Question 1066mhz tri-channel vs 1600mhz dual channel ram? also different timings

Page 3 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.

realflow100

Distinguished
Feb 22, 2013
108
3
18,585
I have 2 computers. different board/socket and cpu

but once is tri channel 1066mhz and the other is dual 1600mhz

I have 4 ram sticks installed in both. CPU-Z reporting correctly in both cases.

800mhz base ram frequency in dual channel system (1600mhz)
533mhz base ram frequency in tri channel system (1066mhz)
I'm running patriot viper III CL9 type gaming memory

Dual channel system ram timings are
9 9 9 24 37 and 1T command rate
Tri channel ram timings are
6 6 6 18 120 with 2T command rate
I dont know if that makes much difference with the timings though.

How much of an overall performance difference between the 2? is it even worth considering?
 

Karadjgne

Titan
Ambassador
The timings of the stock settings are most likely so loose, that it's allowing a 1T command rate. Kind of like wearing shoes 5 sizes too large will allow for really thick socks. If you enable XMP to tighten up the timings and get better performance out of the ram, it's like wearing correctly sized shoes, thick socks won't work out well.

You can Google timings readers, programs like Thaiphoon Burner are very good for telling you a lot about your timings, the OEM, serial numbers etc.
 

realflow100

Distinguished
Feb 22, 2013
108
3
18,585
I see. ok. so don't worry about ram settings too much. I probably won't be able to get much out of my ram anyways in terms of actual performance improvement in games.
(even though Ive seen a review where my ram I particularly have seems to overclock exceptionally well and still stay cool)

also im concerned I will need a new heatsink/cooler for my pc too.. old one or my one from my other computer might not fit in the new motherboard.
What size bracket do they usually come with? Can I just slap any heatsink in it already? since the motherbaord is used. it most likely already has some kind of bracket for a cooler already. but I'm not sure on the screw hole/thread size or spacing. Is it a standard?
Will I need to jury rig a fan duct taped to the heatsink or something if I use my old cpu cooler?
I can get away with doing that maybe. until I get a new cooler next month or something.

Oh yeah also the IO shield. Do cases come with the IO shield separately to install. or do I need to look for one myself. Does the motherboard come with one?
is it fine if im missing the IO shield for a while? is that not an issue right away?
 

Karadjgne

Titan
Ambassador
Intel lga 1156 through lga1200 used the same mounting bracket because the hole spacing in the motherboard did not change. Basically anything 1st Gen through 11th Gen is identical and perfectly swappable as is. The only stipulation being power capacity of the cpu vs power dissipation of the cooler. You really don't want to use the cheap 95w Intel stock cooler from an i3 on a cpu like a 125w 11900k (can hit 250w under extreme loads) or even a 77w 3770k (can OC to upwards of 200w).

I/O shield is a motherboard part. It's specific to each individual motherboard design since it all changes according to the layout of the mobo rear panel. So a msi mpower Z77 I/O shield will not fit a asus Z77 v-lk etc.
 

realflow100

Distinguished
Feb 22, 2013
108
3
18,585
ok my cpu is staying the same. 3770k CPU
I'm not overclocking on the stock cooler but right now I'm not seeing temps go much beyond 60.0*C under full prime95 load on all cores. and under gaming its generally far cooler. almost same as idle temps.

my Xeon W3520 cooler does not have the same mount spacing as the 3770k I'm guessing by the vastly different socket types. I can not devise a way of making that work. which is sad. as the cooler is so much beefier on the xeon cpu. and it'd be a huge upgrade over the tiny 2-pipe 3770k cooler.

think I might have to ziptie or duct tape a fan to the 3770k cooler in the new motherboard :( (the fan is attached to the case with a plastic airflow redirecting shroud thing on the HP 8300 case. and does not attach directly to the cooler.)

its the HP 8300 elite small form factor case and motherboard. stock cooler. Theres not much options to do with it. besides zipties or duct tape.
I dont even know if the fan headers are compatible/right type. so i might have to just buy a new cooler anyways.

The z77 motherboard i plan on getting is used. so I'm not sure it'll even come with an IO shield/plate. I'm guessing they probably left it in the last case that it came out of. instead of keeping it with the motherboard.
 

Karadjgne

Titan
Ambassador
Heatsink has nothing to do with mounting bracket, and neither does the socket. The holes are what the holes are, regardless of socket.

The Xeon has a lga1366 mount. That's everything, a HEDT mount spacing. It's not mainstream lga115x mounting. So no, it won't fit, zip ties will not work. It may be possible to modify the mount to work, depending on the mounting itself, not that much difference between the hole spacing for lga775, 1336 and 115x, but it's enough where it will not work as is.
 

realflow100

Distinguished
Feb 22, 2013
108
3
18,585
My computer takes a long time to get to the windows loading/boot screen. showing a blinking cursor for up to 15 seconds straight.

