• A Great friend to the HardForum with a great kid that he is trying to get a scholorship to continue his schooling. Please give hime a vote! Only 24 hours left! Thanks.
    If you have an VOTE FOR KEENAN!

E8400 Overclocking Results and a question

F

franzkefka

n00b
Joined
Nov 24, 2008
Messages
36
Hello all, just checkin in figured id let you all know how im doing on my 790i e8400 overclocking.

(remember this is under watercooling)
Currently testing with Prime95
478x9 @ 1.381V (via bios specification) everest reports 1.320v under load and around 1.36v at idle

Temps are looking like
Cpu: 52'c
Core1: 52'c
Core2: 52'c
SPP: 48'c
MCP: 49'c

Im running in Vista 64bit
8gb of DDR1333 ((4x2gb Gskill F3-10666CLD9-4GBNQ 9-9-9-24 1.5-1.6v rated)
if you check Gkills site for that memory it doesnt support 790i chipset but works flawlessly passed 8 hours of memtest)
E8400 SLB9J
EVGA 790i Ultra SLI
EVGA 8800 GTS 640mb
2x250GB raid0
1000W zalman PSU

i also had a question reguarding HT Multi in bios as to weather or not i should knock it down as i overclock, im new to intel cpu's and i remember that with AMD's you had to decrease to keep it under 1000mhz x 2
 
Currently at 2 hours into a dual prime run

4300mhz (478x9)
Useing voltages as follows:

Cpu Core 1.38125v (1.320v reported by Everest at load)
Cpu FSB 1.15000v (attempted 1.10v but would not POST)
memory 1.6000v (upper limit of manufacturer's spec)
SPP 1.30v (stock voltage)
MCP 1.50v (stock voltage)

(i also forgot tho specify i am using 99924-1T timing @ 1400mhz on ddr1333)

What should i aim for if this passes?
should i try to work the memory speed up to its max without slacking the timings to gain some Bandwidth or should i aim for more CPU mhz
 
Looks good to me. I think you got a winner there. If the e8400 is stable with those voltages, I personally wouldn't go any higher. I would recommend that you also test/stress your system running various different applications, games, HD videos, etc. to ensure that it is fully stable. Prime stability doesn't always equate to real world stability.

As far as memory is concerned, increasing the speed probably won't provide any substantial real world gains. Then again, I haven't really played around with DDR3 RAM so your experience may differ. I normally run Winrar's Benchmark and Hardware Test and Everest Ultimate's Cache and Memory Benchmark to see what kind of performance gains I've achieved after OC'ing any RAM. They both provide relatively reliable performance improvement or degradation data from my experience.

I want to point out couple of thing about Everest Ultimate. Make sure you adjust TJmax temperature setting to 100 Celcius to get correct temp readings for your CPU cores, if you already haven't done so (Preference -> Hardware Monitoring).

Everest Ultimate doesn't report correct SSP temps, just FYI. As far as I know, there is no software that can detect Nvidia NB temps accurately. Your MCP temps look pretty good though.
 
Ok So now im at 4300mhz
478x9
Cpu Core: 1.38750v in bios 1.32000-1.32800v under load via Everest.
Cpu Fsb : 1.15v
memory : 1.600v
Spp : 1.30v
MCP : 1.50v
Currently 5 hours deep in a Prime 95 run

I had some problems hitting 4342mhz, but i dont think it was temperature or Voltage related, i think the Memory was to blame, im currently running at 1333mhz 9-9-9-24 1t timings as per manufacturers spec ( i also confirmed stability with a 8hour run of memtest+86 over night) But when i was testing 4342mhz (482x9) i was running at ~1390-1400mhz memory (unconfirmed stability). Where as normaly from what iv seen Prime would normaly just report a rounding error on a core or both, It was causeing BSOD's, Id start prime at night before bed, wake up and the computer had restarted and windows error reporting was up on screen, so assumably was a BSOD.
Also saw it Bsod a couple times durring prime runs usualy between 1-3 hours in.
currently the test was started at 14:13 and it is currently 19:03 my time.

From what i read my Voltages are pretty low usualy everyone is up on SPP and MCP and even FSB voltages ( i know to keep the FSB Volts below 1.400v on a 45nm )
So if this were to fail should i attempt to raise the CPU core voltage first or FSB voltage.

