Here you go max spi winXP
Printable View
Here you go max spi winXP
Best Sandra thus far winXP. This is 274 X 11 2225. I could do 280 at 3.6v on a K8Npro
i thought the rules were you need to run cpu at 2400.
lol :)
sorry didn't see the 2400 limit, lol. Either way you can see that high fsb is very obtainable with bh-5 and regardless of what we run, I'm using an FX and dual channel I'll be ahead in Sandra and spi, but I think I got 30s at 2440
Of course, dual-channel, forgot we are on two different sockets here...can you do a single channel run pkrew ?
I could, but we'd still be comparing apples to oranges. I have some pc3700 eb 2 x 512, but it maxes at 260 3225. Slower by over 200 points in 2k1 against 2 256 of bh-5 at 260 2225
would one stick of 512 and one 256 matter? or would this be unfair? I think the cas2 factor would really give pkrew an advantage since I would run cas 2.5 or closer to cas 3 with the eb. do you want to continue?
trans am that's exactly the point. BH-5 does Cas2, EB doesn't.
I thought you were arguing how much better EB was since it scaled '1:1' so high ?
So show us.
ok ok, correct me if I read your post wrong, So you're saying there's no performance increase by using ram in sync with htt(1:1)? you want the highest htt possible and just want the tightest timings possible and use a divider. I am having a hard time understanding. So you could use pc2700 2-2-2-8 with a divider and you'd get better performance? This is good because I am actually learning something here. Sorry about turning a thread about falling cpu prices into a ram battle. This is like Trekkies meets the movie You Got Served. lol!Quote:
Originally posted by LilGator
trans am that's exactly the point. BH-5 does Cas2, EB doesn't.
I thought you were arguing how much better EB was since it scaled '1:1' so high ?
So show us.
What are you talking about ? I can't tell if you are serious, or sarcastic :D
PC2700 of course wouldn't get better performance ... it's still running at 166MHz.
You said BH-5 doesn't have the headroom to go 1:1. It doesn't need to. All you have to do is run whatever HTT you need to get your max CPU clock, then use a divider. Memory is based on the CPU speed. Not the HTT.
Basically what I'm saying is you can push EB to 270-280, 3-3-2-7 or whatever, but BH-5 with a divider at 260 2-2-2-5 will still beat it.
BH-5 to the limit is faster than EB to the limit.
Occasionally you will see some lucky EB that does 300+, but then there is some lucky (and heavily volt-aided) BH-5 that does 280+.
I am serious. :) is this right?Quote:
Originally posted by LilGator
What are you talking about ? I can't tell if you are serious, or sarcastic :D
PC2700 of course wouldn't get better performance ... it's still running at 166MHz.
You said BH-5 doesn't have the headroom to go 1:1. It doesn't need to. All you have to do is run whatever HTT you need to get your max CPU clock, then use a divider. Memory is based on the CPU speed. Not the HTT.
Basically what I'm saying is you can push EB to 270-280, 3-3-2-7 or whatever, but BH-5 with a divider at 260 2-2-2-5 will still beat it.
BH-5 to the limit is faster than EB to the limit.
Occasionally you will see some lucky EB that does 300+, but then there is some lucky (and heavily volt-aided) BH-5 that does 280+.
example eb3700 at ddr533 3-2-3-10 is slower than bh5 at ddr400 2-2-2-5. So what's the point of buying expensive pc4200 if you could buy cheaper pc3200 ram and use a divider. IF this is true then I am under some misconceptions and sincerely appologize to everyone. There's no point to buy ram rated over ddr400 for amd64 because of the integrated mem controller? just buy cas2 and that's it?
When pc4x00 first came out people went through this. With the adata that would do 300+ fsb 1:1 but was performing significantly worse on intel setups with bh-5 2-2-2-x 5:4. Eb isn't the same thing because it can run much tighter timings (3-2-2 or 3-2-3 v 3-4-4-7). Being that cas makes such a small impact on performance many people are opting for eb now since it has the ability to go much higher than bh-5 and generally without vmods eb can often outperform bh-5. However, with volt mods the speed advantage is nulliffied and bh-5 can attain high enough clocks to retake the performance edge.Quote:
Originally posted by trans am
So what's the point of buying expensive pc4200 if you could buy cheaper pc3200 ram and use a divider.
