Christer Posted November 14, 2010 Author Share Posted November 14, 2010 Wired, Did you get around to testing one stick at a time in the same slot? Not yet but I'll get around to it a.s.a.p. It takes time because the system must cool down between the tests and only one test a day can be carried out. That means ten more days, not counting the days I'll be away from home, if I have to test both modules in all four slots to get a "conclusive" result. (Gigabyte GTS recommended running single channel (slots 1+2 or slots 3+4) as a temporary solution. That's why I want to test this first.) In a previous post, I asked about the Command Rate. Right now, the modules are in slots 1+2, single channel and I noted that CPU-Z reports the Command Rate as 2T. It seems like, with the BIOS setting in "auto", one module per channel > 1T, two modules per channel > 2T but in BIOS, it still shows up as "--". The "--" is consistent, at least and that mystery is solved. Link to comment Share on other sites More sharing options...
Christer Posted November 16, 2010 Author Share Posted November 16, 2010 To summarize, with F3 BIOS: Dual channel in slots 1+3 and slots 2+4 > cold start errors, running Memtest86+ v4.10, showing up during the first pass after a few minutes. Single channel in slots 1+2 and slots 3+4 > no cold start errors, running Memtest86+ v4.10 for several passes/hours. I was about to run single module tests but I don't think there will be cold start errors. Four days ago, when there was cold start errors with a single module in slot 2, I had not yet updated the BIOS to F3. However, if you disagree, I'll run single module tests. I had concluded that the modules simply aren't matched to work in dual channel when I had a new message from Gigabyte GTS. I must say that I'm impressed by their support: We did run a complete test to check your reported error with exact same hardware as you, and with us everythings works fine in any slot combination. We did use the latest F5a BIOS with the test. You may try use the latest F5a BIOS as well: http://www.gigabyte.com/products/product-page.aspx?pid=3517#bios Use this procedure when flash BIOS: Download, extract and copy actual BIOS file on to the root directory of a USB memory stick. Enter BIOS by press DEL at boot, then Enter Q-flash utility by press F8 from BIOS screen. In Q-flash change “Keep DMI data” to “disable” before upgrade the BIOS file. The USB memory will appear as a hard drive in Q-flash. After successful flash, reboot, enter BIOS and load optimized defaults. Save (F10), exit and reboot. That supports my conclusion but I'll update the BIOS to F5a and run the dual channel tests. Kind of "back to square one" with fingers crossed. Link to comment Share on other sites More sharing options...
Christer Posted November 16, 2010 Author Share Posted November 16, 2010 While browsing the web for reviews of the TW3X4G1333C9A, at Newegg I found reference to TW3X4G1333C9A G. I checked the box in which my modules were delivered. It had "TW3X4G1333C9A G" printed on the label. I've seen no reference to the G at Corsair. What's the significance? I also zoomed in on the pictures at Newegg and saw a much higher version number (ver9.1) than om my modules (ver2.2). What's the significance? Have I been sold old modules and newer ones work better? Link to comment Share on other sites More sharing options...
Wired Posted November 16, 2010 Share Posted November 16, 2010 G is not part of the part number. G = RoHS compliant. Version numbers have nothing to do with when they were made and only indicate which IC manufacturer was used. Because single and dual channel configurations use different slots, it's still recommended to test one stick at a time. Link to comment Share on other sites More sharing options...
Christer Posted November 16, 2010 Author Share Posted November 16, 2010 Version numbers have nothing to do with when they were made and only indicate which IC manufacturer was used. My conclusion is that they are different and if the version tested by Gigabyte GTS is different than the one I have been blessed with, the version may be the cause of my problems. Because single and dual channel configurations use different slots, it's still recommended to test one stick at a time. If it doesn't work with BIOS F5a, I will! No, I don't work for Corsair! Please, don't misunderstand, I appreciate the input from everyone but maybe someone who works for Corsair has something to add? Link to comment Share on other sites More sharing options...
Christer Posted November 16, 2010 Author Share Posted November 16, 2010 Gigabyte GTS used ver3.13 in their tests. Link to comment Share on other sites More sharing options...
Wired Posted November 16, 2010 Share Posted November 16, 2010 3.13 what? That's not a version number of the memory, as it's always X.Y, no 3rd number (unless they've changed their version numbers recently). Link to comment Share on other sites More sharing options...
