MB OC Report :: EVGA 790i Reference

**3oh6** · 07-18-2008, 08:35 AM

I do them for memory so why not motherboards...right? So here we go with my initial MB OC Report. If you are un-familiar with my OC Reports, learn how to read, I post them bloody everywhere. For those that are familiar with the OC Reports of my past, then you will know that this is going to be half review, half overclocking results thread. Let's get started with a brief overview of the NVIDIA nForce® 790i Ultra SLI® chipset and this motherboard for those that have been living in a cave eating dung beetles for the past few months.

The NVIDIA nForce® 790i Ultra is the latest chipset released from NVIDIA® in a long line of disappointing Intel chipsets. Okay so I may be a bit critical with that statement but...it's true. It was long considered that Intel processors were only going to be able to run well on Intel chipsets for us enthusiasts, that changed drastically with the NVIDIA nForce® 680i SLI® but NVIDIA® still had a ways to go. NVIDIA nForce® 780i SLI® was recently released to much fan fare with tri-SLI® support and better FSB clocking but there was still a lot to be left desired.

This is where things get exciting, the nForce® 790i Ultra chipset is suppose to be everything we had been hoping for all these years. Finally we get a chipset that can clock like the Intel chipsets but supports SLI®. Early reports are that this chipset is definitely living up to the hype but there are of course some potential draw backs...*cough* data corruption *cough*. Either way, it finally looks like NVIDIA® has cooked up an Intel ass whoppin casserole and tonight we dine on this creation.

Directory:

**3oh6** · 07-18-2008, 08:36 AM

This is the boring part but like I said, for the dung beetle eaters with their head in a hole, here are the specifications of the NVIDIA nForce 790i Ultra SLI chipset...and yes, this is stolen directly from the NVIDIA web site. It's only borrowing if you give credit though, so you can find all of this information and more at NVIDIA.com.

One thing I wanted to find out was if the firewire port was 400 or 800 but that seems to be privileged information as I can't find it anywhere. Since the EVGA board is simply the reference board with an EVGA sticker on it, there are no differences what so ever from the NVIDIA board. This is both a good thing and somewhat bad thing. Since they are the same boards, the BIOS is interchangeable so when NVIDIA updates a BIOS, this EVGA can be updated with it. It also means that the EVGA board is exactly the same as any other manufacturers reference boards such as XFX so users can share experiences and information since they are identical boards. Comparing results from the reference boards to the Asus over-priced, hard drive eating, nForce 790i offering is impossible since the Asus board is completely different.

Enough of my blabbing though, if you want to know any further specifications of the board, go here or here and leave me alone

.

As for pricing...yeah, bring your wallets, credit cards, and cheque books, it is that bad. We start off with Canada because that is where I live and well, it is better than anywhere else...especially the west coast (not sure what that has to do with anything but I have been missing the left coast lately).

Wow, Canadian prices don't suck compared to the USA, they are actually on par. For once we don't get bent over, guess I should pull my pants up...sorry, it's just habit. I personally bought my board at a local retailer because I like to support the local guy...and I am lazy, they are literally right across the street. I can see into the store from my bedroom window, waaaaay too convenient for a guy like me with a hardware addiction. That's right folks, straight up retail sample...no handpicked spoon fed monkey stuff like that Dinos guy gets that overclocks way better than anyone else’s

.

(I'm allowed to have inside jokes in my reports so deal with it)

**3oh6** · 07-18-2008, 08:37 AM

There will only be a handful of photos of the package and contents because they don't really do anything for the performance of the board. This won't be the last time you hear this...if you want to know more, find a review

Included in the package is quite the accessory pack. The list of goodies includes:

rear I/O plate (naturally)
1 x round sleeved IDE cable
1 x round sleeved floppy cable
3 x 4-pin molex > dual 5-pin SATA power connectors
6 x SATA cables
add on PCI brackets (serial / firewire / 4xUSB)
tri-SLI bridge
SLI bridge
45cm NB chipset fan (MGT5012AXF-A10)

This layout set is going to be short and sweet. Showing every single IC close up and all that jazz just isn't doing anything for me at this point. Instead, I will focus on general areas and if anyone wants specific component close-ups, just pipe up and I will add it in...deal? Alright, here we go kids...your NVIDIA 790i Ultra SLI fly over brought to you by the 3oh6 ghetto product photography setup.

The primary landmark of this motherboard, aside from the slick trademark NVIDIA color scheme, is the rather unique chipset and PWM heat sink/heat pipe assembly. The largest component standing square on the top of the northbridge is shrouded in a small ??cm fan that is optionally attached by the user. It is probably recommended to run with the fan since you can control it to create very little to no noise. Heat pipes extend out to engulf the CPU socket area and connect to heat sinks on each of the two banks of the PWM MOSFETs as well as a third down towards the DIMM slots. On the topic of DIMM slots, they are well spaced from the CPU socket area but are jammed quite tight to the front side of the motherboard. The 4 main SATA connectors, IDE connector, and 24-pin ATX power connection are tucked between the DIMM slots and the board’s edge.

Continuing down the front edge of the motherboard we find a pair of SATA connectors at a 90 degree angle as well as a fan header and a LED diagnostic display. The south bridge is also tucked into the top of this image and is easily identified as being designed to avoid interference with the three green PCI-E 16X slots for three way SLI. NVIDIA has really done a bang up job with the layout as everything seems to be in the right place. Along the bottom edge here, we find the CMOS battery which will be easily accessible regardless of the GPU setup as well as the onboard power and reset buttons. Another fan header resides down here too. The PCI-E 16X spacing will allow for double slot coolers, as expected, but not a whole lot more than that. There just isn't enough room on motherboards these days but NVIDIA has done well with the space allotted and the slot configuration.

Waahooo...PS2 connectors for both Keyboard and mouse, damn you Asus! Joking aside, we see a very nice rear I/O panel here with E-SATA, Firewire, coaxial SPDIF and TOSLINK alongside the standard stack of USB connections, 7.1 Channel Azalia (HDA), and pair of gigabit ethernet connections. There isn't much else to be put there except I won't be able to hook this bad boy up to my printer as there is no parallel port...damn. The backside is where a real nice feature is, the entire heat sink/heat pipe assembly is held in place with screws...no stinking pushpins here folks. One drawback is the inclusion of some very small capacitors right at the under-side of the CPU socket. CPU heat sink back plates will make contact and I have no doubt in my mind, some will crush these components. I installed a thin strip of armaflex on my Thermalright Ultra-120 back plate and these are the impressions it has left.

