RAM Tweakers thread

[theonetruewill]

Scoring is now updated by Paulieg.

After I was prompted by a thread asking about which RAM to buy I decided to see how far I could push my memory. The results were fantastic and with further encouragement and competition from Smartbomb we both managed to far exceed our expectations and current performance by really putting the pedal to the metal on our speeds and timings. As I found this so useful I thought why not start up a thread solely for helping people to increase their RAM speeds and for all of you guys to show off what insanities you can bring to the memory world. be it DDR, DDR2, or DDR3 - all are welcome.
There will be a leaderboard based on %overclock from specifications and of course just a plain max speed score list too! Get cracking guys- I want to be off the number 1 spot in a few minutes - especially with you Transcend Axeram guys about
CPU-Z screenie needed for confirmation of speeds.


RAM Max Speed! - DDR2
1) Wile E - 1370MHz - DDR2 1200 Transcend aXeRAM (5-5-5-18) - 2.36v - DDR2 Record holder!
2) giorgos th. - 1368MHz - DDR2 800 Team Xtreem (5-5-5-15) - 2.56v - DDR2 800 Record!
3) sneekypeet - 1300MHz - DDR2 667 Crucial Ballistix (6-6-6-18) - 2.3v - DDR2 667 Record!
4) rangerone - 1266MHz - DDR2 1200 Transcend aXeRAM (5-5-5-18) - 2.25v
5) paulieg - 1264MHz - DDR2 1066 Buffalo Firestix (5-5-5-15) - 2.2v - DDR2 1066 Record!
6) DOM - 1260MHz - DDR2 1066 Ballistix (5-5-5-15) - 2.1v
6) erocker - 1260MHz - DDR2 1200 Transcend aXeRAM (5-5-5-15) - 2.1v
6) philbrown23 - 1260MHz - DDR2 1066 Cellshock (5-5-5-15) - 2.4v
9) mab1376 - 1250MHz - DDR2 1066 Transcend aXeRAM (5-4-4-12) - 2.225v
10) infrared - 1225MHz - DDR2 800 SuperTalent (5-5-5-10) - 2.3v
11) Darknova - 1200MHz - DDR2 Ballistix Tracers/Lanfest (5-5-5-15) - 2.3v
11) Arctucas - 1200MHz - DDR2 1066 Mushkin XP-8500 (5-6-5-14) - 2.3v
13) theonetruewill - 1158MHz - DDR2 800 Geil Black Dragon (5-5-5-18) - 2.2v
14) infrared - 1152MHz - DDR2 1066 Ballistix (single-sided) (5-5-5-14) - 2.2v
15) sno.lcn - 1144MHz - DDR2 1000 Team Xtreem (4-4-4-10) - 2.3v UNCONFIRMED
17) tigger69 - 1125MHz - DDR2 1066 Ballistix Tracers (5-5-5-15) - 2.2v
17) EastCoasthandle - 1125MHz - DDR2 1066 Corsair XMS (5-5-5-15) - 2.11v
18) Smartbomb - 1120MHz - DDR2 800 OCZ ReaperX (5-5-4-15) - 2.2v
19) Nitro-Max - 1104MHz - DDR2 800 OCZ Reaper (6-6-6-15) - 2.2v
19) sneekypeet - 1104MHz - DDR2 800 Geil Esoteria (5-5-5-15) - 2.3v
21) Smartbomb - 1100MHz - DDR2 1066 OCZ Platinum XTC (5-5-5-15) - 2.0v
21) Johnspack - 1100MHz - DDR2 1066 Mushkin Ascent (5-5-5-15) - 2.1v
22) CrAsHnBuRnXp - 1095MHz - DDR2 1066 Transcend aXeRAM (5-5-5-15) - 2.2v
23) HTC - 1061MHz - DDR2 800 Corsair DHX Dominator (5-5-5-15) - 2.1v
24) CrAsHnBuRnXp - 1051MHz - DDR2 800 G.Skill (5-5-5-15) - 2.2v
25) marsey99 - 1050MHz - DDR2 1000 OCZ (2x 2GB set) (5-5-5-18) - 2.1v
26) cdawall - 1044MHz - DDR2 800 Patriot Extreme (5-5-5-15) - 2.2v
27) P4-630 - 960MHz - DDR2 800 Geil Ultra (4-4-4-12) - 2.2v
28) cdawall - 923MHz - DDR2 533 Samsung (5-5-5-15) - 2.3v
29) JrRacinFan - 899MHz - DDR2 800 OCZ Gold Cas5 (4-4-4-13) - 2.2v
30) hat - 819MHz - DDR2 533 Crucial (5-5-5-15) - 2.1v UNCONFIRMED
31) Anusha - 801MHz - DDR2 667 Kingston Value (5-5-4-12) - 1.8v
32) p_o_s_pc - 672MHz - DDR2 1066 OCZ Reapers (3-3-3-10) - 2.1v


RAM Max Speed! - DDR
1) sneekypeet - 295MHz - DDR 400 Ballistix Tracers (3-3-3-8) - 3.0v - DDR Record!
2) spud107 - 271MHz - DDR 400 OCZ Platinum (3-3-3-8) - ?.?v
3) repsol - 265MHz - DDR 400 OCZ El XTC Platinum (3-3-3-8) - 2.85v
4) Ninkobwi - 264MHz - DDR 400 OCZ Platinum (3-3-3-8) - ?.?v

RAM Max Overclocks! - Currently DDR2 only
1) sneekypeet - 94.9% - 1300MHz - DDR2 667 Crucial Ballistix (6-6-6-18) - 2.3v - DDR2 Max OC Record!
2) cdawall - 73.2% - 923MHz - DDR2 533 Samsung (flashed to 800) (5-5-5-15) - 2.3v
3) giorgos.th - 71% - 1368MHz - DDR2 800 team Xtreem (5-5-5-15) - 2.56v
4) infrared - 53.1% - 1225MHz - DDR2 800 SuperTalent (5-5-5-10) - 2.3v
5) theonetruewill - 44.8% - 1158MHz - DDR2 800 Geil Black Dragon (5-5-5-15) - 2.2v
6) Smartbomb - 40.0% - 1120MHz - DDR2 800 OCZ ReaperX (5-5-4-15) - 2.2v
7~) Nitro-Max - 38.0% - 1104MHz - DDR2 800 OCZ Reaper (6-6-6-15) - 2.2v
7~) sneekypeet - 38.0% - 1104MHz - DDR2 800 Geil Esoteria (5-5-5-15) - 2.3v
9) HTC - 32.6% - 1061MHz - DDR2 800 Corsair DHX Dominator (5-5-5-15) - 2.1v
10) CrAsHnBuRnXp - 31.4% - 1051MHz - DDR2 800 G.Skill (5-5-5-15) - 2.2v
11) cdawall - 30.5% - 1044MHz - DDR2 800 Patriot Extreme (5-5-5-15) - 2.2v
12) Anusha - 20.1% - 801MHz - DDR2 667 Kingston Value (5-5-4-12) - 1.8v
13) P4-630 - 20% - 960MHz - DDR2 800 Geil Ultra (4-4-4-12) - 2.2v
14) paulieg - 18.5% - 1264 - DDR2 1066 Buffalo Firestix (5-5-5-15) - 2.2v
14) Wile E - 14.2% - 1370MHz - DDR2 1200 Transcend aXeRAM (5-5-5-18) - 2.36v
15) Arctucas - 12.6% - 1200MHz - DDR2 1066 Mushkin XP-8500 (5-6-5-14) - 2.3v
16) JrRacinFan - 12.4% - 899MHz - DDR2 800 OCZ Gold Cas5 (4-4-4-13) - 2.2v
17) infrared - 8.1% - 1152MHz - DDR2 1066 Ballistix (single-sided) (5-5-5-14) - 2.3v
18) Arctucas - 12.6% - 1200MHz - DDR2 1066 Mushkin XP-8500 (5-6-5-14) - 2.3v
19~) tigger68 - 5.5% - 1125MHz - DDR2 1066 Ballistix Tracers (5-5-5-15) - 2.2v
19~) rangerone - 5.5% - 1253MHz - DDR2 1200 Transcend aXeRAM (5-5-5-18) - 2.25v
21~) erocker - 5% - 1260MHz - DDR2 1200 Transcend aXeRAM (5-5-5-15) - 2.1v
21~) marsey99 - 5% - 1095MHz - DDR2 1000 OCZ (2x 2GB set)(5-5-5-15) - 2.2v
23) CrAsHnBuRnXp - 2.7% - 1095MHz - DDR2 Transcend aXeRAM (5-5-5-15) - 2.2v