I also read somewhere that a person had taken my memory which is the
Gskill F3-10666-4gbnq 2x2gb kit Was running 2 sets so 4x2gb at speeds close to 1600mhz on less than stock timings. Just mentioning i dont think theyre capable of that high of a speed never the less running @ 888-22-1t or even 2t.

Personaly if 4300 is the highest i can get this to go even on a decent water cooling system, i can live with that, its kinda funny rather as my last processor also topped out at about a 45% oc was a Opteron 170 dualcore (one of the very good steppings) 2ghz stock, 1.35v, i was running mine at 2.9-2.95ghz at around 1.45-1.5v also had the IHS stripped off it, so bare core on water and even that was physicly limited, didnt run over 45'c at load even up around 1.55v+
 
1.3v for SSP is pretty low. Try upping that to 1.4v -1.45v and stress test again. Nvidia boards can safely handle voltages at this range for the NB. With all the data flowing through your NB, you want to ensure that it is getting enough power.

For my e8600 @ 4.4ghz, I have the FSB voltage at only 1.175. You don't need to go anywhere near 1.4v for these 45nm C2D's. If anything, keep it under 1.3v.

I personally wouldn't raise the cpu voltage any higher. I would rather clock down a bit and find the next best stable overclock. But, that's just me. Try bumping the CPU voltage up a notch or two and test it again if you really want to overclock higher than 4.3ghz.
 
you stated above that Tjmax? should be set to 100 so is it 100 or 105 becuase the above temps were recorded using everest with 105 as the TJmax if i set it to 100 my core temps are about 46'c rather than 52'c
 
Ok so i just found a Post over at the xtremesystems forums, in which they state that TJmax should be set to 95c

So my load temps
95c tjmax
cpu 52-53'c
core1 42'c
core2 42'c
spp 48'c
mcp 49'c
gpu 38'c

100'c Tjmax
core1 46'c
core2 46'c

105'c tjmax
core1 51-52'c
core2 51-52'c

If i had to make an educated assumption, id say that tjmax 95'c is closer to right, due to the fact that my 8800gts is in the same watercooling loop and is directly after the cpu ie no heat exchanger between, and normaly during a dual prime run on my older opteron 170
would stay around 40-41'c during a prime run and currently is at 38'c durring a prime run,

my old notes state that my opteron was doing 300x10 @1.500v durring a long prime run temps looked like cpu 38'c core1 44'c core2 44'c and the Gpu was at 40'c

currently running prime with tjmax set to 95'c 478x9 @ 1.38750v durring a long prime run temps looked like cpu 52'c (deffinitly not reporting proper) core1 42'c core2 42'c and the gpu is at 38'c

its deffinitly not apples to apples, but i cant see my gpu staying below 40'c durring a dual prime if my cpu was pushing out 126'f would deffinitly heat the water up, and push the gpu temp well over 40'c

and Further iv noticed that my top 120x3 exhaust isnt pushing out air thats that much warmer than room temp. where as my old config with a nf4 and opteron would push out well over room temperature air from the top exhaust.

just seems more legit to me, i under stand aswell that the accualy cpu temp could be as high as 53'c and still not be heating the water up due to poor Ihs to core contact, but aside from that my thermal paste installation is good, so aside from the Ihs to core contact possibility i think that tjmax 95 may accualy be closer to the truth than 105 or even 100

thats my two cents

Considering adding an inline water temp probe between the cpu and Gpu blocks to better get an idea of what kind of temps were accualy looking at
 
The TJmax for e8400 is 100 according to intel. Do a search here on hardforum and you'll find the info.
 
Ok thanks for the Info,
im up to 4350mhz 482x9
Voltages
cpu core 1.41250v
cpu fsb 1.200v
memory 1.600v
spp 1.350v
mcp 1.550v

Idle temps are around
motherboard 30'c
Cpu 33'c
core1 31'c
core2 34'c
spp 48'c
mcp 48'c
gpu 36'c

these settings passed 10+ hours of Dual prime, using large inplace FFT's
also just played about 2 hours of CS:s and did a bit of photoshoping no problems so far
I dont have screenies of validation, will bother with that when i get higher, might try for 4400mhz. my load temps were well below that of one running on the best air cooler out there.
i am currently using 100'c as tjmax and those above temps were recorded using 100'c tjmax

Im going to try for 4400mhz using the same settings as above, if it works sweet, if not i dont think im going to push the voltages any higher yet. will prolly fall back to 4350 and try work the voltages down more.
 
