• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

2-2-2 vs 2-3-2 any big difference?

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Krowa 02

Member
Joined
Oct 3, 2002
Location
Long Island New York
Im almost finished choosing the new parts for my new p4 rig, and I have come accross a problem when it comes down to choosing my ram.

I am looking at 2 mushkin models, the cheapest being the dual 512 pack of 3200, with 2-3-2 timings.

Is it worth the extra 30 bucks to get the 2-2-2 stuff? Or would it be possible and safe to up the volts on the 2-3-2 and run it at 2-3-2. Better yet, is it even noticable? I dont want to buy high speed ram, id rather use the 5:4 ratio and run the ram at 200fsb, im aiming at 3ghz, nothing more.

I dont want to spend a whole lot on this rig, as I barely have enough to order the parts I have in my list, I need to sell my current rig to cover myself for the rest of the month ;). But if there are any better choices, I can squeeze a few extra pennies.
 
The main issue of buying Mushkin Black Level 2 PC3500 over Black Level 1 is the deal of what memory chip is used. The Black Level 1 uses Winbond CH-5 chips while the Black Level 2 uses Winbond BH-5 chips.

BH-5 chips are amazing overclockers if you give them good voltage. (although all chips are good overclockers if you give them the volts, the BH-5's seem to be more capable of holding 2-2-2 timings at higher FSB's)

If you don't mind having the best overclocking memory, then using Winbond CH-5 chips should be fine. This goes doubly so if your uneasy doing voltage mods to your motherboard in order to achieve VDD voltages of 3-3.2Volts.

If you want Winbond BH-5 chips, there are also a few other opportunities out there (although they are drying up fast) that are cheaper then Mushkin's Black Level 2 PC3200/PC3500.
Check Prices Here - http://forum.oc-forums.com/vb/showthread.php?s=&threadid=237964
 
http://www.techwarelabs.com/reviews/memory/memory_timings/index_3.shtml

Read that. I did my own tests, and my results were similar. Going from 2-2-2-5 to 2.5-3-3-7 I lost a whopping 0.8% in performance. With that said, save your money :)

Find chips that can get higher MHz, that'll make way more difference than timings as long as they're not horrible. It took till 2.5-4-4-9 to see any real decrease - about 4%. As long as you can do 2.5-3-3-7 or better you'll be fine.

--Illah
 
not really a whole lot of difference in the 2 in real world apps...I would pay money to somethat that could see a difference while running windows common programs ;)
 
So getting faster ram would be a smarter thing to do?

(sorry im posting so late, I was sleeping cause I was to nervous to watch the yankee game, but my friend woke me up with the good news :) )

But back on topic, im only looking to run at 250fsb, so wouldnt it be just better for me to get pc3200? Or is there a different divider that will let me have higher mem speeds?
 
If you are looking at keeping your computer at 250/200 (FSB/MEM), and not any higher, then a good bet would be to buy a quality memory using CH-5 chips and run at 2-3-2. Buffalo PC3200 would be the cheapest, with Corsair XMS, Kingston HyperX, and Mushkin Level 1 the top quality assurance (at high $) picks.

If you want to eliminate memory as a potential bottleneck (meaning if you are going to push much further then 250/200 and want 2-2-2 timings), then the best thing within reason (ie low $) would be to purchase a memory with BH-5 chips.

If you want a possible adventure, then you could also try GeIL Golden Dragon. It has been capable of running at decent FSB overclock, albeit at looser 2-3-3 timings.
 
Illah said:
http://www.techwarelabs.com/reviews/memory/memory_timings/index_3.shtml

Read that. I did my own tests, and my results were similar. Going from 2-2-2-5 to 2.5-3-3-7 I lost a whopping 0.8% in performance. With that said, save your money :)

Find chips that can get higher MHz, that'll make way more difference than timings as long as they're not horrible. It took till 2.5-4-4-9 to see any real decrease - about 4%. As long as you can do 2.5-3-3-7 or better you'll be fine.

--Illah

are those bandwith tests? Of course changing timings won't make a large impact on bandwith, however they will make a huge difference on 3d performance. If you take crap pc4x00 and run 300fsb 1:1 3-4-4-8, or any other crap timings, and you then switch to bh-5 5:4 2-2-2-5 you will gain anywhere between 500-2k in 3dmark2k1. You will loose a good amount of bandwith, but your scores will increase in every meaningful benchmark. Your review was well done, you should go back and do 3d benches like 3dmark01/03 and ut2k3...you basically wasted time saying the obvious...although it is a very good article on the affect of timings on bandwith, but not timings on system performance.

The other problem with ch-5 besides the slightly more slack timings is the fact that they have a whole bunch of compatability issues. If they work for you they seem to be able to hole 2-3-2-5 timings pretty high, but bh-5 will get higher clocks with 2-2-2-5 than any other memory around. Any memory with bh-5 is the best around, until bh-4 comes out.
 
I'll have to give that a try, I'm not a 3D Mark junkie so I only really use Sandra. What about stuff like video encoding, rendering, and SETI - wouldn't that be more bandwith intensive?

Anyways, I'll try testing 2-2-2-5 and see what happens to the 3dmarks. I'm on AMD though so I'm not sure if the benefits still apply vs. Intel.

--Illah
 
...

I am not sure if it makes as much of a difference on amd, but I think it does, being that my 3dmark score shot up about 1k when I switched to bh-5. Would be pretty nice for you to add 3d benches to that, maybe along with s-sandra, seti, encoding, etc...If you do that, you should put it in a brand new thread, because that would realy have to be stickied.
 
That's not my review if that's what you're thinking - I just linked to it :)

But yeah, I tried 2-2-2-5 at my current settings and say a 400 point 3dmark increase, about 2.4%. Considering bandwith was only improved by ~1% that does say that bandwith isn't everything. Still, it's not stable 2-2-2-5 at current speed, might be a vdd thing.

--Illah
 
Back