trodas

Sure, show you they can go. Qimonda rams runs in memtest at default settings at 1772MB/sec. After setting them properly, they run at 2305MB/sec (2370 when used the Pipeline DRQCTL on): CAS - 3 DRAM Bank Interleave - 4 Way TRCD - 3 TRP - 2 TRAS - 8 TRFC - 9 TRRD - 2 TRTP - 2 TWTR - 1 TWR - 2 DRAM Bus Selection - Dual Channel DRAM Command Rate - 1T Kinda difference. The ram divider is locked at 5:6 and with the DDR 400 settings it get to 197MHz, when FSB is 164MHz. Not that bad for a Celeron D 336 (old Presscot core).

So I have to pick carefully. I was not having too much voltage headroom there with 2.1V maximum w/o moding. Witch 667 Crucials are 1G and CL3? Crucial Ballistix BL2KIT12864AA663 promise at DDR2-667 reasonable 3-3-3-12 timings, but 2.2V is stated as requirment: http://www.anandtech.com/show/1951/2 ...and there is not much results that give the "BL2KIT12864AA663" type with the CL3 ratings So you say that I could get the 3-3-3 timings from the rams, even that they are not guaranted to do this? Sounds reasonable (same chips cannot be all that different), but I still like to have the CL3 guaranted somehow. Look, there are even G-Skill DDR2 2GB 800MHz CL4 (2x1GB) F2-6400CL4D-2GBPK for sale now: http://pameti.heureka.cz/g-skill-ddr2-2gb-800mhz-cl4-2x1gb-f2-6400cl4d-2gbpk/ ...but they are CL4. What guarantees I have, that at 667MHz they reach stable CL3 clocks? The memory setting on ASRock 775Dual-VSTA are so complex that I never seen anything like that - ever. And I have my doubts that anyone could help me / explain me these settings. There are simply too much of them :-O Oh, you are so bad! Poor ram, 2.5V ... ts ts So how it get worser from such treatment? Have you measured the degradation somehow? Do it reach lower clocks now? That 3-3-3 timings at 450MHz sounds pretty good. Witch chips / rams precisely you are talking about...? So I could look for them and see. No 32M SuperPi stable? Then back down few MHz and try again? Somewhere the stable clock must be found, unless these rams are permanently damaged, witch would be unfortunate. Try...? Sure it can be a good option, but since I will be seriously clock-limited, then I have to get the absolute lowest DDR2 ram timings that exist for the best performance at given clocks. So CL4 is not really an option. I have my eyes on 3-2-2-8 DDR2 rams - even these that claim 3-3-3-12 can run 3-2-2-8 at 320MHz: http://www.anandtech.com/show/1951/7 http://www.anandtech.com/show/1951/9 "Crucial Ballistix, using Micron's latest DDR2 chips, is competitive with any DDR2 memory that we have tested in the recent past. The performance appears to be better than early versions of Micron's Fat Body D chips" PS: "even though DDR2-667 SDRAM with extremely low timings like 3-2-2-8 is just a little faster than DDR2-800 SDRAM with 5-5-5-15 timings" http://www.xbitlabs.com/articles/memory/display/ddr2-oc1ghz_10.html ...that is why I seek the best timings possible, since I fear that I will be limited to DDR2 667 speed.

Hmmm, only 512MB versions have the 3-3-3 timings: http://www.amazon.com/Ballistix-PC2-5300-3-3-3-12-UNBUFFERED-DDR2-667/dp/B000ACJQNA?tag=duckduckgo-d-20 The 1G versions Crucial Ballistix DDR2-1066 (PC2 8500) BL2KIT12864AA1065 have timings 5-5-5-15 (pheeew), yet use Micron D9 chips and looking very pretty: http://www.bjorn3d.com/2007/06/crucial-ballistix-2gb-pc2-8500/ It would require some testings, if they can run tight timings on lower memory clock speeds... What did you do to the poor ram chips, that you say that they are pertially degraded? Confes, infidel!