That's pretty impressive OC for e8400. I'm guessing that you have a E0 stepping.

Try playing more graphical demanding game than CS:S for several hours. Run Crysis at the highest playable settings if you have the game. Another good game is Company of Heroes. When CS:S passed with flying colors, CoH always crashed to desktop when my system wasn't fully stable after playing an hour or two.
 
Just to let you know, Large in-pace FFTs does not stress the CPU as much as the Small FFTs in Prime.

Make sure you stress test with Small FFTs.
 
So far its run Cs:s for about 4 hours straight no problems there, also played a few hours of COH with the res set to 1920x1200 4xAA mostly high settings, also played some crysis no problems in any of them as far as crashing goes during gameplay, i did have an issue with crysis crashing when i exit the game but it seems to have gone away.

Im at 4350mhz 482x9 @ 1.41250v and im starting to work the SPP and MCP voltage down to its minimum stable settings.

And as far as the large vs Small FFT's go everyone has theyre say on which to use, i usualy use Small fft's my self but recently be using Large fft's , and as far as that goes 100% load across both cores for more than 10hours with out crashing or generating errors, is stable to me.

Iv yet to see a core fail in prime except for once, it usualy just Bsod's reguardless of large or small.
where my old amd would fail a core if you steped the voltage down one below your stable settings this one just Bsod's i dont really know why memory checked out fine ran all 8gb of it with the timings tightend down to 8 8 8 22 1t for around 18 hours in memtest not a problem.
so its deffinitly not the memory.

once i get the voltages droped to theyre lowest stable settings im going to start working towards 4400mhz.
 
Back at it some more,

Iv gotten it to 4340mhz 482x9 @ 1.41250v - 1.15fsb - 1.30vspp - 1.500vmcp
been running quick primes all morning 1 to 2 hours
started at 4300mhz 1.41250v 1.10vfsb 1.30vspp 1.500vmcp
but it was crashing before i got a load on it
so i bumped the FSB volts up and it got stable at 4300 @ 1.41250v
still working it up in 20mhz intervals till it fails.
when it does ill push the fsb down in 5mhz intervals till it gets stable again
then work the voltages up a little more, from what i read tho this chip is pretty good, but deffinitly not amazing

Oh BTW 1.41250v was set in bios actual voltage reads at 1.35200v in everest durring priming
so yeah thats a pretty terrible .06050v droop
 
new update

im currently working on 4300mhz 478x9
Cpu volts = 1.38125v (bios setting) 1.3200v durring prime (via everest)
FSB Voltage= 1.15v
I had prime instability with the spp and mcp volts at 1.30v and 1.50v (Failing prime after 16m resulting in a BSOD)

SSP voltage= 1.350
MCP voltage= 1.550

currently 1hour into a prime run
room temp is 19.7'c
cpu - 49-52'c
core1 - 48-49'c
core2 - 48-49'c
spp - 48'c
mcp - 48-49'c
motherboard - 30'c

Seriously considering the Thermalright IFX cooler's for the north south bridges and mosfets and some chip sinks for the top side mosfets

i think the Mcp/Spp temps are a bit high, or atleast high for my taste
and im sure they would benefit from the additional cooling

maybe even get me into the 4400mhz range.
 
I think there's a problem ,cause with air ,I am at 4180 mhz ! I am running a (Granted ) E8500 and a Biostar TP45 with everything @ auto ,never even touched the bios save speedstep disabled .( my cooling is with a Big Typhoon ) ? Each time I start up (just put it together ) I up the FSB ,1 ...Temps are 42 idle right now ,they have hit 55,56 with real temp under load . With my ram I will not ( probably ) go higher change bios in Dram,and Voltage menues .So there you have it ,all on auto .pretty good I think ....