Not cool man...not cool.

**3oh6** · 07-18-2008, 08:39 AM

Updated Air Cooling:
Alright, so we've seen the stock cooler and yeah, it's pretty beefy. To be honest, I think it is more than enough for any but the most extreme of clockers. Without a volt mod for the northbridge, I highly doubt that any additional cooling is really necessary. Either way, that sucker is coming off and my standard chipset cooling combo is going on.

Ahhhh, look at that big nasty greasy fingerprint on the side of the main northbridge heat sink. Man, wipe that stuff off before you photograph it

. This heat pipe setup is pretty much the best that I have seen on a motherboard. It is very solid with the flat heatpipes and un-like the typical Asus heatpipes I am usually pulling off, there is no need for flattening the assembly out for good contact. The bridge between the south bridge and northbridge heat sinks is a big fat and flat heatpipe, a very ingenious method to provide exceptional cooling while maintaining a low profile to allow for them big booty green video cards.

Another item worth mentioning is the amount of thermal paste used. Let's just say there was 'enough', one might say five times more than necessary, but that would be just one person. After finally cleaning each core off, we can see we have the fancy new nForce 790i Ultra SLI on the left and relatively old nForce 570 SLI south bridge on the right. Now that we have the stock heat sinks off, let's put something else on and see how it fits.

I have gone with my two favorites that haven't let me down yet. On the northbridge will be the Noctua NC-U6 and on the south bridge, a Thermalright HR-05-SLI. The first thing that stands out is the simple fact that the HR-05-SLI is in all kinds of the way for any video cards. Well, the top slot will fit a dual slot and the bottom slot will be fine but the middle slot is all but useless...two out of three isn’t bad. The northbridge on the other hand is very nicely spaced and even with the relatively large wingspan of the NC-U6, the first DIMM slot is completely un-impeded. There should also be plenty of room for large CPU coolers.

The last of the alternative cooling will require something on the PWM MOSFETs. I only say this because there are heat sinks on the MOSFETs with the stock cooling and not because the area gets hot because it really doesn't. Either way, I have gone with eight Swiftech MC-14s. They fit perfectly over all components and do more than enough of a job because again, they are not even warm to the touch under heavy load from what I have found thus far. Even without much of a breeze going over them.

This setup will work well but for tri-SLI, I will have to swap that south bridge cooler for a low profile solution like the EverCool EC-VC-RE. Looks like we have come full circle from the good old DFI Ultra-D days. Now if only motherboard makers would go back to well made, active, low-profile south bridge coolers that aren't part of the heat pipe theme park up front; then we wouldn't have to replace it.

Updated Water Cooling:
With air cooling results maxed, which can be seen in the appropriate section below, it was time to move over to water cooling for the SPP (Northbridge). The MCP (Southbridge) doesn't really have a need for additional cooling and for 3-way SLI testing I actually put a P5B-Dlx south bridge heat sink on which seems to be working just fine. Here are a couple photos of the updated water cooling.

The water cooling loop components used are as follows:

Swiftech MCP655
Thermochill PA120.2
Swiftech MCW30
Swiftech MCRES-MICRO

The Swiftech MC-14s are still doing a mighty fine job so they are staying put. Thermalright has announced a new set of PWM heat sinks for this motherboard and I have received a sample from Thermalright. There is only one problem, they don't fit. The one piece at the rear of the motherboard fits and mounts perfectly. The top side, however, doesn't have the ability to mount. Thermalright apparently has overlooked the fact that the hole layout for the two banks is different. I am working with Thermalright in hopes of getting it sorted. I am hoping that they simply sent the wrong mounting gear but hearing back from them has been rather slow. This section will get updated once it is sorted out.

**3oh6** · 07-18-2008, 08:40 AM

This is obviously not your normal section in a review but hey, this isn't a review dammit. It is my idea of a fancy information and results thread. At this point, all of the volt mods for this board aren't really 100% confirmed but that is simply because I am a little busy with all kinds of things so I haven't tested them all. As time goes by and I get off my butt and test the mods, I will post them here. Until then, feel free to hit up Dinos22s thread over here where there is a bit of a start on all the vmods for this board including what I don't have posted here.

vDROOP
It was posted that a 50k ohm VR was needed for this mod so that is what I used but you could likely use a 20K or even 10K ohm VR without issue as it is tuned to almost 0 resistance to eliminate vDROOP. As it stands, tuning the circuit down to about 600 ohm did the trick to stabilize the vCORE from idle to load. I could also lower resistance even more and raise vCORE at idle and load from what is initially set. You will want really small gauge wire because this one is a tight solder. Rookies need not apply. Here are the photos:

vCORE
I started this out thinking 100K ohms was necessary but a 50K ohm VR is perfect. So please excuse the second 50K ohm VR in the photo that I ran in-line as it wasn't necessary and later removed. This one is very straight forward and the capacitor we solder to is an 0805 so decent size. It is pretty easy compared to the vDROOP mod. You can use any ground, I have used the ground pad of the un-populated transistor behind the CPU fan header.

vSPP
Dinos22 post was again bang on and the soldering point was correct. Measured resistance of the circuit was 776 ohms so I went with a 20K ohm resistor. After the mod, resistance dropped to 745 ohms and voltage was 0.06v higher than selected according to the BIOS monitor. I still don't have a confirmed vSPP read point on the board so I am just going with the BIOS reading as it raises with lowered resistance. Very easy to solder to the IC leg and a perfect mounting point, as seen, alongside the I/O tower. The vSPP mod has definitely helped with FSB clocking and memory clocks but additional cooling is required.

**3oh6** · 07-18-2008, 08:42 AM

"Into the great wide open, under them skies of blue..." as Tom Petty would say, in our case it is the BIOS of blue. The BIOS we are looking at is not the EVGA BIOS, it is the P03R2 NVIDIA reference board BIOS. EVGA has since released P04, P05, P05b, and P06. P04~P05b show no real differences and the only additional options are the inclusion of the P1/P2 options that were already available in P03R2. P06 seems to be a bit different story with stories of bandwidth lowering for users. Perhaps chipset latencies loosened a bit to help with the corruption issues that were apparently fixed in P05 (I think). Either way, P03R2 should look exactly the same as the newer BIOSs so these will not be updated any time soon.