...yep, these Cell Shocks are looking good, but it will be hard to find any seller of them. So I have to keep an eye for alternatives too. Of course the problem is, that when using the target CPU, there will be some ram divider, as the mobo (ASRock 775Dual-VSTA) does support FSB 1066, but only a 667 for DDR2 rams. That tend to more that suggest, that the ram overclock will be quite limited (I will push the FSB up as high, as it can, with way worser mobo it get to 300MHz easily w/o any troubles and w/o even trying to mod the mobo in any way: http://hwbot.org/submission/2863476_...75i65g_300_mhz ), so I need tight timings. 3-3-3 must be always reachable at somewhat lower clocks. As far as voltage goes, on the High settings I get 2.1V into the DDR2 rams in this ASRock 775Dual-VSTA mobo. So unless we talking about voltage mod, the max. voltage will be 2.1V. Somewhat limiting as well, but these Cell Shock results ("The results around "3-3-3-5 @ 800 is only good at 2.0-2.1v.") would be satisfaction. With DDR(1) and Celeron D 336 (old Presscot core) is the ram divider at 5:6, eg. FSB 164 gives a 197MHz on rams: http://hwbot.org/submission/2911678_trodas_reference_clock_775dual_vsta_164_mhz It is, however (FSB 1066 vs max supported DDR2 667 rate) very likely, that such nice divider won't be used when I run Core Duo Extreme X6800 in the board. So when people are selling rams, they mostly focus on the maximum overclock with relaxed timings - example for the Team 800C3: http://overclock.pl/forum/threads/11118-wyprzeda%C5%BC-Micron-DDR2 "650 5-5-5-15 @ 2.4" - not really what I'm looking for. Also for the 3-3-3-8 timings I find only Geil's GX22GB6400EB3DC ... however they need 2.15 to 2.5V for it, witch is way too much...

Guys, I wonder, what 2x1G sticks to use in my ASRock 775Dual-VSTA to get the maximum possible ram speed. There are some cool rams out there, that can do very nice speeds, like these - 413MHz, 3-3-3-5 1T timings: http://hwbot.org/submission/582356_mayk_pcmark_2004_core_2_x6800_%282.93ghz%29_13547_marks (but user Mayk is not registred on this forum, so I cannot ask, what rams he used) - 294MHz, 3-3-3-9 1T timings: http://hwbot.org/submission/2738623_michaelnm_3dmark2001_se_geforce_6800_gs_%28nv42%29_agp_42240_marks (user michaelnm asked about witch rams he used) - 465MHz, 3-3-3-1 (?) 2T (?) timings: http://hwbot.org/submission/590221_roro_superpi___32m_core_2_x6800_%282.93ghz%29_9min_19sec_340ms (no user roro on this forum is registred) So I would like to know, what rams get, to get similar/same/better results. The aimed CPU is Core Duo Extreme X6800 with FSB 1066MHz, mobo ASRock 775Dual-VSTA (DDR2 667 supported, so 333MHz) and 3-3-3 timings is a *must* reach at such clocks. All I managed to find ATM is CellShock CS2221440 rams, witch are praised a lot in tests: http://www.memorybenchmark.net/ram.php?ram=MSC+Vertriebs+GmbH+CS2221440+1GB&id=2592 Show capability of 3-3-3-8 timings (at 1T up to 358MHz, witch is pretty good): http://www.modlabs.net/page/cellshock-cs2221440-ddr2-ekskljuziv-made-in-germany#.VZWd_oLw7Gg ...and overclockers pushed them a lot too, 440MHz, 3-3-3-5 with 2.51V (ugh) and unknown command rate: http://www.xtremesystems.org/forums/showthread.php?201773-Micron-D9-GMH-GKX-JKH-etc-club/page91 The results around "3-3-3-5 @ 800 is only good at 2.0-2.1v." would be great, but Google search returned nothing to sale (not even on eBay) for "CellShock CS2221440", so I probably have to look out for alternatives too. Anyone who can suggest a good 2x1G DDR2 rams with the 1T capable of 3-3-3 timings...? Not need to have TRAS 1, lol. That is probably some sort of error, lol. Cannot imagine that this is a real value :banana:

religious logic

Hoooray, it works now! Thanks!

Absolutely NO LUCK on submiting any score at all. Flustrating. VERY flustrating.

Using Celeron (single core) is just too easy to get the score: That works just fine. However to my sheer absolute flustration, there is no way to submit score into HWbot now! WTF ... This is insane. ... Anyway, slight bump for any ideas about dualcores!

What the hell... What happend? The picture of the mainboard is GONE: http://hwbot.org/hardware/motherboard/775dual_vsta/ ...and I find that is IMPOSSIBLE to submit score with Celeron D 336 and this ASRock 775Dual-VSTA mobo. What is going on?!