You should be way closer to 5 Ghz.
 
its beyond doubtful that this cpu will do over 4400 - 4500mhz max im limiting my self to a max of 1.43v set via bios.

further more im not quite sure about your board but auto on these 790's means if you increase the bus speeds and what not it will increase voltages aswell on its own you might wanna check that out on your board, as i would feel very uncomfortable letting the board set its own voltages.

currently testing

4400mhz

489x9 @ 1.41250v (via bios) 1.3520 durring prime load (via everest reading)
currently using 1.15v Fsb
1.45v spp
1.650v mcp

i also toyed around with the GTL Vref settings

using 1.41250v cpu 1.15v fsb 1.35v spp 1.600v mcp

my results were

-40mv Gtlvref prime fail 18m
-50mv Gtlvref prime fail 28m
-55mv Gtlvref prime fail 23m
-70mv Gtlvref prime fail 8m

so noteing the trend that times at which they fail worsen from -50 either direction i have chosen to stick with -50 on all lanes.
and i moved upto the aformentioned voltages and am testing with higher mcp volts to see if thats the culprit before moving up in cpuvoltage

if the times in which prime fails do not increase with the added spp/mcp voltage i will back it down again and test the next notch higher on cpu voltage

i may accualy get to 1.425v and call it quits and back down to
4300 @ 1.38v and call it a day.
 
Alright guy, i ended up scraping 4.4ghz not worth pumping that much Cpu voltage.
was taking upwards up 1.43v set in bios to get mildly close to stable at 4400mhz

where as i'm setting 1.38750v in bios to run at 4300

basicly im 100% stable at 4300 478x9 @ 1.38750v

ill post up my Notes in a later post maybe tomarrow when i scan them in, but im
using the following voltages to achieve 100% stability
cpu core 1.38750v
cpu fsb 1.150v
memory 1.600v
SPP 1.350v
Mcp 1.550v

GLTVref set to -50mv on all lanes

also managed to get my Gskill DDR3 1333 2x2gb kits to run upwards at 1630mhz
@ 9-9-9-24 1t (5 40 12 18 25)

But im wondering if anyone else noticed this
my bench mark scores scale marginaly with increases in SPP and MCP voltage
for example

the above settings 1.35v spp 1.550v mcp
Everest memory
Read 7813mb/s

1.40v spp 1.600v mcp
Read 7803mb/s

1.450v spp 1.650 mcp
Read 7818mb/s

This one is where it starts to make me wonder
1.400v spp 1.600v mcp 1.625v memory
read 7882mb/s


And if someone could answer this one for me, a reason to this madness would supprise me!

Heres the test
478x9 @ 1.38750v 1636mhz memory 9 9 9 24 1t 4x2gb
fsb 1.15v
mem 1.600v
spp 1.35v
mcp 1.550v
Everest memory test
Read : 7813mb/s
write : 10049mb/s
copy : 7278mb/s
latency : 82.7ns

333x9 @ 1.250v 1333mhz memory 9 9 9 24 1t 4x2gb
fsb 1.10v
mem 1.600v
spp 1.35v
mcp 1.550v
Everest memory test
Read : 9104mb/s
write : 7017mb/s
copy : 7103mb/s
Latency: 65.1ns

someone please tell me, how using the same memory timings 9 9 9 24 1t @ 1.600v 1636mhz is going to give a higher latency than at 1333mhz

1636mhz over 1333mhz
Read -1291mb/s
write +3032mb/s
copy +0175Mb/s
latency +17.6ns

So how do i Gain a huge amount of write speed, but loose a considerable amount of read speed as well as gain almost 18ns of latency by overclocking my memory.

on a Side note i also did a considerable amount of Game benchmarking at stock and my max overclocked speeds, and have come to the conclusion that this cpu rocks my 8800gts in that im completely GPU bound.
3000mhz vs 4300mhz
CSS VST : 223fps At 3000mhz 1920x1200 4xAA 8xAF maxed settings
224fps At 4300mhz

COH was nearly identical my average fps differed only .7fps between 3000mhz and 4300mhz

So 9800gtx sli here i come
 
Those Everest benchmark numbers are way off for your system at your current overclock. It all points to your system having some sort of conflict due to your current settings. No way should you see those numbers.

For reference, this is my Everest result with DDR2 ram on my 780i.

everest3.jpg


So with your DDR3 1333 ram OC'd to 1600 with e8400 OC'd to 4.3ghz, you should be getting better results than you currently are.