After about 7 to 12 hours of running time the pc starts having problems opening programs insufficient system resources (cant even open mspaint or anything)
After a minute or so of this. screen goes black for a moment. and pc reboots. just giving a generic "blue screen" error in event log.

I think the memory is either incompatible or very slightly faulty. maybe the issue being 1600mhz ram and being ran at incorrect frequency by the different motherboard? 1066mhz and the timings are incompatible or something?

I removed the patriot 4x8gb ram modules and put back in the old 2x 2gb modules I had instead. so only have 4gb ram now. it booted LOTS faster. blinking cursor only showed for a second or 2 before windows started booting up.

Do you think its incompatible? Or slightly faulty? I tried running a memory test on it and it seemed fine when doing that.
Or maybe its just too much ram? is 32gb incompatible/too much for the computer model to handle? Why was it starting and letting windows boot?
is the "type" of ram incorrect like unbuffered or something?
Does it need to be the specific registered/unregistered type is that also causing problems?

When I pulled the 4 modules out they were all quite hot/warm and toasty. Not normal. they should barely be above ambient temperature.
why would the bios allow the computer to boot with incompatible ram?
 
Last edited:

Karadjgne

Titan
Ambassador
No, that cpu does not require 'special' memory. But it's specs call for 800/1066, native speed and you are sticking 1600 in there. The memory controller is likely having fits, the voltage is high (hot dimms) and it's resulting in lack of performance. If you applied (dunno if it's possible) some OC to the cpu, (at least as far as bumping up system agent, or vccio) that'd give the memory controller a boost and might better deal with it.

It's like a go-kart. They are built to run with a lawnmower engine. Get a strong one and they are pretty zippy and quick. That 1600 is like dumping in a V8 car engine instead. You are doing nothing but spin tires, the weight balance is all wrong and the thing is impossible to control.

But, you can always drop frequency. Try dropping those 4x 8Gb down to 1066 or 1200, non-xmp, the sticks have jadec tables for timings built into do that.
 
  • Like
Reactions: realflow100

realflow100

Distinguished
Feb 22, 2013
108
3
18,585
Not possible to change the timings. the bios is dumb. All you can do is disable turbo boost/speedstep/disable multi core. No xmp or anything.
Its already running at lower frequency. 1066mhz. but its still unstable. CPU-Z is showing it only running at 533mhz in the memory tab so I can't set it any lower. its already running lower.
its not possible to overclock the CPU in the Z400 workstation motherboard. the bios doesnt support it either. the bios is locked. its crap. theres no memory timings or anything to change.

voltage is the same though 1.5v same as in my other pc. but the timings are different when ran at 1066. its running at tighter timings. maybe thats making it unstable or something?

Which is more likely? unstable ram timings/frequency settings. or faulty ram?
its always about about 7-12 hours when the issue happens and i get insufficient system resources error. and then windows crashes and restarts after about a minute or 2 later.
 
Last edited:

Karadjgne

Titan
Ambassador
DDR stands for Dual data rate, if you think of data being transmitted on a frequency, that's a sine wave, DDR transmits on rise and fall, not just rise like the old SD-Ram (single data). Cpu-z, Windows (now) and a few other programs report the Data Rate, not the Dual data rate, so cpu-z saying 533MHz really means 1066MHz according to the ram.

Timings are a range per speed. You'll see that with xmp or jadec tables. The slower the Hz, the smaller the timings. 1600 was often 9-9-9-27, but the same ram at 1866 would have been 10-11-10-30 instead. Tightening the timings would be 1866 at 9-10-9-27. It's all relative. Corsair Dominator ram was the cats meow, it'd run 1600 at 6-7-6-21 etc. So the timings aren't tighter as such, just smaller, per respective speed.

1600 cl9 was a very small hair slower than 1866 cl10, but 1600 cl8 was the same amount faster than 1866 cl10.

Everything takes time. The timings are like the time it takes to open the door, how many paces to cross the room, open the next door. The speed is how fast your feet move. So even if you move your feet very fast, if the timings are really loose, that's a lot of very short steps, (think geisha). You can have really slow moving feet instead that uses less paces (giant steps) to cross the room, so ends up faster overall.

If all your errors are occurring after a long time period, all day, I'd suspect a memory leak more than a hardware issue. A memory leak is badly implemented software that's running, but not releasing the ram when it's done with a command.

So if you open a tab, and it uses 100mb of ram to open, that ram is reserved for that tab opening. When you close it, it cancels the reservation and the ram returns to the pool. With a memory leak, the reservation isn't canceled. That 100mb is still reported as being in use, so the next service or process to run has your 32Gb, minus 100mb. If it's a constantly used service, like Antivirus checker that's pulling 100Mb at a time, and not releasing it, it won't take long before that 32Gb gets used up, reserved for the Antivirus service, and then runs out of resources, leaving windows with zip for ram, so crashes.

What you do is wait 10 hours or so, then check ram use and available ram, figure out what program is soaking it all up. After that disable the program, reinstall it, update it, do whatever it requires to fix it.
 
  • Like
Reactions: realflow100

realflow100

Distinguished
Feb 22, 2013
108
3
18,585
Nothing is using any excessive amount of ram. I have left task manager open the entire time.
Its not using more than 4-5gb out of the 32gb that I have available. Typical amount that it normally uses.
My pc does not "slow down" over time either. it runs just as fast from the moment windows boots. till it gives the error. about 7-12 hours later. which the error just spontainously happens. then I get a bluescreen shortly after.
This time it happened while in a game and clicking "view friends profile" in the game. and that caused a NEW error to appear. I forgot what it was. but maybe something about "requested resources"

its got to be some kind of strange incompatibility. or instability problem.
I went through my pc just now and updated every chipset usb audio and video driver I could find for it.
including updating the bios version from (2013 date)
to the latest (2018 date) bios version.
I dont know if that will do anything to fix it though. windows may need to be completely reinstalled if windows is corrupted/broken.
 

realflow100

Distinguished
Feb 22, 2013
108
3
18,585
I got the new power supply. motherboard. and case for my computer to fix it. but my old cpu cooler/heatsink wont work. i need new cpu socket bracket adapter for it.
so i just bought a whole new cpu cooler + bracket combo. 18$ with 10 heatpipes and fan lol.
so now I have to wait another whole day for the cpu heatsink/cooler to get here before I can try to get my pc fixed/running again.
I'm still using the crappier xeon z400 computer setup right now.
 

realflow100

Distinguished
Feb 22, 2013
108
3
18,585
I made a new post/question about another issue ive started having. related to usb sound card and left and right audio channels "swapping" back and forth every few seconds.

I think it might even be related to some of my blue screens/crashes and weird errors that ive been getting. it might of gotten damaged when my computer psu and motherboard died together.
 
Last edited:

realflow100

Distinguished
Feb 22, 2013
108
3
18,585
I got my computer fixed!!!! Hurray!!!! :D New motherboard case psu and cpu cooler/heatsink!!!! :D

Even managed to overclock it to 4.5ghz and my temperature is below 70*C under a prime95 Small FFT 100% load :D
 

realflow100

Distinguished
Feb 22, 2013
108
3
18,585
Needed to up the voltage a smidge. wasn't QUITE stable. I got lucky the first round of stress testing oops! 1.310v now and max temp of 76C on hottest core. but averages between 60-69C under heavy stress test.
in gaming its usually under 60C! amazing :D almost water cool temperature when gaming :D
 

Karadjgne

Titan
Ambassador
Silicon has its safe voltage limits. As long as you try to remain well under those particular limits, all good, regardless of what the actual number is. DDR4, for instance, should not be taken over 1.5v for daily use and expect any longetivity out of it, so staying under 1.45v is best, but 1.4v is perfectly acceptable. Just because XMP sets 1.35v doesn't mean 1.37v is an issue.

There's always ways to get what you need, most issues are not with the components, but with owners perceptions. 70°C under stress is just fine, good even, you could go as high as 80°C and be perfectly acceptable to the cpu, but ppl perceive 80° as too hot, that's not a component issue but an owner issue.
 

realflow100

Distinguished
Feb 22, 2013
108
3
18,585
How about for DDR3?
I was thinking about trying to get a bit extra out of my ram but im not sure what the safe/recommended limits are.
and what I should aim for as a starting point.
patriot viper iii ddr3 1600mhz cl9 memory is what i have. 4x 8gb so 32gb total
 

Karadjgne

Titan
Ambassador
Ddr3 limit is 1.8v. Original releases were 1.65v and that lasted a year or so, but most commonly it's 1.5v.

You have an intel. The ram has a sweet spot of its own. Intels don't use ram the way AMD does, the ram is basically no more than a storage cache to an intel. So there's very few programs out there that will or can actually take advantage of higher speed ram.

As long as the ram is decently quick, doesn't stress the memory controller and doesn't have lousy timings, it's going to be about as good as it gets for real world applications. Stable ram is a better quality than faster ram to Intel and Windows. 1600/9 or 1866/10 is within a few hundredth of a nanosecond difference, or realistically none at all to anything but a benchmark.

If you have 1600/12 or 1600/13, then trying to get closer to 1600/9 might show small benefits in some load times, boot times, but that's about it.
 

realflow100

Distinguished
Feb 22, 2013
108
3
18,585
I managed to get a reasonable overclock on my ram!!! :D
2000mhz 11 12 12 28 and a few other timing tweaks. stable! :D

Also got my cpu up to 4.6ghz :D

edit I needed to drop the frequency of my memory to 2000mhz for some reason there was no errors in normal memory tests. even after many hours. at 2133
but prime95 large fft and occt large extreme had errors in like 1-2 minutes. so something was not sitting right with the memory overclock.
after that its fine so far. no errors. occt fine prime95 large fft fine.
 
Last edited: