Binning ddr1 memory tccd bh5

[Mythical tech]

I just got a DFI NF4 ULTRA D and a few kits of tccd off ebay and am about to get 6 kits of Bh5 from various members on the forums. I am wondering if any one here is willing to share their experience of what timing and frequency is considered a very good kit and what is decent kit for both Bh5 and tccd memory ic's. Any and all help is welcome and appreciated. 

[TerraRaptor]

For winbond, I would count 270 2-2-2 with just 3.3v a very good kit. For tccd, although I have not that much experience, I would also consider IMC quality.

[TaPaKaH]

Over the last few years I've tested something like 80-85 sticks of BH5. I usually start by checking the sticks individually for three passes of test #5 at 250-260-270-275-280-285-290 at 2-2-2-5, then match the sticks based on required voltage and voltage tolerance, and then proceed with testing 32M in OS at 250-260-270-275-280.
So far, only six sticks out of the 80-85 would pass test #5 three times at 285 2-2-2-5 with any voltage (usually, 3.70-3.80V), and only two would do it at 290 2-2-2-5 (usually, 3.90-4.00V). The same sticks would do 32M at 275-280 2-2-2-5, usually around 3.70V. All of these are 256MB sticks. Best 512MB ones do around 280 for #5 and 270-275 for 32M.

[I.nfraR.ed]

I haven't tested as many modules, but I would aim for the numbers mentioned by Sam.

On s.939 usually I get to run 512MB modules at the 270-280 range. Most of the times in the 3.6 - 3.8V range, all tight.

512MB modules will give you and advantage in scores, but are a little harder to clock. For older platforms 256 modules are a must.

Edited May 20, 2019 by I.nfraR.ed

[TASOS]

I agree with the comments above.

Keep in mind that the quality of the cpu for socket 939 is crucial.

Also you have to test if a cpu with 512kb cache (instead of 1mb) is better for your memory testing.

An other aspect is the bios that you will use , to have the proper memory table for BH5 or TCCD.

 

[Mythical tech]

Thank you all for your help. What would you suggest for a bios for each ic. I am using the ultra d board 

[TASOS]

1 hour ago, Mythical tech said:

Thank you all for your help. What would you suggest for a bios for each ic. I am using the ultra d board 

https://www.techpowerup.com/articles/other/114

[TaPaKaH]

8 hours ago, I.nfraR.ed said:

512MB modules will give you and advantage in scores, but are a little harder to clock.

True. The difference in 32M on 939 is around 3-4 seconds but 256MB sticks usually clock 5-10MHz better. So in the end I'm not even sure which option will be fastest. 

What is more interesting, is that cross-sided and single-sided 256MB sticks have the same performance. Maybe it's just me, but for the last 15 years I was living under a notion that cross-sided is always faster

[Mythical tech]

I am very thankful for all of the information. I am getting BH5 from @TaPaKaH for benching with s939 and s426. I want to mess with max frequency if anyone has kits or single sticks I will probably buy them.

[ObscureParadox_3XS]

58 minutes ago, TaPaKaH said:

True. The difference in 32M on 939 is around 3-4 seconds but 256MB sticks usually clock 5-10MHz better. So in the end I'm not even sure which option will be fastest. 

What is more interesting, is that cross-sided and single-sided 256MB sticks have the same performance. Maybe it's just me, but for the last 15 years I was living under a notion that cross-sided is always faster

I suppose it doesn't really make a difference if the number of IC's is the same, you have the same theoretical data throughput with number of chips and clocks etc.Maybe there is a slight difference in latency but I'm not so sure.

[Mythical tech]

3 minutes ago, ObscureParadox said:

I suppose it doesn't really make a difference if the number of IC's is the same, you have the same theoretical data throughput with number of chips and clocks etc.Maybe there is a slight difference in latency but I'm not so sure.

Well if it acts like dual ranked ddr4 you may get a little bit more performance but from looking at the available info it does not look like it is going to be much different than single sided.

[I.nfraR.ed]

BTW, I have found one XS thread with some more info about possible clocks: https://www.xtremesystems.org/forums/showthread.php?276672-Winbond-BH-5-The-Final

 

So it turns out my result kind of matches what @TerraRaptor said - 270 2-2-2-5 3.3V.

PS: @Kotori's teamgroup xtreem (tccd) stick, that holds the current ddr record on the bot, scaled up to 3.7-3.8V IIRC, while mine was starting to show random crashes as soon as I give it 3.4+ volts.

Edited May 20, 2019 by I.nfraR.ed

[ZFeSS]

As TaPaKaH says, Voltage scaling make sense. I've got a stick of CH-5 that do 265@3.3v and a stick of BH-5 capable for 255@3.3v. But if you up voltage to 3.8 both sticks are equal and capable for 275 2-2-2-5.

[Alpi]

On 5/20/2019 at 10:55 PM, I.nfraR.ed said:

BTW, I have found one XS thread with some more info about possible clocks: https://www.xtremesystems.org/forums/showthread.php?276672-Winbond-BH-5-The-Final

 

So it turns out my result kind of matches what @TerraRaptor said - 270 2-2-2-5 3.3V.

PS: @Kotori's teamgroup xtreem (tccd) stick, that holds the current ddr record on the bot, scaled up to 3.7-3.8V IIRC, while mine was starting to show random crashes as soon as I give it 3.4+ volts.

Omg ! That's a scaleing from a tccd !! :O  Mines not showing much above 3.0 and starts to suffer at 3.2+.