You may want to downclock a bit to see if you are creating a conflict due to high OC for your CPU and RAM. Try these voltage settings for 4.0ghz OC:

CPU: 1.35 (this can be less, I'm sure, but set it to this for now)
FSB: 1.2
SSP: 1.45
MCP: 1.55
Memory: 1.6

Run your memory linked and sync'd to FSB. In BIOS, set your "FSB - Memory Clock Mode" to "Linked" and "FSB - Memory Ratio" to "Sync".

You can set your memory timings to auto or manually configure them to your current specs. I do advise running them in 2T for now.

Run Everest again and see how this works out for you.
 
alright so the memory timings are 9 9 9 24 1t 5 40 12 18 25 7.7us
so i guess im gunna mess around with the later half of the timings and see if slacking them off a bit makes a diffrence in bandwidth, as it might be "choking" the bandwidth meaning theyre set to low for 9 9 9 24 1t @ 1636mhz and thats whats causing the decrease im not sure
 
So another test

4300mhz cpu (all Voltages remain the same in both tests)

4300cpu 1636memory 9 9 9 24 2t auto timings the other half 5 41 12 19 26 7.7us
read : 7713mb/s
Write : 10038mb/s
copy : 7157mb/s
latency: 83.3ns

4300cpu 1338memory 9 9 9 24 2t Auto timings the other half 5 34 12 18 22 7.7us
read : 7139mb/s
Write : 10099mb/s
copy : 6634mb/s
latency: 90.4ns

but for some reason if i decrease the Cpu mhz

3000cpu 1338memory 9 9 9 24 2t Auto timings the other half 5 40 12 18 22 7.7us
read : 9104mb/s
Write : 7071mb/s
copy : 7103mb/s
latency: 65.1ns

So im begining to wonder if maybe not enough Cpu voltage, or maybe FSB termination voltage is too low, or maybe some other chipset voltage. any thoughts appriciated at this point.
 
ok so i linked and sync'd
1600mhz for 3600mhz cpu and 1600mhz memory 9 9 9 24 2t auto for the rest
read : 10611mb/s
write : 8490mb/s
copy : 8281mb/s
latency : 56.9 ns

ill be back in a few minutes going to try a divider and see if i still loose a huge amount of bandwidth. id like to beable to run at 4300mhz and keep the ram at around 1600 without loosing that much bandwidth.
 
Ok so Anything other than Linked and sync'd gives me the same kinds of memory bandwidth
tryed a 5:4 divider and it works the same as unlinked 1911 / 1638 so i guess im stuck with that shoddy bandwidth. cause im totaly not dishing out 300$+ a kit for ddr3 2000 since anything over 1600mhz isnt supported in 4x2gb configs on these boards.
 
Scratch what I said about linking and sync'ing at 4ghz since that would have given you 1777mhz OC for your memory. I suck at math, lol. I was thinking x10 multi for your cpu but e8400 has only x9.

Your results at 3.6ghz linked and sync'd looks like where they should be.

Ok, now keep your FSB to memory linked and sync'd at 3.6ghz and tighten up your memory timings. Try 8 8 8 24 2T (rest on auto). Keep your memory voltage to 1.6v. Run everest after this and see how much your score improves by. You may want to run memtest to see if it will run stable with these tighter timings at 1600mhz. Increase in voltage by a notch or two might be needed as well. There's a possibility that your memory might not run with these timings at all. Revert back to 9 9 9 24 if that is the case.

Once you've tested this out, OC your CPU by 100mhz at a time, keeping your memory unlinked and "auto" at 1600mhz. After every 100mhz increase, run everest and check results. If the numbers improve, go up another 100mhz and test again. Repeat this process until you see negative results in Everest.

Also, keep in mind that you have to increase your SSP voltage to ensure that the 790i's north bridge, which also houses its memory controller, is getting enough juice to handle all the data flowing through. I absolutely and HIGHLY recommend that you raise this to at least 1.45v. It will not kill your 790i north bridge at this voltage (and it's still in the green). If you're worried about temps, you should not be overclocking at all since overclocking WILL result in increase in temps. You can figure out how to control the temps later after you've successfully managed to get your system optimized.