Main Menu // Standard CMOS Features // Advanced BIOS Features

The BIOS is your typical Award BIOS and we all know there is nothing really exciting here so well, yeah, let's leave it at that. The layout is very obvious and all the overclocking resides in one section which is nice. Speaking of the overclocking section...that’s what is up next.

Advanced Chipset Features // System Clocks

As mentioned, everything for overclocking is right here, including the ability to save 3 timing/voltage profiles. I would prefer it saved all the BIOS settings but you can't complain with at least the ability to save the important stuff.

The System Clocks page provides us access with CPU multiplier control as well as the PCI-E and SPP<->MCP frequencies. To get half multipliers simply enable the CPU N/2 Ratio which becomes available when the 8x multi is selected with the E8400. The PCI-E frequency options are 100~200MHz at 1MHz increments. The SPP<-> frequency can be adjust from 200~230MHz in 1 MHz increments then from 230~500MHz in 2MHz increments. The HT multiplier options are 1x~5x for both upstream and down from the SPP to MCP.

FSB & Memory Config

Moving on to the FSB & Memory Config section we are getting into the meat of the overclocking diet. I like how there is a display showing what CPU multiplier we have selected and what the ultimate frequency will be...what I don't like is that NVIDIA continues to use the QDR FSB frequency. We get it, you want to be different from Intel chipsets but come on...quit making me use a calculator in the BIOS. In nTune you list FSB as the actual frequency, why not in the BIOS? Hmmmm? Either way, FSB is adjustable between 400~2800 or 100MHz~700MHz so the BIOS selections won't be holding us back any time soon.

The other major feature of the NVIDIA chipset that we are all well aware of is the fact that we can 'un-link' the memory from the FSB for all kinds of crazy FSB/memory frequency combinations. It isn't a true un-linked memory where we can adjust both independently as ratios are still at the heart of this ability but there is definitely a greater memory/FSB freedom with the NVIDIA chipset than there is the Intel chipset. Of course, we also have standard linked ratios of 5:4, 3:2, and Sync (2:1).

At the very bottom of this page before the Memory Timing Settings we see the elusive P1 and P2 options. These are the much discussed options that appear to be disabled at 475FSB (some users report at 450FSB+) and beyond resulting in horrible memory bandwidth. With the P03R2 beta BIOS and up, we have the option to override these settings and manually keep them enable keeping memory bandwidth and performance where they belong at higher FSB frequencies. This of course puts more stress on the chipset and may require more vSPP or even limit overclocking for some users.

Memory Timing Setting // CPU Configuration // System Voltages

The Memory Timing Setting page is just that, the section for adjusting memory timings. They have the timings broken up into two sections with the main timings at the top and secondaries at the bottom, plus, the timings are in 'typical' order keeping it nice and simple. This will make it less confusing for the rookies to get their uber l33t memory timings set and should lead to less whining in the forums about RAM not running at what it is rated for saving the 'Beards, Tony, Eric, Greg, Michal and others some hassle. The other nice feature is that the column to the right shows what is currently set...very nice to have sometimes. I will mention that the timing list is limited and there is no tRFC available from the BIOS but you 2x2GB owners fear not as this board sets ridiculously high tRFC at all frequencies and Memset allows the tuning of that down for us benchers.

Looking at the CPU Configuration section we can see the typical CPU features listed here for enabling or disabling along with the ability to shut down individual cores of the processor, another nice touch.

The last of the screenshots above is that of our voltage options. Again, nothing fancy going on here with our basic voltages but we do have individual GTL reference voltage adjustments. These range from -155mv ~ +155mv at 5mv increments. The rest of the voltage ranges are as follows...

Code:

CPU Core    = 0.51250v ~ 2.00000v in 0.00625v increments
CPU FSB     = 1.10v ~ 1.45v       in 0.05v increments
Memory      = 1.500v ~ 2.275v     in 0.025v increments
nForce SPP  = 1.30v ~ 1.55v       in 0.05v increments
nForce MCP  = 1.500v ~ 1.750v     in 0.050v increments
GTL Ref	     = -155mv ~ +155mv     in 5mv increments

Integrated Peripherals // Power Management Setup // PnP/PCI Configurations

Yeah, well...what do you want me to say about these sections? Ohhhh, you can enable or disable all of the integrated peripherals. I think the screenshots speak for themselves so I will just go sit in a corner until you finish looking at them.

System Monitor // Dynamic Fan Control

The System Monitor page isn't loaded with temperature readings but we have the basics. What is odd is that newer versions of Everest pick up a couple more like SPP/MCP temp in Windows so I'm not sure why they are not listed here. As you will see, Everest picks up SPP and MCP temps which are very nice to have access to. It also picks up vDIMM and the rest of the voltages and since v4.50.1383 of Everest, the vDIMM reading as well as all other voltage readings appear to be bang on. Older versions of Everest provide an elevated vDIMM reading. The one interesting line here is the CPU Current which shows the Amps that the CPU is drawing. I am not sure how accurate it is but it certainly scales with clocks and vCORE.

Dynamic Fan Control gives us full control over the CPU, Chassis, chipset, and auxiliary fan speed. The options available are pretty straight forward but having this control from the BIOS without the need for additional software to be installed in Windows is always a welcome feature for 24/7 users. That wraps up this boat ride so...like..."good trip eh?".

**3oh6** · 07-18-2008, 08:44 AM

Before we get to any results this is the system I will be using for the initial air cooling results in this thread. Once going sub-zero nothing will change except for the cooling...naturally.

MB: EVGA 790i (BIOS P03R2)
CPU: Intel C2D E8400 (L746A381)
CPU Cooling: Thermalright Ultra-120
GPU: HIS HD3870X2
NB Cooling: Noctua NC-U6
SB Cooling: Thermalright HR-05-SLI
PWM Cooling: Swiftech MC-14s
PSU: Silverstone Zeus 560W
HD: Seagate SATAII 80GB 8MB NCQ
OS: Windows XP Pro SP2

Finally, time to show what this bad boy can do...or bad girl I guess. It's just that with such a dark PCB I would hesitate to consider this motherboard to be female. Does anyone know where you can check the sex on this thing? It's not in the manual.

Initial FSB on Air w/E8400
Before we begin, I just want to mention that the highest FSB this particular CPU was able to achieve with the Maximus Extreme or P5K3-Dlx was right around 535MHz for validation and 530MHz for benching on air. That sure isn't the case with the 790i board. We start off with the least stable and will work our way down. That means we begin with the highest CPU-Z validation I could get.

Initial Max FSB on Air w/E8400 - CPU-Z Validation :: 575MHz
click for validation page...

This is a great sign to start off with. I also tried with various lower memory ratios and a lower CPU multi to rule them out but was unable to get anything over 575MHz so it definitely seems to be the wall for this chip on air. With that said, the Maximus Extreme or P5K3-Dlx can't touch this board with 45nm C2D processors for FSB clocking. In reality, there is very little effort involved. Some legwork by Dinos22 got the FSB clocking and then some adjustments from his work got the final push. In the next couple screenshots, I will have voltages as well at the GTL Reference Voltages listed that were used for this suicide screen.

Initial Max FSB on Air w/E8400 - SPi 1M :: 567MHz
click for full size...

Next up is max 1M SPi benchable FSB and as expected based on the CPU-Z validation a solid 567MHz was stable enough to get a couple 1M runs off as well as the screenshot. Again, this is quite impressive and should lead to a single 32M SPi benchable FSB of 555MHz+.

Initial Max FSB on Air w/E8400 - SPi 32M :: 558MHz
click for full size...

As suspected this setup was capable of a single 32M SPi run at just under 560MHz FSB. I was kind of hoping for over 560MHz but I guess 558MHz is close enough. If going sub-zero holds true to history, we could be flirting with 600MHz FSB under the single stage phase or dry ice for 32M but it is going to be tight.

Initial Max FSB on Air w/E8400 - OCCT 12 Hour Stable :: 550MHz
click for full size...

Last but not least is complete stability using a very long run of OCCT on the default infinity setting. As we can see, 550MHz had no problem holding up over the length of the night with OCCT hammering away. I can't say enough about this motherboard and what it is not only capable of as far as FSB goes, but what it un-locked in this CPU that was previously held back heavily by a pair of Asus high end motherboards. Keep in mind, this is still without voltage mods and using the Noctua NC-U6 & Thermalright HR-05-SLI for chipset cooling.

Max FSB after vSPP mod w/E8400 (Single Stage Phase Change) - SPi 32M :: 575MHz @ 1.61v w/Noctua NC-U6
click for full size...

Performing the vSPP mod for higher north bridge voltage has definitely helped increase the legs of this motherboard. The only problem is that the temperature is holding back further gains with more voltage. Anything over 1.61v seems to overload the Noctua NC-U6 but with better cooling, as we will see, the board again inches closer to 600FSB for 32M SPi stability.

Notes About FSB Clocking

nTune sucks but it does help sometimes reach higher FSB clocks...when it works
So far, nothing but nTune allows us to adjust FSB within Windows
In order to reach high FSB I had to POST with higher vCORE than necessary for CPU stability, but then lowering vCORE once in Windows never led to stability issues. nTune is useful for some things, such as voltage control in Windows.
As you might have expected, some memory ratios limit FSB clocks and POSTing ability at higher FSB frequencies...but for the most part, you can use what you want where you want it.
Higher SPP voltage did lead to higher FSB clocks, this should continue to scale with additional cooling on the SPP
This is a big one, setting too much vSPP will result in C3/C1 or double booting errors. That is right...too much vSPP causes this. A vSPP volt mod is about the only way around it as you can then POST with whatever voltage the board will allow, then turn the vSPP up before going to Windows.

**3oh6** · 07-18-2008, 08:45 AM

Alright, FSB has us covered up to 1100MHz with this board for memory without having to un-link it. I wish I had time to investigate the entire memory clocking realm of this board and performance hits, if any, there are at the various ratios but life just doesn't allow that. Heck, I have a couple care packages from my motherboard manufacturing nemesis coming in as we speak so I am all but out of time with this board as it is. Needless to say, I have managed a voltage limited Dual 32M SPi OC Report of my latest Corsair Dominator PC3-14400 (non-XMP), but first, a look at highest CPU-Z validated, and 1M stable...

Max RAM Frequency w/Corsair Dominator PC3-14400 - CPU-Z Validation :: 1100MHz
click for validation page...

Naturally it took a bit of effort and vDIMM set to 2.125v (2.1v ish actual) to get the magic 1100MHz validation but I managed. I tried getting Everest but it was a no go so this frequency is extremely un-stable. Once I get some more vmods done to the board and increase vDIMM, I am sure I can get 1100MHz benchable with this kit and others. Heck, loosening timings up could make a world of difference...but 8-7-6 at this frequency would be so much nicer. Mmmmm, 550*9 1:2 1100MHz 8-7-6 with P1/P2 enabled

Max RAM Frequency w/Corsair Dominator PC3-14400 - 1M SPi :: 1086.1MHz
click for full size...

A bit of a drop off from the max CPU-Z and I was actually able to get a 1090MHz 1M done but when trying to run Everest bandwidth the system would reboot. Next up is single 32M and then an abbreviated dual 32M SPi OC Report on this kit of memory.

Max RAM Frequency w/Corsair Dominator PC3-14400 - 32M SPi :: 1070.0MHz
click for full size...

Only 2.153v (actual) was used because more voltage didn't help. I am finding that perhaps the vDIMM circuit on this board is a little weak at higher voltages because there seems to be stability issues going over 2.200v set in the BIOS. It could just be this kit of memory as well so further testing will be needed. I will try a few other kits and if the same results are found, then it will be Maximizer time for this board.