Christer Posted November 16, 2010 Author Share Posted November 16, 2010 3.13 what? That's not a version number of the memory, as it's always X.Y, no 3rd number (unless they've changed their version numbers recently). It seems like they have changed the version numbers. Gigabyte GTS attached a picture to their response to my question. Let's hope that I manage to attach it to this post. Link to comment Share on other sites More sharing options...
Christer Posted November 17, 2010 Author Share Posted November 17, 2010 BIOS F5a was not the solution. In dual channel, slots 1+3, Memtest86+ v4.10 returned errors at cold start. Reboot > still errors. Power off + Restart > no errors. I will run the modules, one at a time in slot 1 and slot 3 to see what happens. No matter what happens, since the same hardware (motherboard, processor and RAM) work in dual channel in my friends computer and in the one Gigabyte use for their supportive tests, I will RMA the modules via the shop where I bought them. Link to comment Share on other sites More sharing options...
Christer Posted November 18, 2010 Author Share Posted November 18, 2010 I found no space for questions regarding software so, I'll just post it here: Running Memtest86+ v4.10, dual channel in slots 1+3 and slots 2+4 returned errors at cold start. Running Memtest86+ v4.10, single channel in slots 1+2 and slots 3+4 returned no errors at cold start. I have not yet finished running single modules in the different slots. Since there are no errors at cold start running both modules in single channel, I don't expect errors at cold start running a single module. However, running module "A" in slot 1 returns 3981 MB/s, running module "B" in slot 1 returns 4089 MB/s. Are these values an indication of "mismatched" modules? The problem seems to be running dual channel but is it a pair of "mismatched" modules or is it the memory controller? I visited the "Memtest86+ Official forum" and in Which module is defective? the recommended method is: Try to swap the two modules to see if addresses will change, as it could be the memory controller the responsible (if the addresses of errors don't change). I also visted "Memtest86.com" and in Troubleshooting Memory Errors it is stated: 2) Rotating modules When none of the modules can be removed then you may wish to rotate modules to find the failing one. This technique can only be used if there are three or more modules in the system. Change the location of two modules at a time. For example put the module from slot 1 into slot 2 and put the module from slot 2 in slot 1. Run the test and if either the failing bit or address changes then you know that the failing module is one of the ones just moved. By using several combinations of module movement you should be able to determine which module is failing. Which is correct? Can I swap the two modules between channel 0 and channel 1 and by the reported failing memory addresses be able to determine if it is a problem with one of the two modules or, if it is a problem with the memory controller? If the modules are individually "okey" but "mismatched" for dual channel, will that trigger an error in the memory controller? Link to comment Share on other sites More sharing options...
Christer Posted November 24, 2010 Author Share Posted November 24, 2010 I've been away from home and have not had time to carry out tests in all slot/module combinations. Tests on BIOS F5a: Dual channel, slots/modules 1A + 3B Memtest86+ v4.10 Memory: 4094 M - 4601 MB/s Settings: RAM: 666 MHz (DDR1333) / CAS: 9-9-9-24 / DDR3 (64 bits) Cold start errors. Reboot - errors. Power off > Restart - no errors. CPU-Z DRAM Frequency 669.7 MHz, Command Rate (auto in BIOS) 1T Single module, slot/module 1A Memtest86+ v4.10 Memory: 2046 M - 3981 MB/s Settings: RAM: 666 MHz (DDR1333) / CAS: 9-9-9-24 / DDR3 (64 bits) No cold start errors. CPU-Z DRAM Frequency 669.7 MHz, Command Rate (auto in BIOS) 1T Single module, slot/module 1B Memtest86+ v4.10 Memory: 2046 M - 4089 MB/s Settings: RAM: 666 MHz (DDR1333) / CAS: 9-9-9-24 / DDR3 (64 bits) No cold start errors. CPU-Z DRAM Frequency 669.7 MHz, Command Rate (auto in BIOS) 1T Single module, slot/module 3A Memtest86+ v4.10 Memory: 2046 M - 3981 MB/s Settings: RAM: 200 MHz (DDR400) / CAS: 9-9-9-24 / DDR3 (64 bits) Cold start errors. Reboot - errors. Power off > Restart - no errors. CPU-Z DRAM Frequency 669.7 MHz, Command Rate (auto in BIOS) 1T Single module, slot/module 3B Memtest86+ v4.10 Memory: 2046 M - 3981 MB/s Settings: RAM: 200 MHz (DDR400) / CAS: 9-9-9-24 / DDR3 (64 bits) No cold start errors. CPU-Z DRAM Frequency 669.7 MHz, Command Rate (auto in BIOS) 1T Running a single module, each works fine in slot 1 but the transfer rates differ. In slot 3, module A gives errors but module B works fine. In slot 1, Memtest86+ v4.10 reports the RAM as 666 MHz (DDR1333) but ... :confused: ... in slot 3 as 200 MHz (DDR400). A BIOS issue? A bug in Memtest86+ v4.10? (CPU-Z reports 669.7 MHz in both slots.) Link to comment Share on other sites More sharing options...
Christer Posted November 25, 2010 Author Share Posted November 25, 2010 Yesterday, I fitted two 2 GB modules of a different brand. It was not a matched pair, just two modules off the shelf. This morning there were no cold start problems. I'll do a second cold start test tomorrow morning to confirm. Link to comment Share on other sites More sharing options...
Christer Posted November 26, 2010 Author Share Posted November 26, 2010 Yesterday, Memtest86+ v4.10 was running for 12 hours from cold start, no problems. This morning, no problems. My conclusion is that one of the TW3X4G1333C9A modules is defective. Link to comment Share on other sites More sharing options...
Christer Posted December 2, 2010 Author Share Posted December 2, 2010 I'm not into overclocking but I'm willing to pay "a bit" more for compatibility/stability. If we disregard the problems I've had with the TW3X4G1333C9A, would the CMP4GX3M2B1600C8 (auto detected, no fiddling in the BIOS): - increase the probability of a stable system running at SPD speed (1333 MHz) and SPD latency (9-9-9-24)? - default to 1.5V at SPD settings? Next, would the CMZ8GX3M2A1600C8 offer better compatibility/stability than the CMZ8GX3M2A1600C9 if both kits are running at SPD settings (auto detected, no fiddling in the BIOS)? Finally, there's the CMZ6GX3M3A1600C8 kit of 3x2GB. Will there be a "CMZ4GX3M2A1600C8" kit of 2x2GB? Link to comment Share on other sites More sharing options...
trackrat Posted December 2, 2010 Share Posted December 2, 2010 From personal experience the CMP4GX3M2B1600C8 seems to work fine as far as booting to the SPD specs which all of the RAM should. You have a dual-channel AMD mobo so you can't use the CMZ6GX3M3A1600C8 or similar triple-channel kits. The CMZ kits are just coming out from what I gather so you'd need to be sure and pick one that is AMD compatible if you go that route. The default DDR3 industry/mobo voltage is 1.5 V but higher frequency rated RAM usually runs at 1.65 V. Based on my experience the CMP4GX3M2B1600C8 will run @ 1333 MHz. @ 1.5 V. Link to comment Share on other sites More sharing options...
Wired Posted December 2, 2010 Share Posted December 2, 2010 Next, would the CMZ8GX3M2A1600C8 offer better compatibility/stability than the CMZ8GX3M2A1600C9 if both kits are running at SPD settings (auto detected, no fiddling in the BIOS)?Same stability. Finally, there's the CMZ6GX3M3A1600C8 kit of 3x2GB. Will there be a "CMZ4GX3M2A1600C8" kit of 2x2GB?http://www.corsair.com/products/vengeance/ Link to comment Share on other sites More sharing options...
Christer Posted December 2, 2010 Author Share Posted December 2, 2010 Guys, thanks for your comments! I've seen the Vengeance kits on offer and was surprized that there is a 6 GB kit, based on a 2 GB module (3x) but not a 4 GB kit, based on the same module (2x). I don't think I'll need 8 GB ... :sigh!: ... but I want to run dual channel! Link to comment Share on other sites More sharing options...
Christer Posted December 10, 2010 Author Share Posted December 10, 2010 I have received a new kit TW3X4G1333C9A. I fitted the modules in slot 1 and slot 3 running in dual channel. Tuesday to Thursday, I have done three coldstarts, running Memtest86+ v4.10 for ~12 hours without errors. Today after the fourth coldstart, still no problems. There's one "peculiarity" that I fail to understand. Running Memtest86+ v4.10 with a single module: For both Corsair kits, in slot 1, for one of the modules 4089 MB/s was reported, for the other module 3981 MB/s. In slot 3, 3981 MB/s for both modules. For the "other brand" (two modules off the shelf, not a matched kit), 3981 MB/s was reported for both modules in both slots. In dual channel, slot 1 and slot 3, for both Corsair kits and also for the two "other brand" modules, 4601 MB/s was reported. Why do the Corsair modules differ in slot 1, single module (4089/3981 MB/s)? I was under the impression that they should be "matched"! Link to comment Share on other sites More sharing options...
Christer Posted December 17, 2010 Author Share Posted December 17, 2010 I have received a new kit TW3X4G1333C9A. I fitted the modules in slot 1 and slot 3 running in dual channel. Tuesday to Thursday, I have done three coldstarts, running Memtest86+ v4.10 for ~12 hours without errors. Today after the fourth coldstart, still no problems. There's one "peculiarity" that I fail to understand. Running Memtest86+ v4.10 with a single module: For both Corsair kits, in slot 1, for one of the modules 4089 MB/s was reported, for the other module 3981 MB/s. In slot 3, 3981 MB/s for both modules. For the "other brand" (two modules off the shelf, not a matched kit), 3981 MB/s was reported for both modules in both slots. In dual channel, slot 1 and slot 3, for both Corsair kits and also for the two "other brand" modules, 4601 MB/s was reported. Why do the Corsair modules differ in slot 1, single module (4089/3981 MB/s)? I was under the impression that they should be "matched"! I repeat my question, marked in red color and bold. I would appreciate an explanation ... :[pouts: ... ! Link to comment Share on other sites More sharing options...
Christer Posted December 23, 2010 Author Share Posted December 23, 2010 I repeat my question, marked in red color and bold. I would appreciate an explanation ... :[pouts: ... ! One final bump before giving up! To be honest ... :evil: ... the lack of response makes me inclined to regard the "matched kits" as a hype designed to increase sales. Since it's Holidays ... :feedback: ... I'll put that conclusion on hold for a while longer! Link to comment Share on other sites More sharing options...
Wired Posted December 23, 2010 Share Posted December 23, 2010 No idea, might want to RMA them. Matched kits are definitely NOT hype. Way too much evidence to support it. Link to comment Share on other sites More sharing options...
Christer Posted December 23, 2010 Author Share Posted December 23, 2010 I RMA'ed the first kit which didn't work in dual channel and I thought the "4089/3981 MB/s" thing was the problem. When I received the second kit, it worked (and still does) but these modules too are differing in performance (4089/3981 MB/s) depending on which slot they go into. They should have the same performance, whichever slot, right? Memtest86+ v4.10 gets "other data" wrong but I'm having trouble convincing myself on this one. I'm trying not to put words in their mouth but when I asked Gigabyte, they responded "trust CPU-Z"! I had to resort to "slight provocation" ... ;): ... but thanks for not leaving this hanging! Link to comment Share on other sites More sharing options...
Wired Posted December 24, 2010 Share Posted December 24, 2010 Wait, so if you take stick A, and place it into slot 1, it says 4089 MB/s, and if you place the same stick into a different slot it says 3981 MB/s? Link to comment Share on other sites More sharing options...
Christer Posted December 24, 2010 Author Share Posted December 24, 2010 Wait, so if you take stick A, and place it into slot 1, it says 4089 MB/s, and if you place the same stick into a different slot it says 3981 MB/s? Yes but I have only tested the first slot of each channel, slot 1 and slot 3. If I take stick B and move it between slot 1 and slot 3, it says 3981 MB/s in both slots. It was the same with the first kit as it is with this second kit. Link to comment Share on other sites More sharing options...
Wired Posted December 24, 2010 Share Posted December 24, 2010 That makes no sense: Stick A / Slot 1: 4089 Stick A / Slot 3: 3981 Stick B / Slot 1: 3981 Stick B / Slot 3: 3981 Is that correct? Link to comment Share on other sites More sharing options...
Recommended Posts
Archived
This topic is now archived and is closed to further replies.