For now, focus on making sure all your components are receiving the voltages they require, especially your northbridge (SSP).
 
ok so i tryed what you said heres what happened

4000mhz cpu unlinked auto memory set to 1638 999242t (known stable)
1.38750v cpu 1.10vfsb 1.45vspp 1.600vmcp 1.600v memory

read 8777
write 9420
copy 7782
latency 72.7ns

4100mhz cpu unlinked auto memory set to 1638 999242t (known stable)
1.38750vcpu 1.10vfsb 1.45vspp 1.600vmcp 1.600v memory

read 7451
write 9636
copy 7034
latency 86.4ns

4300mhz cpu unlinked Auto memory set to 1638 999242t (known stable)
1.38750vcpu 1.10vfsb 1.45vspp 1.600vmcp 1.600v memory

read 7713
write 10038
copy 7157
latency 83.3ns

as i increase the Cpu fsb upwards past 4000mhz the memory bandwidth will start to increase again
but still with the read speed being much lower than the write speed, and very far off from what i read in the normal.
 
another test of link sync
1600fsb 3600mhz cpu 1600mhz memory
1.38750vcpu 1.15vfsb 1.60v memory 1.35vspp 1.550vmcp
Read 10515
write 8499
copy 8282
latency 58.1ns

1645fsb 3701mhz cpu 1645mhz memory
same as above settings
Read 8351
write 8700
copy 7493
latency 74.1ns

any idea's ill report what 1500mhz fsb does on link sync in a minute
 
Another thing you can try is downclocking the memory from 1638mhz in small increments while you keep the CPU OC'd at 4.3ghz (it is stable at this speed right?). Try downclocking 25-30mhz and test with Everest each time. What you are trying to find here is a FSB to Memory divider that your system agrees with since it does not like the divider that it is forced to use with FSB 1911:RAM 1638. I'm making an informed assumption here so this may or may not work.

Although it may have passed memtest (that is if you did actually test it at 1600mhz), it could be your memory not agreeing with the OC to 1600+ since it is only rated for 1333.
 
its passed just about 20hours of prime and 13hours of memtest at those settings
im just about to wrap up a huge amount of testing tomorrow i will scan all of handwritten notes and post them for you to look over iv tryed just about every setting from link sync to
setting the memory to 1333 and raising the cpu from 1333-1645 which all of my tests show that anything over 1600mhz cpu bus it gets into that weird range of memory readings.
weather i put it at 1645memory or 1333 memory as soon as my cpu speed passes 3600mhz the memory bandwidth goes to hell.

Im not sure if it has to do with running 4x2gb sticks or just having all the memory banks populated i will repeat my tests tomarrow using the same settings i did in all the previous tests, except im going to try only 2 sticks, then only 1 stick of ram
to see if its the fact that im fully populated on all banks.
 
Have you tried keeping the memory locked at 1600mhz and only overclocking CPU starting from 3.6ghz? For example, set your FSB to 1644 and keep your memory on "auto" at 1600mhz. Your memory frequency will fluctuate in BIOS when you start moving your FSB up. It may show something lower than 1600mhz. This is fine, let the system do its thing. Just set your memory initially at 1600mhz and FSB to 1600mhz, and ONLY increase the FSB and DO NOT touch the memory frequency.

If the outcome show positive results, repeat this process by increasing the FSB by additional 44-45mhz and test with Everest.
 
After re-reading your last comment, it could possibly be the issue of having all the slots populated with 4x2GB ram. Running only two sticks might work out better, but this is only speculation at this point. Give it a shot and see how it goes.
 
Alright i was working on this from about 5 in the afternoon to 1 this morning, i tryed almost everything i could think of, i even upgraded the bios from P06 to P07, and pretty much
the memory scales upwards untill i pass beyond 1600mhz Qdr Fsb on the cpu. i can set the memory to 1645 and start at 1333 fsb and go up in any ratio till i hit 1601fsb on the cpu then the memory bandwidth gets funky.