Dual 32M SPi OC Report w/Corsair Dominator PC3-14400
As mentioned, this is an abbreviated session since time is constrained but it gets the general view of how this board clocks memory as far as dual 32M is concerned. Since initially putting this MB OC Report together, I have done a complete OC Report on a kit of Crucial Ballistix PC3-16000 and the results are absolutely ridiculous. You can find that report in my signature at the end of this MB OC Report or right here if you are the impatient type. For now, this was the first memory clocking I did on this board with the Noctua NC-U6/Thermalright HR-05-SLI cooling setup and no volt mods...here is how it shook out:

Dual 32M at 1060MHz with only 2.102v is a mighty fine accomplishment and 960MHz at 7-6-5 is pretty decent as well. One thing to note, however, CL7 stability is far below what we see here. At 8-7-6 I could achieve 100% stability at 1045MHz w/2.10v and 825MHz at 6-6-5, but with 7-6-5, I was only able to squeeze out 917MHz which is quite a deviation from the gap at the other two timings shown above. Hopefully a new BIOS alleviates this issue otherwise some CL7 kits coming out are going to have a nightmare with compatibility with this board if they are rated above 900MHz. I could also have been overlooking something, in which case I get a boot in the ass. Here are the screenshots, from left to right...

6-6-5 @ 830MHz // 7-6-5 @ 960MHz // 8-7-6 @ 1060MHz

**3oh6** · 07-18-2008, 08:46 AM

This is an entirely fresh section of this MB OC Report and it will solely deal with the clocking and compatibility of 2x2GB kits of memory clocking. This seems to be the hottest topic surrounding motherboards these days as a lot of users are going the 4GB route. I will keep a diary of sorts of the work I have done with 2x2GB kits on this board so don't mind the vacancy right now, this section will get some tenants rather soon. Starting off with a little Mushkin Ascent 2x2GB PC3-12800 action...this kit of memory is a review sample sent directly from Mushkin and used in a review for HardwareCanucks.com. If you want to see more results from this memory, including full stability testing, just fire me a PM or ask in this thread.

click for full size...

The Mushkin Ascents are a very solid 2x2GB DDR3 offering I have found, and they absolutely love the 790i, in fact, a little too much. Here is the dual 32M SPi breakdown of what they were able to achieve on this board.

At first glance you may think that I am not done with the clocking of this kit...well, that is complete the opposite. To be truthfully honest, this board is done all together. The initial results up to 2.01v were from when I was still completing the review of this memory. I then came back to them specifically for this section, but after the CL6 testing, in which the memory absolutely took off...the board stopped cooperating. It still runs fine and all, but getting clocks north of 500FSB seems to be an issue now. Further more, CL7 is clocking worse than it ever did. I have taken a break from this motherboard for now in hopes it just needs a rest but I have a feeling this MB OC Report has gotten all of the results it is going to.

Here are the screenshots to accompany the results posted above. In all of the screenshots, voltages and all settings should be visible for those interested:

6-6-5-X 1T / 6-5-5-X 1T:
NVIDIA 790i:
Click link for full size screenshot...
740MHz :: 1.86v | 775MHz :: 1.91v | 800MHz :: 1.96v | 820MHz :: 2.01v | 850MHz :: 2.10v | 880MHz :: 2.20v
As mentioned, at CL6, this memory not only surpassed my PC3-16000 2x1GB kit, but did so with less voltage and continued to scale with voltage as far as I was willing to go with it. At one point, I was ready to solder the Maximizer to the EVGA board just to go further with voltage on this kit but decided against it. At an impressive 880MHz 6-6-5, this kit had no problems running dual 32M and likely could do single 32M at close to 890MHz.

7-6-6-X 1T:
NVIDIA 790i:
Click link for full size screenshot...
875MHz :: 1.86v | 895MHz :: 1.91v | 916MHz :: 1.96v | 930MHz :: 2.01v
These above screenshots weren't origianlly intended for the 32M OC Report of this kit but after the review and the CL6 clocking, the board went south so I wasn't able to re-do the CL7 dual 32M clocking and also wasn't able to push past 2.01v. It is a shame because at the pace this kit was going, 970MHz+ looked to be a possibility at 2.20v or less.

8-7-6-X 1T:
NVIDIA 790i:
Click link for full size screenshot...
975MHz :: 1.86v | 1000MHz :: 1.91v | 1000MHz :: 1.96v | 1010MHz :: 2.01v
It is the same story for the CL8 clocking. I only went up to 2.01v to ensure I didn't damage the memory before the review was finished and the plan was to really push them for this report. We might also be looking at the limit of this particular 790i sample as the gains really slowed down above 1.96v with both 1.91v and 1.96v showing similar clocks, just with tighter timings on the higher volts. Even still, 1000MHz at 8-7-6 with a 2x2GB kit for dual 32M is quite nice, especially for a PC3-12800 kit.

**3oh6** · 07-18-2008, 08:48 AM

Yes, I am cursing in church, walking on someone’s grave, dating the boss's daughter...running an ATI video card in this NVIDIA motherboard. Something that should just never been done, well, I did it and here are the results. Things got switched up a little bit this afternoon. Some 1/2" neoprene was found locally so I had to put it to good use and slapped a Chilly1 SS phase change on the E8400 and took this board for a sub-zero test drive with the top down. First, a shot of the setup which is still with just the Noctua NC-U6/Thermalright HR-05-SLI cooling and no volt mods...yet.

All results are without LOD or any Mip Map tweaking...I NEED LOD adjustments with the HD3870X2 in

I told you that dual slot coolers had no problem with the HR-05-SLI mounted on the MCP. It actually worked out rather well with the HD39870X2 fan acting as an active cooler for the Thermalright chipset cooler, not that it is really required. Enough babbling, on to some results...

AquaMark3 :: 307705 @ 5031MHz 945MHz//1017MHz

3DMark 03 :: 74838 @ 4995MHz 918MHz//1017MHz

3DMark 05 :: 32339 @ 4950MHz 891MHz//999MHz

3DMark 06 :: 21112 @ 4950MHz 918MHz//999MHz

Apparently I didn't save a screenshot of 01, not sure how but, it's not here. Either way, 01 sucks because I still can't get LOD set in XP and well, 01 just sucks on the HD3870X2 so we aren't missing much. As for the rest of the scores, they are all slightly lower than what I achieved on the P5K3-Dlx but that is primarily because the HD3970X2 wasn't clocking near the same. On the P5K3-Dlx I was able to run all benchmarks at 945MHz/1017MHz, that clearly wasn't the case here. Either way, still some great numbers and plenty to be excited about with this setup.

**3oh6** · 07-18-2008, 08:49 AM

2 x BFG 9800GTX OC in SLI
We will start off with some dual 9800GTX action in XP and then move on to a 3-way with Vista. The main reason for this is to get some 3DMark 01 and AM3 run in XP. For all of the benchmarks in this section, CPU cooling will be accomplished via the Chilly1 single stage. Here is the first real session with the CPU frequency hovering around 5GHz and the motherboard still on air cooling and no volt mods...

If you look close, no Thermalright HR-05-SLI. I ended up grabbing the south bridge heat sink off of a P5B-Dlx. Temps are 15C higher but the system seems stable so I am going with it. Here are the results from the first session with un-modded cards...at the time of doing these runs, these were all personal bests. It was quite the exciting afternoon and some of the scores I am still quite proud of.

AquaMark3 :: 314964 @ 5073MHz 841MHz//1240MHz//2050MHz

My personal best AM3 score just kept going up with each session. I apparently had a better mount of the single stage as I was able to bench a good pinch higher than last session. Turns out, it was just a bit of a luckier run as FSB is what was really holding CPU clocks back.

3DMark 01 :: 104615 @ 5022MHz 831MHz//1241MHz//2050MHz

http://www.madonion.com/compare?2k1=9368983
And the personal bests kept falling like no body's business. This I have to admit I was quite proud of. I am not as experienced a 3D bencher as I am SPi but I worked my ass off for 100K and managed it at less than 5GHz...getting almost 105K on a single stage is just the icing on the cake

3DMark 03 :: 88802 @ 5022MHz 820MHz//1251MHz//2050MHz

http://service.futuremark.com/compare?2k3=5820974
Chalk yet another PB up to the list. This score isn't really that impressive but 03 is more geared towards the 3SLI so I am not concerned. Although a pair of 9800GTXs easily crush an HD3870X2 in 03. Of course, adding a third card can't compete with a pair of the HD3870X2s though and the newer 4xxx series super cards blow NVIDIA out of the water here.

3DMark 05 :: 34247 @ 5051MHz 820MHz//1231MHz//2050MHz

http://www.madonion.com/compare?3dm05=4218905
3DMark 05 was another PB at the time for me and another strong benchmark for the 9800GTXs in SLI. All things considered this is a pretty good score for stock cards and only 5051MHz on the CPU clock.

3DMark 06 :: 23657 (PB) @ 5000MHz 820MHz//1231MHz//2050MHz

http://www.madonion.com/compare?3dm06=6457507
Last but not least is 3DMark 06. I admit I didn't spend much time on 06 but when you just have a dual core, it is like fighting an uphill battle on a skateboard. 23K though is about 2K better than the best I could squeeze out with the HD3870X2.

That pretty much wraps up the initial SLI results. There will be further updates later on after volt mods to the cards as well as the board and the additional cooling. If you thought this little E8400 was doing good up to this point with just a single stage strapped to its back, you'll certainly like the next couple of sections

3 x BFG 9800GTX OC in SLI
So the 3-way with the ladies is done. It went much quicker with only 03 to run since 05 and 06 show zero gains from 2 to 3 cards but likely has more to do with Vista being a dog. 03, however, is a completely different story. Now this isn't the highest NVIDIA 03 out there, but it sure is top of the ORB and BOT for 9800GTXs for now...and this is fully on the up and up, no mipmap, just LOD, OS, and hardware tweaking

What a bloody mess...but getting air into the heat sinks is not an easy task. Temps are the biggest issues so I am not sure how much volt mods are going to help the 3-way SLI setup. For just 2 cards they should give a nice boost though.

3DMark 03 :: 108025 (PB) @ 5032MHz 820MHz//1250MHz//2000MHz

http://www.madonion.com/compare?2k3=5821535
So I guess that means I joined two 100K clubs on this day. Sure the HD3870X2s are blowing NVIDIA out of the water in 03 but this 3 card setup puts up a bit of a fight. Either way, top of the BOT and ORB for multi 9800GTXs by a solid 10k is good enough for me

**3oh6** · 07-18-2008, 08:51 AM

This section will house the random benchmarks that I run on this motherboard that have occurred since the volt mods and additional cooling on the motherboard. Up to this point, all results we have seen are from an un-modded board (with the single FSB screen shot as an exception) with only upgrades to the chipset cooling but sticking to air. In all honesty, all of those results were capable on the stock cooling. It wasn't until the vSPP mods happened that the Noctua NC-U6 was even required and then going to water on the SPP is what really opened things up for my board. I start off this section with a little head to head showdown between my best 4GHz 32M run on the Maximus Extreme and the best 4GHz 32M on the 790i. The OS and benching methods are identical as this was to try and showcase just how close the 790i can get to the Intel chipsets clock for clock in SPi 32M.

You all know how this goes down, 4000.0MHz is the limit, 32M SPi is the game, lowest time is the goal. Same OS setup and same tweaks used for both runs. I am sure you are all familiar with my 32M SPi benching process so I won't go into details. The combatants are the Maximus Extreme & of course the EVGA 790i. Here is how it went down, click each time for the full screenshot...

Asus Maximus Extreme
500x8 1:2 1000MHz 7-7-6-18

EVGA 790i Reference
533x7.5 1:2 1066MHz 8-7-6-18

Obviously we have a couple different setups but that is the point of this whole challenge. Use each boards strengths to its advantage. The 790i excels at high memory frequency where the Maximus Extreme was able to run 7-7-6 at 1000MHz. The ME edged the 790i out but it certainly is a tight battle. It is definitely nice knowing that the 790i, although a small margin slower clock for clock with the Intel big boys, is still certainly competitive.

I wanted to do a 5GHz 32M challenge with the P5E3-Premium but my sample of that board was a complete dud and 555*9 wasn't even close to possible. Don't ask me about it but if you want a cheap P5E3-Premium for a quad core, fire me a PM before I make an X48 key chain out of it

PB 32M SPi
I will keep this roped off area to update my 32M personal best time with this motherboard. It seems like this time drops every day with this board as there is always more I can squeeze out of it. The next step is better cooling on the SPP in order to get higher clocks and take advantage of more volts. Either that or a higher multi CPU but until I get an E8600, I am done dumping cash on E8500s that don't even come close to the power of this golden little Q808A E8400 of mine...and yes, 1.56v is accurate for vCORE at damn near 5.3GHz. FSB is the single limit right now from this CPU going much higher...keep in mind, this is still under a single stage at about -25C (if that).