Now testing these nice Qimonda rams on ASRock 775Dual-VSTA mobo Things are looking good - first using testing Celeron D 336 2.8GHz, these rams can do at PC3200 (200MHz clock) 1T, dual channel (of course) and 4 Way interleave. 8 Way was a bit too much, sadly... Now off to finetune the other timings in hope to get even more speeeeeeeeed from them. Memtest ram speed jumped from default setting 1772MB/sec to 2133MB/sec

porn on mobile

To be honest, not quite sure, but that is what this test is all about, right? If overclockability will be higher after recap, then it is a viable strategy The Vcore input caps are very important for the whole regulator. Yet capacity is only one of about 5 major capacitor properties. These are: - capacity (the number of energy it can store - the higher the better) - voltage (as close to the working voltage as possible, beware of DC bias in ceramic caps, it could be 80% capacity off...!) - ripple current (maximum current the capacitor can deliver steadily w/o overheating itself, the higher the better) - ESR (resistance of the capacitor - the lower the better, ideal will be zero for infinite current ) - ESL (impedance of the capacitor - the lower the better, as that enhance frequency characteristics) - optimum operating frequency (polymers max. ripple at 120Hz is x0.05 (!), while for most elytes it is x0.5 coeficient ... you get x1 at 100kHz, but also polymers can go higher w/o losing the specs, elytes cannot) (also important is Leakage current, temperature range, dissipitation factor, endurance, shelf life... shall I continue? ) So with that in mind, let's compare these caps: Chemi-con KZG 1000uF 16V - max. ripple: 1870mA - ESR: 0.021 Ohms - leakage current: 160 uA Nichicon RNU 330uF 16V - max. ripple: 6100mA - ESR: 0.008 Ohms - leakage current: 1056 uA So even that the capacity is notably lower (3x lower that the original), the maximum ripple current is 3.26x higher. Also ESR is 2.62x lower, while leakage current is much higher with the polymer (6.6x higher), witch is basic polymer flaw (also a slight noise in signal, witch can be cured by using ceramic caps to bridge these caps...). Therefore my question was - is more important ripple current, or capacity in this particular place? The answer was - ripple current: http://www.badcaps.net/forum/showthread.php?t=46164 But test determine what happen I'm also very curious, because the noise and limited capacity could seriously impair the VRM design. Even Intel made some CPU VRMs with input elyte caps. Examples are Intel D955XCS or even overclocker board Intel D975XBX "Bad Axe" or Intel D925XCV - there phases and elytes on the input: So the future tell us, if the focus on ripple current is valid or not. And that would be an interesting test, don't you think? Keep in mind, that I have very well determined the stable OC, benchable OC (semistable) and max. OC So I will notice any change for better (witch I hope for) or worse (witch would trigger me to replace the caps for elytes). Sadly the time to get the caps is the problem. I would need something like "being sponsored by Nichicon" :-) ...or at least someone cooperative enought in USA (Digikey) for the purchases of hard to get caps...

Thanks! Quite messy, but working bench. What do you expect, heheh :-) News? Waiting for caps. That is usually the longest part of all my projects, because getting the right components is not only cost-prohibitive, but also very problematic and lenghty. Caps are ordered from Mouser.com, but no information about the order since 15.6.2015: https://cz.mouser.com/OrderHistory/OrderDetail.aspx?qs=9C%2fvjDbyW3V1fmZ4G2SNFokJWM6STzNO Dunno, how long this will took AND there are not even all the caps I need for the mobo ordered. These are missing (not only for this project, but so some other projects as well): 5x UCY2G470MHD6 (Frank PSU) 2x RR50G122MDN1PH (R 9600XT) 1x RR50J821MDN1PH (R 9600XT) 12x RF80G331MDN1PX (R 9600XT) 11x PLG0E222MDO1TD (ASRock 775i65G) 3x RR50J471MDN1KX (ASRock 775i65G) But since I have no other source that could help me out with the parts, then we have to wait... Sad, huh? ATM I could only show you that the clock generator is Winbond W83194BG-SD: (my ASRock 775i65G R3.0) ...or that the Vcore regulating chip is RT8841: (my ASRock 775i65G R3.0) Witch is sad. If the contolling chip was a L6714 chip, then the Vcore mod would be easy, because that chip have Offset pin, where when you put another that ZERO resistor, you get % of voltage increase. 40k is about 10% Vcore increase, but you can set more: http://www.controsensi.it/Mods/4CoreDual-SATA2/Contents_Eng.htm http://forums.vr-zone.com/hardware-depot/192571-volt-mod-done-right-asrock-775dual-vsta-l6714.html However only other ASRock boards have the L6714 chip. Such as the ASRock 775Dual-VSTA, witch I get now too: http://forum.hwbot.org/showthread.php?t=137706

perverse justice

My idea of replacement caps for the ASRock 775i65G R3.0 mainboard: ------------------- 4x Chemi-con KZG 1000uF 16V d8 -> 5x Nichicon RNU 330uF 16V (RNU1C331MDN1) (+ one empty near the PSU connector) 5x Samxon GE 1500uF 6.3V d8 -> Nichicon LG 2200uF 2.5V (PLG0E222MDO1TD) 1x Panasonic FL 680uF 4V d8 -> Nichicon LG 2200uF 2.5V (PLG0E222MDO1TD) 19x OST 1000uF 6.3V d8 -> 5x Nichicon LG 2200uF 2.5V (PLG0E222MDO1TD) -> 11x Nichicon R5 560uF 4V (RR50G561MDN1) -> 3x Nichicon R5 470uF 6.3V (RR50J471MDN1KX) 11x OST 100uF 16V d6.3 -> 8x Nichicon S8 1200uF 2.5V (RS80E122MDN1) (one empty) -> 4x Nichicon S8 560uF 6.3V (RS80J561MDNASQJT) + 11x ceramic caps on all Vcore caps + sound cap: ceramic 22uF 6.3V 1206 (JMK316AB7226MLHT) + 2-4x taltalpolymer caps on Vcore bottom near CPU: 470uF 2.5V (2R5TPF470M6L) - if I can fit them there somehow...

Slight bump for any ideas, about how to get the PCMark04 run sucesfully more frequently? BTW, the Win7 system is 32bit, my bad Once per like 100 runs is not very good result, lol.

Superb!

The board is known to HWbot: http://hwbot.org/hardware/motherboard/775dual_vsta/ ...but there is no picture. Take there the original from ASRock site: :celebration:

But is not HWbot about records and the best cards...? So if company made the BEST card, should not get the image of it to represent the card in question? Or you see in News pictures of the last runners on Olympic, or the first ones? Please answer this specific question. That is, where the root of the issue is. And if the product works best - it is still advertising, because HWbot show it on the first place. And user get points for that. So you already have not-so subtle advertising of the best products already. And since the ref design suxx, its noisy and perform worsest of all the designs - why show it? It exactly bear the "logic" of "let's photograph the losers, not the winners of Olympic games." So what is next? The average scores get most of the points, or the slowest scores will be rewarded most for the trying...? ... Sure it is NOT up to me to decide, but I want to make my opinion clear and I'm willing to defend it. In the end - how it is fair, when Asus made the best GTX 970, that is not there the image of their card? Did not they made the best ones? How that is fair? Or it is not (the whole HWbot) overclocking about finding the *BEST* hardware, fastest chip, best voltage mod and best cooling to achieve the seemingly impossible? The *BEST* result? Push the hardware over the limits and make it perform lightingly fast?

Well, to present an oposition - no. I did not think that THIS is fair. Fair is show image of the best card. And that is not the reference card. Othervise what do winner get? Nothing? Is that fair? Who can call that fair? Do HWbot give out points like a candy to mediocre results? No? Why then even try to build best card, when reference card win? That did not seems to be to be fair at all to me. This actually support mediocrity. Sort of like political corectness. An straightjacket for freedom of speech. ... Not to mention that the ref images are ughly (they are nothing like the Fury cards from AMD, witch are beautifully designed to impress) and photoshop fakes. That is another issue. ... I just see things differently. The winner should be rewarded. Somehow. When there is a chance - why not? We did not have the sales data to determine with is the most common GTX 980 (970) card there it is, but we do have data to find out, witch is the fastest of these. And reward the creator of the fastest card by at least the image, that it will shown the best card from these GTX 980 cards, there it is. That seems fair to me. Or do you see pictures of losing / last athlets in the news of Olympic games, or do you see the winners? (that is the logic that drive me to this choice and seems I did the mistake of openly explaining it, hoping for sort of approval... and getting something different )

And why do you think that?

I still believe that my method of choice was better :-) But you are the boss that determine the end result. Still... that is unfair for eVGA that wins this round

Interesting choice Wins what? :-)

I just checked and in *shock* realized, that GTX 970 and GTX 980 have no image on HWbot! WTF! That call for fix. http://hwbot.org/hardware/videocard/geforce_gtx_980/ At first I was like - well, I have to be objective and choose generic nVidia image to not side with some company (or being accused to do so). But upon checking the nVidia images there: http://www.geforce.com/sites/default/files-world/NVIDIA-GeForce-GTX-970-front.png http://www.geforce.com/sites/default/files-world/NVIDIA-GeForce-GTX-980-front.png I come to realize that these pictures are fakes. There is only the "7" chaging to "8" and each other pixel(s) are the same. So these nVidia images I did not want to be used, as they are clearly photoshoped fakes. Shame on you, nVidia! So to get more interesting images, I have to pick a sides. I decided that what company produced card, that get WR in Firestrike Extreme test, wins. In case of GTX 980, it is eVGA: http://hwbot.org/submission/2766654_steponz_3dmark___fire_strike_extreme_geforce_gtx_980_10459_marks 2300MHz core, absolutelly stunning result! For GTX 980, the image is this: http://cdn.videocardz.com/1/2014/09/04G-P4-2988-KR_XL_7.jpg (a bit cleaned up) Enjoy! :ws: