that I could have gotten into the high 12's with it! All those saved clock cycles do add up!

Russ

 

Remember Russ, SuperPi is single threaded... All that's getting me that is high clock speed and more cache.

 

I know it's single threaded! Try tightening up all the lesser memory timings and it does make a difference, at least with the smaller L2 cache of the E6750 it does! That and I need to go back and correct that 14.725 to 14.236! LOL!! Damn Dyslexic fingers! ROFL!!

Russ

 

You know guys...I really need to make my next build an intel! Im sick of people bashing Amd. Its the smoothest platform i've ever PERSONALLY got to play with. Granted, the intel systems I did get to play with could have been other components causing the bottleneck...because they were intels has always stood out to me. I am SO NOT bias. Im simply noting what i've witnessed.
Heres a good question. Whats better, a 2.0Ghz pentium 4, or an Athlon 1.4Ghz thunderbird. Cause guess what. The 1.4Amd machine, SMOKES the pentium machine! Thats one of the reasons ive always looked at the name "INTEL" with a quizzical look. But im also aware that a component CAN vary greatly between identical components. E.g. one processor may OC to 4.0Ghz where another can barely hit 3.7Ghz! Perhaps the 2.0 pentium was built on a monday LOL! Or, there is simply another component in dire need of attention. Thats currently my task now. And im looking at its PSU VERY VERY closely!!!

 

On this page we were bashing the earlier AMDs, the Phenom IIs aren't bad.
Also, 2Ghz P4 vs 1.4Ghz Thunderbird, the Inte was 10 times over the better choice, the Thunderbirds were hot, unreliable, and had all sorts of problems. The Athlon XP that succeeded the was amazing, and handily beat every Intel that rivalled it. (Palomino, Thoroughbred, Barton, vs Willamette, Northwood)
The Athlon64 came out and gave Intel a further serving, especially against the Prescotts which were worse than Northwoods. The Athlon X2 (Manchester/Toledo, Windsor, Brisbane) destroyed the Pentium D (Smithfield, Presler).
The Core 2 Duo's arrival though firmly turned the tables. The Conroes and Allendales beat the AMDs hands down in every department, and it proceeded from there. Intel have been the dominant performer since July 2006. AMD are close to clawing it back with the Phenom II, but they need to squeeze more out of the architecture and copy these gains to the dual core market for that to work. Right now, the only Dual cores they have to offer are the older Phenoms that couldn't.

In the Intel days you'd struggle to get a 3Ghz chip to 3.5. 3.4 would be doable with lots of tweaking. AMD's Athlon64 architecture was seeing gains like 2.2->2.7. The Core 2 Duo was what really set the benchmark though. A 1.6Ghz E2140 would make 3.2Ghz without too much fuss (100%). My 1.8Ghz E4300 ran at 3.2Ghz fine for over a year (78%)
Even with Quads, you can usually squeeze up to 40% out of them with a good board. Right now, the best AMD seem to be able to do is about 25. For the Quad cores that's not too bad, but for dual cores, it's a bit on the low side.