People with ADATA PC4000, 512 MB sticks, what is your unbuffered bandwidth?

And CAS settings. At the highest fsb you can go, must be two 512 MB sticks. I dunno, maybe not many people have this ram.

Many people dont know were to get these sticks in the USA. I know a store in the UK that sells them but none in the US and most of the members here are American so that might be the problem. I dont know much about these sticks but i hear they are good.

ive got 1 stick off Adata pc3200 see sig but i think the pc4000 is just as good for overclocking i think ive seen ppl in here with the cpu limiting the overclock and not the ram on a P4 system

I'm betting the unbuffered bandwith of 512 MB ADATA sticks are the same as the Samsung PC4000, but the Hynix chips respond to voltage whereas the Samsung don't. Thus you can force 512 MB ADATA to do high overclocks if you got the VDIMM.

So far, I have not seen any difference in unbuffered memory going from 2.5-4-4-5 to 3-4-4-5, Samsung or Hynix chips, so I suspect it's how manufacturers program a chip . What does it matter if ADATA PC4000 at 3-4-4-8 gets the same unbuffered as OCZ PC4200 EL at 2.5-4-4-8? The ADATA sticks are $130 cheaper ($300 versus $430). And the ADATA should do over 280, 1:1 with volts, which is the max the ELs did in the Anandtech review (and you'd have to assume they were supplied cherry-picked sticks).

I got unbuffered data on SS ADATA PC4000, so I'd like to see unbuffered data on DS ADATA, e.g. twin 512 MB sticks. I bet it's the same as the high-priced Samsung stuff.

I am also a little curious to know how the ADATA PC4000 is doing. A couple weeks ago I picked up 2x512 of generic Hynix PC3200 for $234 CAN ($179 US). Tough to beat the price and its doing DDR500 just fine. I get Sandra unbuffered 3100-3200 and Memtest shows 3023. Curiously, 3-4-3-8 is more stable for me than 3-4-4-8.

Clevor: How does your ADATA do?

Haven't tested the limits of my ADATA PC4000 since I don't have a CPU/board that can go high enough. So far 282, 1:1, 3-4-4-5 at 2.9 volts. It can run up to DDR470 or so at 2.5-4-3-5. They are SS sticks so they can go pretty high. Some people have hit 300, 1:1 with 3.2+ volts.

My 4PCA3+ goes to 3.32 VDIMM but the CPU can only go 279 on that board.

You are doing pretty good for PC3200. I have some ADATA PC450 that are good for DDR480, 2.5-3-3-6, at 2.7 volts. Won't do CAS 2 no matter what though.

At 282 with only 2.9V, it looks like yours is a canidate for 300. You just need to find the magic cpu

250 is all I really wanted from my cheapo solution and it is all I get. I don't have error-free memtest above 250. I can actually do CAS2.5 right up to 248 but so close to DDR500/1000fsb I had to make it an even 250 . I know my cpu is good to 260 but I'm happy to stay where I am 24/7 since anything more is inefficient w.r.t. vcore.

Any thoughts on why I would have more stability with tRP=3 than tRP=4? I have done the double bios flash to enable full time PAT but I don't think that has anything to do with it. I'm not really worrying about it, though.

Cheers

I think you are talking about CAS to RAS? I've found most ram, both Hynix and Samsung PC4000, don't like to run 3.0 at all. You will see I run 2.5-4-3-5.

And yes, most of the stuff like OCZ PC3700 Gold or PC3500 stuff or even my ADATA PC450 may do DDR500, but they can't do much higher.

If I were you, I'd swing for some PC4000 as you could easily hit 260, 1:1, and the P4P800 is a good overclocker.

The best bang for the buck now is ADATA PC4000, if you can get it (Hynix chips). They are reasonably priced and consistent in quality.

No, actually it is the RAS Precharge that I find works better for me at 3. I usually give my timings in the same order as given by CPU-Z since many people check their timings with this program.

I hear ya about RAS to CAS delay. I have to have that at 4 for anything 215mhz and higher.

I gave a lot of consideration to ADATA's and OCZ's PC4000 for 1:1 but 1GB for the amount I paid that could get me to 250mhz 1:1 was too good of a deal to pass on. I'll probably look again if/when a majority of PC4000 is running tight timings reliably at DDR500+.

On a side note, I'm glad I went 1:1 with the double bios flash. For me, having the PAT even with loose timings actually feels faster than my pc3500 did in 5:4 with tighter timings.

Cheers

There is an Intel issue at 5:4 with high performance ram at tight timings. I can just feel the board wake up at 1:1 with PC4000 ram. The P4C800-E and 4PCA3+ can prevail through that and put up stellar benchmarks. I don't know what they are doing right.

I just tested a DFI 875P Infinity board last night. First tried XMS3500. I could actually run 280, 5:4, DDR440, 2-3-2-5 Memtest clean at 2.7 volt limit of the board!!! But 3DMark persistently locked up. I couldn't figure it out. If I dropped down to 270 it would pass, but Memtest was clean and the CPU was good to 279-287. I just could not get 3D stable. I was about to give up, put the board back in the box and sell it tommorrow.