I also tryed 2x2gb and 1x2gb using the same settings and it was no different
tryed over volting the ram, tryed increasing the spp volts, mcp volts, so far nothing has made the memory bandwidth what is is at 1600mhz cpu or less

i tryed setting the memory to 1645 started at 1333 cpu worked up and the memory scales till i hit 1601cpu

tryed setting the memory to 1333 and started at 1333 cpu worked up and the memory scales till it hit 1601cpu

tryed linksync it works but once i hit over 1601cpu the memory does the exact same thing it did in any other tests iv done


On a side note, i dont think its an isolated issue i caught a piece of a review on memory bandwidth with the 790i and his graphs looked 100% like mine would, but im wondering if maybe we just have junk memory controllers.

i mean aside from the memory bandwidth being screwy past 1601mhz i havent had any issues with stability, plays games and does prime for 20 hours memtest stable.

iv just never seen anything like this were increasing the fsb would bottleneck the ram regaurdless of ratio, voltage, or any other setting.
 
when i said i was setting the memory i was refering to unlinked and physically setting the memory limit to 1333 or 1645 both of which i know do mem test at those speeds and timings.. kinda aggravating tho if i knew id get terrible memory bandwidth over 1600mhz cpufsb id have gone for a 780i and some ddr2 and saved a few bucks. because in all seriousness my old opteron 170 with just ddr500 accualy pushed better read speeds than my ddr3 1333 @ 1645, if i try to get the most potential out of my cpu.
it wouldnt bother me so much if i could get it to do maybe 4000mhz cpu and get decent memory b/w but having to clock down 700mhz from my known stable overclock to maintain a decent amount of memory bandwidth is insanity :mad: :mad: :mad:
 
Sorry, there's a lot of walls of text in this thread and it's making my head hurt :)

So when you aren't running a 1:1 ratio your memory speeds are crap? So, run a 1:1.

I also noticed you are running some HIGH latency on your memory, I'm not experienced with DDR3, is it normal to be that high?
 
9 9 9 24 1t isnt exactly high at ddr 1600 speeds, i mean im sure that Corsair or some other brand can do better timings than that at higher speeds but i wasnt and am still not about to buy a 300$ 2x2gb kit, and at this point i dont think it would matter, because it doesnt matter what speed i run the memory at as long as my Cpu FSb is at or below 1600mhz and it doesnt have to be 1:1 can be any ratio as long as the cpu fsb is below 1600mhz
 
Evga has just released P08 BIOS for the 790i. Check out their forums for download. It appears that the new BIOS improves memory compatibility and better supports memory running at 1600mhz. Can't say that the new BIOS will actually help you or not, but you never know. Also, you can find some really good information on their forums regarding 790i related (or any nvidia) issues. They have many folks who are well versed in nvidia system tuning. I highly encourage utilizing their forums.
 
One thing to consider is whether or not if you will actually notice a difference running your CPU overclocked but with reduced memory efficiency/bandwidth compared to running your system linked and sync'd at a lower overclock. Since you cannot run your system fully optimized due to whatever the reason may be (memory controller, the memory itself, BIOS, and etc.), choosing the setting that provides the best results in real world applications is probably be the best course of action until you can find the solution to your problem.

Couple of things you can try on each system setting:
- Run some games and use FRAPS to monitor FPS.
- Winrar benchmark (found under Tools -> Benchmark and hardware test)

Believe it or not, the Winrar benchmark, provides good real time performance data, which shows you how optimized your current system overclock (especially memory OC) actually is, in the most simplistic form - how many KB of data it can process per second. Better optimized your system is, better this number will be.

Another thing I like to observe is windows startup time, particularly the number of passes the windows startup bar makes. From my observation, not having memory at its optimal efficiency/bandwidth directly had an effect on this.
 
Alright all sorry for no updates, iv looked all over the place for information on this Problem, and havent found anyone with issues like this, so im guessing i either have a bad board, or My memory is to blame.

But from what iv read, it apears that the 790i does not support 2gb sticks? at over 1600mhz but that doesnt seem to be the problem. as like i said before the memory bandwidth scales just fine as long as i dont push the CPUfsb beyond 1600mhz.

so basicly it would seem i could run any memory speed
just not any cpu Fsb.

i installed P08 aswell to no avail, several windows reinstalls for unrelated issues no difference. even in Memtest 1600 linked and synced gives ~5160Mb/s
if i push 1601 or anything higher i drop off to 4700~ or less
 
You must log in or register to reply here.
Back
Top