588MHz x 9 5290MHz / 1102MHz @ 8-7-6
click for full size...

I call this run "I strike the empire back..." because I was listening to Mos Def - Black on Both Sides while benching this session. There is now water cooling on the north bridge in the form of a Swiftech MCW30/PA 120.2/DD5/MCRES loop. It has obviously helped and allowed for more vSPP voltage. The maximum 32M FSB went from 575MHz up to 588MHz. At this point I am FSB limited as nothing will get me a 32M higher than 588FSB.

Along with the water-cooling on the SPP, this board also acquired a bit of a useless milestone...15K in Everest bandwidth Read:

Like I said, it's a little bit useless but fun to be the first to do so none the less...at least first that I have seen. You can see the entire thread about this little milestone here.

PB 1M SPi
Like the 32M Personal Best spot above, I will keep my 1M PB updated on this board in this spot here.

596MHz x 9 5362MHz / 1117MHz @ 8-7-6
click for full size...

This setup just kicks and kicks and kicks some more. It is like my old hopped up 2L 8v VW engine that was built for a 10K KM race season and ended up turning out damn near 45K KM at an average of 7K RPM before finally giving out in a smoke show attendees of the event will never forget. Taking the oil out of the engine at 8500RPM might have been the ultimate factor

. Either way, this board just churns out the results every time I fire it up.

2x9800GTX SLI - vGPU Mods Done
This is the section you have all been waiting for, well, maybe not...but it sure is the most fun section of the MB OC Report. I figured I would leave the un-modded results in the other section and post these here. The CPU has also been changed from most of the other screenshots as was mentioned a little higher up. You can easily tell which results are with this CPU as the voltage is staggeringly lower. This new CPU is an Q808A476 E8400 and is an absolute monster on air or under the single stage. It has really helped push my 3D scores to above expected heights on this motherboard. All results in this section are personal bests as you might have expected.

Aquamark 3 :: 328310 @ 586MHz x 9 5272MHz // 880 / 1235 / 2172 // 1099MHz @ 8-7-6
click for full size...

3DMark 01 :: 111026 @ 580MHz x 9 5221MHz // 860 / 1235 / 2122 // 1087MHz @ 8-7-6
click for full size...

http://service.futuremark.com/compare?2k1=9371715

3DMark 05 :: 35578 @ 580MHz x 9 5222MHz // 860 / 1230 / 2122 // 1088MHz @ 8-7-6
click for full size...

http://service.futuremark.com/compare?3dm05=4238851

3DMark 06 :: 24615 @ 579MHz x 9 5216MHz // 850 / 1230 / 2098 // 1086MHz @ 8-7-6
click for full size...

http://service.futuremark.com/compare?3dm06=6577938
The cards are volt modded, the board is volt modded, everything is at the absolute limits that this setup could provide with this cooling...and the results speak for themselves. The idea of almost cracking 25K with a dual core under a single stage would have been laughed at before I got this setup and I didn't think in a million years I would see 110K+ in 01 without LN2 but with the help of a killer 790i motherboard, a monster E8400, and a weeklong lesson in 01 tweaking...it seemed to come rather easily. Hopefully you enjoyed these results as much as I did because I am officially hooked on 3D benching now...wallet, look the **** out!

**3oh6** · 07-18-2008, 08:52 AM

There it is...my first MB OC Report and a decent look at the 790i chipset. Of course there is so much more to cover but I didn't intend for this report to be all encompassing, just to share my results and thoughts on the board. Again, big thanks goes out to Dinos for his initial work with this board before anyone else had it. Makes for getting started real easy when some of the basic leg work is done and done well.

Major Issues
There are of course a few issues that can't go without being discussed. The biggest being OS corruption or hard drive corruption. I personally haven't experienced any issues that couldn't easily be explained as pushing memory too far. With that said, I have loaded more drive images in the past two weeks with this board than I have with all others in the past 2 years...and I re-image a lot. Clearly there is something going on and I suspect a lot of the corruption issues are simply unstable memory or chipset latencies. Recent BIOS updates have alleviated these issues for most users and like I said, I didn't once corrupt an OS that I didn't expect to from pushing memory. I can also still bork an OS in a single boot if I want to. The other negative items on my list are the capacitors shown on the back of the motherboard. They are in terrible spots and back plates will contact them. Contact means grounding or straight up destruction. I am not sure how something like this could have been overlooked and as of yet, no one has been found to have done so in forums. Perhaps I am just being over critical of this design but I still don't like anything being back there.

Thoughts on the BIOS & Overclocking
There have been a lot of BIOS revisions come out since most of the work in this MB OC Report was done but nothing has improved results. P03R2/P04/P05 have all been really good to me and I currently run P05 for all work done on this board. As for overclocking, there are some quirks and oddities with the 790i reference design but nothing that can't be worked around. The ability to clock the FSB with something other than nTune in Windows would be nice but nTune gets the job done most of the time, sometimes, okay...once in a while.

There are some other issues with POST sometimes giving random POST codes hanging and 11 being seen far too often, but a simple reboot followed by save and exit and you are usually good to go. That is the other major plus this board has that seems small, but when clocking and benching, is soooo huge. The BIOS recovery on this thing from a failed POST is perfect. Hit the reset button and bam, back to safe boot settings. No power down, no clear CMOS, nothing, just reset and your back up. Even if you get a partial POST that hangs, hit reset then reset again while at C3 and you will get the safe settings. Can't say as I have had a more enjoyable overclocking experience than with this board.

Overall, the board is solid and really nice to work with. Oh, and BTW, SPi performance is nearly as good as the Intel chipsets thus far. Dinos said it first, I am just confirming. I really like this board and won't hesitate to fan the flames of the haters any chance I get

**Gunslinger** · 07-18-2008, 09:06 AM

That is one of the best, most thorough reviews/reports I've read in ages, very well done.

**mad_skills** · 07-18-2008, 09:11 AM

Great job man, really thorough and complete, same as your ram reviews. Well done.

**h@RRy** · 07-18-2008, 12:48 PM

That was far and away the best review I have seen in my life

Thanks 3,
h

**loonym** · 07-18-2008, 01:42 PM

Wow! Thanks so much for this great work. I'm setting up an xfx 790i and this is wonderful. Incredibly thorough, well thought, and just a fantastic read!

**3oh6** · 07-18-2008, 03:53 PM

wow, thanks guys!

glad you are enjoying the new Report. don't hesitate to ask anything about the board or if there is something specific you want to see. i have been beating the living day lights out of this guy for a while now and am more than happy to help anyway i can. i am working on the 2x2GB Mushkin Ascent ram clocking as we speak.

**Pellepel** · 07-19-2008, 08:40 PM

WOw this was such a great review, I had to finish it in one session

. I am definitely thinking of getting this board now, but then again you mentioned about the backplate issue crushing the capacitor on the back

. Now I'm not sure what I could use as an alternative to my D-Tek Fuzion v1?

Preferably I like to keep my cpu block, because I just want a board to replace my awesome Striker Extreme. Anyhow good job on the review, very thorough and informative!

**~~Zucker2k~~** · 07-19-2008, 09:24 PM

Awesome

**pumbertot** · 07-19-2008, 11:11 PM

can you post the voltages including GTLs for those different overclocks>
would vdroop pencil work on this board?

**3oh6** · 07-20-2008, 06:57 AM

Originally Posted by Pellepel

WOw this was such a great review, I had to finish it in one session

. I am definitely thinking of getting this board now, but then again you mentioned about the backplate issue crushing the capacitor on the back

. Now I'm not sure what I could use as an alternative to my D-Tek Fuzion v1?

thanks Pellepel...as for the backplate crushing components, i haven't seen a single instance of this in any forums yet. i was almost positive this would creep up when i first got this board but apparently the standard backplate that so many use for blocks/heat sinks (like the one for the Ultra-120 int he pics) doesn't cause a problem. the EVGA forums is full of people whining about issues with their boards but not one has said anything about crushed components back there...perhaps they aren't seeing them and that is the source to some of their problems, but not likely.

if you are thinking of jumping into the 790i you definitely want to have a stroll through the EVGA forums as well as maybe XFX if it exists to see what you are getting into. a lot of it is user error and newer OC'rs not knowing what they are doing but there are definitely some issues with this board. i admittedly only bench with the board, but there are some quirks with FSB and voltages that i tried to explain here. SLI has never been a problem for me and with enough work on GTL and FSB voltage, gaming at higher FSBs isn't an issue either. those seem to be the major problems people are having from what i see. with that said, my benching consists of stability testing with games, Prime, OCCT, HCI memtest, and all kinds of SPi and 3DMark work for memory reviews and what not, so it sees pretty close to a regular 24/7 rigs type of work. HTH

Originally Posted by pumbertot

can you post the voltages including GTLs for those different overclocks>
would vdroop pencil work on this board?

voltages for all the SPi SS are there in the monitoring software or a small text file that is open in the SS. the 3D benches there isn't but they are going to be about the same at the various FSBs.

GTLs aren't documented and i couldn't list them all but for the most part i have been really simple with them. -80/-80 up to about 500FSB+/-, -110/-110 up to 550FSB+/-, and -155/-155 above 550FSB. for 24/7 stability, that is a different story and going over to the + side has provided some success for stabilizing COD4 around 450FSB.

apparently the pencil vDROOP doesn't work or is hit & miss at best. have a look here...

http://www.xtremesystems.org/forums/...d.php?t=181677

the thread is getting long and old so maybe start at the back with the newer information. but my guess is that the resistance gets lowered so much that doing it with a pencil is all but impossible because we are lowering the resistance with the VRs to almost zero...which the pencil just can't do, and definitely not with any kind of precision. i actually advise against the pencil mod because you would surely lower resistance to 0 in order to get the vDROOP eliminated and that just can't be good...but i don't really know for sure so take my opinion, and that is all it is, for what it is worth. i am a lazy cat and still haven't taken disconnected the vDROPP VR on my board to find out what it is actually set at to eliminate vDROOP. maybe this week will be the week

**TheGoat Eater** · 07-20-2008, 07:17 AM

Originally Posted by 3oh6

Notes About FSB Clocking
-nTune sucks but it does help sometimes reach higher FSB clocks...when it works
-So far, nothing but nTune allows us to adjust FSB within Windows

-In order to reach high FSB I had to POST with higher vCORE than necessary for CPU stability, but then lowering vCORE once in Windows never led to stability issues. nTune is useful for some things, such as voltage control in Windows.
-As you might have expected, some memory ratios limit FSB clocks and POSTing ability at higher FSB frequencies...but for the most part, you can use what you want where you want it.

As always a great write up - hats off to you for it -learned a lot

Glad to see the posting of the review on this site

- was specing the suggested cooling solution last night though i need a new Pot too so IDK what to get first

My results - hope its ok

Anyway - about quoted text...Very true - with the system tools I did the following - unmodded board and DICE cooled with OCZ D9 JNL

my board is capable of -
E8500 Q807

585 FSB - Super Pi 1M ok
569 FSB - Super Pi 32M and Pi Fast ok
*See here*

565 FSB average for 3D Benching
*See here*

**GAR** · 07-20-2008, 08:14 AM

Very nice review, good job and good write up!!!

**loonym** · 07-20-2008, 11:02 AM

Been playing with an e8400 and ocz flex xlc 2x1gb. I can't keep from corrupting the os , and quite easily entire c drive, unless I set ridiculously loose timings and don't go over fsb 1600. I can run much higher, 560-570fsb easily but it destroys any files and I need to load restore image at next reboot. Are there settings I can use in advance mem timings that can help this? I've been using auto for all but basic timings because I don't know how the others work.

edit: bios is xfx P05

Thread: MB OC Report :: EVGA 790i Reference

Thread Tools

Search Thread

Rate This Thread

Display

MB OC Report :: EVGA 790i Reference

Quick Specifications & Pricing

Package & Layout

Cooling Alternatives

Mmmm...Burning Solder: Voltage Mods & Read Points

Blue Martini's: BIOS Screenshots

Appetizer: FSB Clocking

First Course: 2x1GB Memory Clocking

First Course Continued: 2x2GB Memory Clocking

Second Course Blasphemy: Red Bleeds On Green

Main Course: 2 Turn Tables & A Microphone

Dessert: Random Work & Benches

After Dinner Deep Thoughts...By 3oh6

Tags for this Thread

Bookmarks

Bookmarks

Posting Permissions