CPU-Z and Memset screenshots are mandatory, no screenshot = no scoreboard entry

[theonetruewill]

theonetruewill - 1140MHz - DDR2 800 Geil Black Dragon (5-5-5-15) - 42.5% overclock.

[JrRacinFan]

Would it be good to include voltage used with ram timings and speeds? I keep mine underclocked for temp reasons, it gets hot in this apt. I will post sometime later when i can get a good clock on mine.

[Darknova]

The thing I have to point out here is that there are diminishing returns. I'm sure I could overclock my RAM to 1200Mhz, but I won't, because I'll lose my low tRD, which is the only timing that has a constant effect on latency (ie 2.5ns less for every notch u go down eg. 6 -> 5).

I have booted at 1200Mhz, but couldn't go with less than a tRD of 7, which actually performs worse than 1100Mhz at 5.

Heh, I spent a lot of time reading about this. I'll link if required.

[Dr. Spankenstein]

and if we could include Memset with the SS it could help with sub-timings.

@onetruewill

If it's based on the % of overclock, don't be so sure those aXeram's are gonna make it! (1% of 1200=12MHz)

Your speed is very nice! I haven't heard one bad thing about those Black Dragon's.

[Darknova]

I would like to make a suggestion actually.

As well as a pure speed section, how about Latency and Mem Read sections?

2 basic tests which show a lot about the memory beyond the speed it's running.

As well as a full list of all timings given by Memset.

[theonetruewill]

Originally Posted by Dr. Spankenstein

and if we could include Memset with the SS it could help with sub-timings.

@onetruewill

If it's based on the % of overclock, don't be so sure those aXeram's are gonna make it! (1% of 1200=12MHz)

Your speed is very nice! I haven't heard one bad thing about those Black Dragon's.

The MAX speed scoreboard is where the AxeRam's will shove everyone else back.

The thing I have to point out here is that there are diminishing returns. I'm sure I could overclock my RAM to 1200Mhz, but I won't, because I'll lose my low tRD, which is the only timing that has a constant effect on latency (ie 2.5ns less for every notch u go down eg. 6 -> 5).

I have booted at 1200Mhz, but couldn't go with less than a tRD of 7, which actually performs worse than 1100Mhz at 5.

You may post whatwver you want - this is not just about a scoreboard - but also sharing information - we may add a scoreboard for read write times etc. Thanks for your input.

Would it be good to include voltage used with ram timings and speeds? I keep mine underclocked for temp reasons, it gets hot in this apt. I will post sometime later when i can get a good clock on mine.

The voltages sound like a very good idea - implemented.

[Darknova]

Originally Posted by theonetruewill

You may post whatwver you want - this is not just about a scoreboard - but also sharing information - we may add a scoreboard for read write times etc. Thanks for your input.

Ok, well I'll start you guys off with a bit of reading material upon the issue of tRD.

Anandtech's Discussion about tRD

Anandtech's article about the X48, and why it IS an improvement over the X38

I will just run a few benches, and get back to you with my results, I will do a number of benches at different tRDs to show you the impact.

[Nitro-Max]

Run everest cache and memory benchmark too see what read/ write /copy speeds you get.

[Darknova]

Alrighty, so here are the benches.