Either way you go, eb and bh-5 are both damn good chips, and since bh-5 is no longer readily available, and with the good pricing on eb, there isn't realy any need to argue over which is better, especially in a thread about dropping amd prices...which by the way kicks ass, and I will be ordering a cpu as soon as the dfi s754 mobo is released (still haven't decided between nc or cw, being that I can't find any head to head comparisons of max oc v max oc).
Look at what the guys on top of the ORB are running. It's not PC4200.
gouda96
Believe me, you want a ClawHammer. Newcastles are not good performers.
BTW: To anyone who may care, I just picked up an Asus A8V and FX-53 for a price I couldn't refuse. This weekend I'll be testing the PC Ice phasechange cooler on the FX-53.
Also, THE S939 FX-53 IS A CLAWHAMMER THAT HAS BEEN RENAMED A SLEDGEHAMMER!!! Further proof that FX is the only way to go if you are thinking S939.
See, Sledgehammers require reg memory (only Opterons and S940 FX's are truly Sledgehammers) and they are all made with a dark gray ceramic package. Clawhammers and Newcastles are made with organic green or brown packages which is the same as the S939 FX. Also, the die of the FX-939 is identical in size to the S754 Clawhammers I've seen naked. For these reasons I believe that the S754 3700+ is an FX-53 S939 with a different package (less pins) and one of it's controllers disabled.
Check this out for more info:
http://www.ocforums.com/showthread.p...95#post2748995
Never even asked the guys what they are running there BH series ram at before telling to replace...Quote:
Originally posted by trans am
I am serious. :) is this right?
Besides... as you can see you're telling a guy who constantly says "Go to the orb" as the end all be all test. Go take a peek and report back what you see.
The main thing here with the way the fsb is integrated.... you are pretty much always running a divider off the cpu speed for the mem anyway... always. Second thing is that Ras to Cas of 2 is worth big MHz. For instance, BH5 could probably top your mem bandwidth right now with maybe 225 to your 246 if the EB isnt tightened to 2. See though, I'm not telling you to get new ram, but you should test a good deal at ras of cas of 2 if you can and give up quite a few MHz on the fsb for it.
Who is on top??? DDR500 being shown
http://images.anandtech.com/graphs/b...10646/2747.png
link to full charts http://www.anandtech.com/memory/showdoc.aspx?i=2130&p=7
granted its only a small sample of ram thats avil.
I'm glad this happened because I learned a lot. I am going to save this page because I think it will help a lot of others new to AMD64 and end the misconceptions. Sorry again, I stand corrected. I'm not used to the level of maturity on this forum and appreciate the politeness and the time you took explaining things in detail instead of simply getting flamed and told to Fvck off. warm regards.
what chip is the EB stuff using??
Micron ic's except for the 1gig sticks.
If you would have bothered to look at the test setup for the stupid barchart you posted you would have noticed they were using a P4. This thread is about A64s, not the same thing at all. P4s have different requirements for RAM. What works well on a P4 does not necessarily work at all on the A64.Quote:
Originally posted by iboomalot
what chip is the EB stuff using??
Pretty chart though! Maybe less booming/more reading before pulling that trigger tiger. :rolleyes:
why don't you get off your a$$ and show some proof to dispell what I've shown??
You have nothing to offer except your opinion.
show us how EB isn't the best or top of the list???? proof would be nice.
Quote:
Originally posted by trans am
"I am serious. is this right?
example eb3700 at ddr533 3-2-3-10 is slower than bh5 at ddr400 2-2-2-5. So what's the point of buying expensive pc4200 if you could buy cheaper pc3200 ram and use a divider. IF this is true then I am under some misconceptions and sincerely appologize to everyone. There's no point to buy ram rated over ddr400 for amd64 because of the integrated mem controller? just buy cas2 and that's it?"
You've got it now :) 1:1 with slack timings is slower than 5:4 with tight timings. It's even the same on the P4 (though the P4 does take a hit with dividers), but A64 of course has no 1:1 or 5:4 as I said earlier, so there is no performance hit running a divider and there is no magical thing that happens with 1:1 like everyone seems to think. Now in your comparison EB with 533 def. would be faster than BH-5 @ 200MHz, it's too big of a MHz gap for the timings to make up for it. EB is the best memory available with 3.0v or less, but 3.2v+ BH-5 is much better. Also, it really gives you unlimited multipliers if you think about it.Quote:
Originally posted by trans am
I'm glad this happened because I learned a lot. I am going to save this page because I think it will help a lot of others new to AMD64 and end the misconceptions. Sorry again, I stand corrected. I'm not used to the level of maturity on this forum and appreciate the politeness and the time you took explaining things in detail instead of simply getting flamed and told to Fvck off. warm regards.
Lets say you have an 8x multi chip, and you want a '10x multi'.
All you have to do is set your memory mode (divider) to 166, and move the HTT up to 240. Bam, you have your memory at 200MHz, and your CPU at 2000MHz (Just like 10x200) with only an 8x multi :)
Lets say you're like me, and you want the 3700+ 12x multi ;)
Then you set your memory speed (divider) to 133, and move the HTT up to 300. And again your have your memory at 200MHz, and your CPU is running 2400MHz (Just like 12x200) with only the 8x multi. Obviously if you move the HTT higher you'll have to drop the LDT to keep your HT within its spec (or as high as you know it to go)
Virtual 12x/11x/10x/9x multies are all possible with an 8x multi chip using that method.
[Virtual 9x Multiplier] - [9x200]
8x Multi Chip: Set Memory to 183 and raise HTT to 219-220MHz
[Virtual 10x Multiplier] - [10x200]
9x Multi Chip: Set Memory to 183 and raise HTT to 219-220MHz
8x Multi Chip: Set Memory to 166 and raise HTT to 240-241MHz
[Virtual 11x Multiplier] - [11x200]
10x Multi Chip: Set Memory to 183 and raise HTT to 219-220MHz
9x Multi Chip: Set Memory to 166 and raise HTT to 240-241MHz
8x Multi Chip: Set Memory to 150 and raise HTT to 266-267MHz
[Virtual 12x Multiplier] - [12x200]
11x Multi Chip: Set Memory to 183 and raise HTT to 219-220MHz
10x Multi Chip: Set Memory to 166 and raise HTT to 240-241MHz
9x Multi Chip: Set Memory to 150 and raise HTT to 266-267MHz
8x Multi Chip: Set Memory to 133, and raise HTT to 300MHz
Hope this helps, and thanks for listening/thinking outside of what you knew to be true before :)
some A64 comments since you can't accept the above testing
OCZ is breaking new ground with their latest Enhanced Bandwidth series, and 3700EB extends the EB performance envelope. While rated at DDR466, we were able to reach a stable DDR524 on our Intel test bed and an even more remarkable DDR550 on our AMD nForce3-250 test platform.
higher it got. the wider its bandwidth and on a A64 it clocked even more than the intel setup.
I wonder how well 3700EB would do with some extra voltage ;)
Even EB at DDR550 (probably 3-2-3) is no match for BH-5 (2-2-2) at even DDR500-520.Quote:
Originally posted by iboomalot
some A64 comments since you can't accept the above testing
OCZ is breaking new ground with their latest Enhanced Bandwidth series, and 3700EB extends the EB performance envelope. While rated at DDR466, we were able to reach a stable DDR524 on our Intel test bed and an even more remarkable DDR550 on our AMD nForce3-250 test platform.
higher it got. the wider its bandwidth and on a A64 it clocked even more than the intel setup.
I wonder how well 3700EB would do with some extra voltage ;)
hummmm look at the followingQuote:
Originally posted by LilGator
Even EB at DDR550 (probably 3-2-3) is no match for BH-5 (2-2-2) at even DDR500-520.
http://images.anandtech.com/graphs/c...50643/2388.png
that shows stable timings at 400 mhz notice the EB is below the 2-2-2-5 muskin and corsair memorys.
http://images.anandtech.com/graphs/c...50643/2392.png
this chart at 433 mhz the EB has caught up and passed the muskin by a small amount and still below the Corsair.
http://images.anandtech.com/graphs/c...50643/2402.png
by 466mhz EB is passing the Corsair
http://images.anandtech.com/graphs/c...50643/2399.png
by 500 mhz EB is on top Period.
Since the EB can goto 550 mhz on the A64 system Iam sure it stays on top of the heap due to its improved scaling as speed increases. This happens even at the looser timings.
As I said I wonder how well EB would do with improved voltage to help keep its timings low.
My old Geil is rated at 2-3-2-6 and I run it at 2-2-2-5 so Iam sure the EB could be tweaked timing wise.
Nothing against BH-5 chips those things whale on high voltages.
I'd like to see some low timing PC4000 EB :D