Then I decided to try ADATA PC4000. Result? 283 fsb at 1.49 VCORE 3D stable. Stuck in an MO step 2.40C, never characterized before. It errored out at 279, 1.58 VCORE on the 4PCA3+. It is currently running 285 at 1.54 volts! And I'm not through yet; I will test how much higher I can get tonight. I have to run 5:4 on the ram since the sticks won't do 283-285, 1:1 at the max 2.7 VDIMM of the board.

But this DFI board is the best overclocking board I have seen so far, and I almost sold it based on performance with BH-5. I see the same problem with the Abit, Gigabyte, AOpen, and MSI 875P boards (all of which I got rid of).

Fortunately I can toggle between PC4000/BH-5 to find which works best with a board. Both ram work well with the Asus and Epox boards, BTW.

I gave a lot of consideration to ADATA's and OCZ's PC4000 for 1:1 but 1GB for the amount I paid

Click to expand...

The price of 1 GB of PC4000+ ram is ridiculous. Even Buffalo PC4200 will run close to $400. And forget about getting PC4000 that might do over 260, 1:1, because they are binning that and selling it for more as PC4200. Every stage of ram is being binned and sold for more profit. In the old days you might luck out on some sticks, not any more

How do these ADATA 512Mb sticks do with a maximum Vdimm of 2.8V? See my sign, I'm thinking about replacing my Corsairs to 2×512Mb ADATA and I would like to run 280-285 1:1. My CPU can do it, that's for sure.

Anyone has some benchmark results with settings something like that? Currently I am getting about 6000/6000 in Sandra with 220MHz 2-5-3-2.

thanks

It has been clearly established that the 5:4 mode with the timings you are running will outperform 1:1 with the timings the Adata is capable of.

Your real problem is lack of voltage. Do the 1-resistor volt mod to your IC7 and solve the problem, not the symptom. If you are looking to go to 1GB of ram as your primary concern, remember that 512mb sticks don't work nearly so well as 256s do on 865/875 boards, and either the Adata or a 512MB BH5 will likely need the volt mod even worse than your existing 256s do.

I strongly prefer the 1:1 mode over 5:4, but the timings the Adata is capable of just don't allow the potential to be realized.

Thanks for your reply.

I've heard that in these systems memory clock is more important than timings, so is this untrue? Heard about the 512Mb modules issue as well.

If I do the vmod on my IC7, how far would I go? 230-235MHz with the same timings possible? I don't want to go over 2.9-3.0V since I am not intending to risk my Corsairs' life.

now thats really strange because ive ran countless benchmarks using DS and SS dimms. in my opinion, DS always performs better on the 865/875 platform. now on the Asus P4C800 boards, ive found that 4 DS dimms performs best. on the Max3 however, 4 DS dimms dramatically hinders overclockability. atleast in my case it does. i couldnt even get 220fsb 5:4 on the Max3.

performance ratings from best to worst are as follow(excluding Max3):

4 DS Dimms *best*
2 DS Dimms ---------
4 SS Dimms --------
2 SS Dimms *worst*

These sticks are for 1:1 only...as far as I know.

I got the answer to my own question, here it is, with two 512 MB sticks:

270, 1:1, 3-4-4-8=3500-3600 unbuffered
250, 1:1, 3-4-4-8=3300 unbuffered
464, 5:4, 3-4-4-8=2800 unbuffered

These are whopping scores, for ADATA sticks that are actually PC3600. See my new post on this topic regarding 'ADATA PC3600 at PC4368'.

It has been clearly established that the 5:4 mode with the timings you are running will outperform 1:1 with the timings the Adata is capable of.

Click to expand...

I was one of the guys at the forefront with benches to prove this, but I was comparing DS BH-5 to SS ADATA PC4000. Now with DS ADATA stuff versus DS BH-5, I am getting better unbuffered with the slow CAS timings like 3-4-4-8.

Unfortunately, the cost of two 512 MB PC4000 sticks is beyond the budget of most mortals. I would say only 5% of us would spend $400 for such ram. I'm just glad my XMS3500 sticks were leftovers from my 845PE days so they are all DS.

I really need a volt mod on the DFI 875P Infinity board, which only goes to 2.7 VDIMM. If I could get 2.9-3.2, I could probably drive my ADATA PC450 to 292, 1:1, where I might see 4000 unbuffered. Hopefully the guys on the Extreme forums will pick up on the DFI boards and figure out how to voltmod it. I think it's easy. Check out the Bleeding Edge forums on the DFI boards doing well there also.

wow, that is nice

I have 2x256 MB GeIL PC4200 running at 292MHz FSB 1:1- my CPU craps out at the exact same FSB as my RAM. The best mem Ive ever had, its running 3-4-8-2.5 right now with 3.2VDimm on the EpoX 4PCA3+