All tests done with my E8400 at 3.3Ghz, RAM at 1.1Ghz with timings of 5-5-5-15, tRFC @ 42, and the rest of the subtimings consistent.

The only variable here is the tRD.

I have done both Latency and Read via Everest, using the same version throughout. I will do extra benches if people require, but I thought this would be enough for now.

So, let's start with a tRD of 8.





tRD = 7





tRD = 6





And finally, tRD = 5





I'm pretty sure the results speak for themselves, but as you can see, between each run there is a decrease in latency of ~2.5ns and a read bandwidth of ~300Mb/s

[Dr. Spankenstein]

^^For 4 gigs, thats pretty awesome!

Haven't had much time to fiddle with the aXrams, but here's a prelim:

[Darknova]

tRD of 10??? Ouch dude...that's really going to hurt your perf!

You're running a divider of 2:3, so you should be able to use a tRD of 5 or 6

Oh, and Spank, I want a SS of your latency at 10, and 5

[Smartbomb]

Fantastic, here where my best results on OCZ ReaperX PC6400.
MAX SPEED 1120mhz 5-5-4-15 TRC34 2T
MAX PERFORMANCE 1066mhz 5-4-4-15 TRC26 2T

[Darknova]

Originally Posted by Smartbomb

Fantastic, here where my best results on OCZ ReaperX PC6400.
MAX SPEED 1120mhz 5-5-4-15 TRC34 2T
MAX PERFORMANCE 1066mhz 5-4-4-15 TRC26 2T

Pic of Memset, and a latency run of Everest please

[Nitro-Max]

just use the cache and memory benchmark in everest /tools/ cache memory benchmark that will give you all the results in one screenshot.

[HTC]

Here's my current:

[Darknova]

CPU-Z doesn't give enough memory timings to be useful. Use Memset instead, gives all subtimings.

[HTC]

Originally Posted by Darknova

CPU-Z doesn't give enough memory timings to be useful. Use Memset instead, gives all subtimings.

Have a link for it?

[Darknova]

Memset - http://www.xtremesystems.org/forums/...ad.php?t=92190

[theonetruewill]

Pleased with the feedback so far - Can I hear votes for adding a Max memory read and write section. Please vote either way as it's a lot of work to keep these scoreboards up to date- I have done quite a few in the past. If there is enough interest, I will of course do it - but if not then sorry, too much hassle.

Also DarkNova's suggestion of requiring the everest cache memory benchmark as necessary - I will give this some thought- mostly because CPU-Z is easier I am against it - but I accept the merits of the further timings displayed as being far more useful, especially in a thread like this.

[Darknova]

Originally Posted by theonetruewill

Pleased with the feedback so far - Can I hear votes for adding a Max memory read and write section. Please vote either way as it's a lot of work to keep these scoreboards up to date- I have done quite a few in the past. If there is enough interest, I will of course do it - but if not then sorry, too much hassle.

Also DarkNova's suggestion of requiring the everest cache memory benchmark as necessary - I will give this some thought- mostly because CPU-Z is easier I am against it - but I accept the merits of the further timings displayed as being far more useful, especially in a thread like this.

Firstly, I think we only need Max Overclock, Latency, and Read sections.

Second, I suggested Memset, neither CPU-Z nor Everest shows the subtimings, and most importantly tRD.

Third, Nitro-Max suggested using Everest's Cache benchmark because you get Read/Write/Copy and Latency in one window

[HTC]

Here it is: this time, with an 8 multi, instead of 6.



Notice the motherboard temp on probe and aisuite: WTF is that?

[Darknova]

Divider of 5:6....try lowering your tRD (Performance Level) to 5 HTC. See how that helps. I would say a latency of ~50ns, and bandwidth of ~9600Mb/s

[Dr. Spankenstein]

@ onetruewill,

A: Your call.
B: Yes, definately.

[theonetruewill]

Darknova I need your RAM's exact name if you want to go on the score board - I'm not sure which of the ones in your specs it is- same with HTC.
That goes for everyone please, even if it Generic cheap ass memory- write 'Generic'