Quote:
Originally Posted by high5
execatly what I though. I was just also wondering what score would reach my BE-5 that are stable @ 285 htt. should be more fair since uccc also reaches those speeds
Printable View
Quote:
Originally Posted by high5
execatly what I though. I was just also wondering what score would reach my BE-5 that are stable @ 285 htt. should be more fair since uccc also reaches those speeds
That's why it's a little bit of a different approach here. Think about it like an upgrade from 1GB to 2GB. You know your CPU's max and now you're looking at different memory types and try to decide what is better in the real worrld. UCCC will most likely go further on pure clock (but looser timmings) that CE-5 (low latency). That's why I was testing both at, like high5 said, "it's best" given the CPU speed stays the same. If I tested both at same frequency then Infineon would look much better in most benchmarks (maybe besides games). It would be same as locking CPU at 3GHz and trying to see what is better - 273MHz 7-4-4-3.0 or 250Mhz 5-2-3-3.0. But then I could have tried 7-3-4-3.0 (UCCC) and 5-2-3-2.5 (CE-5) as well. And most likely results would be similar, again - except for 3D01 and AM3 which favor low latencies a lot.Quote:
Originally Posted by MonkSP
------
I don't think that testing with dual-core will show much different trends (it's a memory comparison not sc vs dc one and multi-threading will not make memory matter a lot more. But I could have done "gaming part" a little bit better. At first I wanted to show that at everyday settings (1600x1200) video card bottlenecks the system so memory does not seem to play a huge role... But at 1024x768 we might see bigger differences - although I'm not expecting anything larger than 2-3 fps. I'm gonna do it today, probably.
Yep, it would be nice to get a hold on some Micron capable of 291MHz 5-3-3-3.0. I'm pretty sure it would take a lead in benchmarks where memory matters a lot. Unfortunately, I don't expect to be lucky enough to get Rev.F that will do 291MHz. I'm sort of working on getting a set of Micron -5B:F based sticks for my next OCing Report and I we will see what happens. It's not gonna be anything that will do 291Mhz, though...Quote:
Originally Posted by high5
Sorry about that typo, I just fixed it ;)Quote:
Originally Posted by ahmad
Then I could do it for both CE-5 and UCCC as well and have same effect, I would think.Quote:
Why run at Tras of 5 and Trrd of 2 when you can do 1/0 with the CE-5? You get slightly better SPI times with Tras @ 1.
I don't quite see the point of that (and a chance - no $ for FX) as dividers DO NOT affect performance whatsoever. It's not AXP anymore ;)Quote:
Originally Posted by perry_78
i'd be glad to run a few of the non-gpu related benches for you if you'd like. my cpu tops out at just over 3ghz, so i can't match your cpu speed, though i can change the timings to match the UCCC that you tested, just to see the difference between UCCC settings and 5b F settings :DQuote:
Originally Posted by bachus_anonym
As promised, I have just added set of Game Benchmarks at 1024x768. Check them out if interested :)
Also, I have almost figured out Battlefield2 timdedemo benchmark and might have results later. I've been thinking about running XVID video encoding test using VirtualDub and LameMP3 audio encoding with dBpowerAMP. Anything else useful to run?
Some 2D/3D rendering. Sciencemark, Sysmark also would be great to see :)
Great work bachus!
Ah, good old AXP days when an htt over 250 was a wonder even on some (i.e. my) NF7 boards :)
With this amount of ram testing you do, an fx should be leant out to you!
very nice comparison, but keep in mine in general most UCCC can only do 260~280 while the Infineon can only do 252~260 :D UCCC is dirt cheap now (GSKILL HZ $158 @newegg) so I'd say it's a much better deal :D
Hehe, that would be nice but I can live without it. For RAM testing, my Opteron is just enough as I honestly dismiss a claim that FX has a superior memory controller and all memory will clock better with it. For hardcore benching it's a different story, though ;) Anyway, back on topic...Quote:
Originally Posted by perry_78
You're about right about averages, Ben. I would just raise low value for CE-5. If it's not a very weak set, PC4000 rated CE-5 should do on average 260Mhz in 3D. Very good set is 270Mhz and whatever higher is rare but exceptional. That's why I mentioned in my 1st post that sets are "somewhat above average"... Corsair claims to have secured some good Rev.C ICs from Infineon (direct shipments) so we might hopefully see more good clocking CE-5 from them (PC4400PRO).Quote:
Originally Posted by ben805
I was thinking about doing this comparison at 3000MHz instead, but then decided that it's XtremeSystems and cranked up to 3200Mhz to make it more interesting :D
I will add more benches soon, I had a bit busy weekend here :) Thanks for checking this out guys :toast:
I like this comparision, shows just how meaningless latency and bandwidth are for A64's. Buying all that expensive RAM just aint' worth it at all folks as a ~1% performance gain is so low as to be within the margin of error for most tests, your money is better spent on cooling or a better video card or CPU if you want to improve performance.
Last time I checked the url it still said xtremesystemsQuote:
Originally Posted by mesyn191
It seems like once the game or whatever became GPU dependent, the extra speed from the Samsung could feed it slightly better. Very nice comparison, thanks for publishing it!
Thanks for ur work and effort
Very good comparison:toast:
Indeed it does, but if your performance gains are so low you can't prove they're actually there than that isn't xtreme, that just plain stupid.Quote:
Originally Posted by ahmad
Good job.
Here we can see that tight timings rox :)