*Official* NBR Desktop Overclocker's Lounge [laptop owners welcome, too]

Everyone on overclock.net said I’d never get my memory stable at even XMP of CL15 4000Mhz 1.5V. But, it’s running just fine. Already passed bootable memtest 86 just fine.

 

Everyone at overclock.net doesn't know what they're talking about. Only a few do. I had no doubt you could do that with good RAM sticks because I'd already done the same with the Corsair sticks. The people at overclock.net that were paying attention saw it. But, I doubt they were the ones saying you couldn't.

Congrats. It takes good RAM, good motherboard and good CPU to do that. The people that have tried and can't are probably missing one of those things. Take one of those away, and it just ain't gonna happen.

 

Ive got it at 4000Mhz C15-15-15-36 1T. It doesn’t even break a sweat.

 

Great!

How about some Aida64 runs to show us what they put out?

 

BAH! $1000 or $1200 for a 20GB 3080, that wouldn't have interested me, and I doubt I'd need that....seems too much dosh for that anyway. People can buy what they want though.

 

Coming soon. I lost my entire OS with the dead M.2. So, I am getting my windows setup right with this, drivers etc. trying to recover some files too. I’ve only done memtest for stability so far.

 

Too bad they don't offer the same sticks in 16GB x 2 kits, with solid black heat sink and no RGB. If they did I might want to buy some.

 

Ok, first run. I will buy a Aida64 key for next time. not sure if this is good or bad. I am going to try and tweak the most out of it though.

 

GDDR6X is both expensive and power hungry. Because of that, I wouldn't expect the 20GB variant for much less than $1K. But we'll see considering that release will be a reaction to AMD's release.

 

So I think my numbers are some what high. Shouldn’t my latency be around 49?

 

I dropped the timmings some tRAS, and tRFC.


Testing 15-16-16-28-360-1T

Massive latency drop of 10ns from using

15-16-16-36-700-1T

 

That’s a high graphics score for a 2080Ti!

 

I am trying to get numbers similar to you guys, my numbers are pretty low. I have managed to get my latency a little better. But, my L3 copy, and write speeds from the CPU are about half as fast as other people running much higher timmings. My standard memory read / write speeds are also pretty low too.

My timmings are 15/15/15/28/1T/290.



Any ideas? Other higher scores I see are running X32 or X33 Mesh frequency. I am only at X30. So, I have lowered my CPU frequency and started back to the basics trying to figure out what is causing these lower L3 cache numbers.

 

That is a direct result of the higher core clock (and VRAM clock to some extent) that is only possible with the shunt mod, firmware with higher voltage, and most of all colder temperatures to stop the silly room temperature thermal throttling crap.

Try tightening up your tFAW and setting tREFI to the maximum value allowed. What is your voltage set at? Even though it is stable, if it is set lower than ideal the RAM may not be performing optimally. If it is 1.450V or less, try bumping it to 1.500V and see if that helps. I ran more than 1.600V for extreme memory overclocking and there is nothing unusual about doing that.

Also, it might be the BIOS version you are running. I found BIOS v1.23 to be crap. BIOS v1.20 was a lot better overall even though v1.23 has a couple of nice features that were newly introduced. I love the "Always show boot option" feature, but it wasn't enough to make me move to v1.23 because the performance hit was evident.

 

Hey thanks! I only run 1.5V that is default XMP voltage for this memory, and I will look in to bios 1.20.

My tREFI is at 15,600. And tFAW is lowered to 44 from 48.

OK, I will adjust these even further!

 

Happy weekend to you all New round with new benches for @Prema Mod Team.

https://hwbot.org/submission/455959...hd_graphics_620_(mobile_kaby_lake)_4152_marks


https://hwbot.org/submission/455959...hd_graphics_620_(mobile_kaby_lake)_2997_marks


https://hwbot.org/submission/455967...raphics_620_(mobile_kaby_lake)_2062.65_frames


https://hwbot.org/submission/455967...raphics_620_(mobile_kaby_lake)_8093.86_frames

https://hwbot.org/submission/455967...raphics_620_(mobile_kaby_lake)_4940.38_frames

https://hwbot.org/submission/455967...raphics_620_(mobile_kaby_lake)_3065.05_frames

https://hwbot.org/submission/455966...raphics_620_(mobile_kaby_lake)_1690.56_frames


https://hwbot.org/submission/455962...graphics_620_(mobile_kaby_lake)_25515_3dmarks


https://hwbot.org/submission/455961...graphics_620_(mobile_kaby_lake)_20506_3dmarks


https://hwbot.org/submission/455961...graphics_620_(mobile_kaby_lake)_13832_3dmarks

 

It is so nice having a good memory set, like when you adjust timmings it doesn’t “Black screen” lol. I’ve never experienced this in my life.

 

THANKS!! I just hit 122GBPS!! Stock mesh, stock cpu.

 

Ok! This is much better! Crazy how these (2) timmings could make such a huge difference.

 

Rip msrp cheaper 3080 cards! Sure to electronic filtering issues, 3080 cards are crashing and having issues. Igor of Igor's lab covered it this morning.

 

Happy weekend to you also, our crazy overclocking Norweigian Brother. You're like a benching machine, man.

Great job! Yeah, those two settings can make a big difference. If you didn't already try it, goose the XMP default voltage a bit and see if the speed and latency improve or diminish. Since default is 1.500V try like 1.520V or 1.525V. That often makes memory and CPUs run a little faster.

 

Not quite. The cards that are crashing are the cheap Palit, Gigabyte and Zotac and it does appear they may have cheaped out or not have recalled their cards in time to fix.

My EVGA XC3 hasn’t crashed once while gaming as it has 5 SPs and 1 MLCC. FTW3 changed from 6 SP to 4 SP and 2 MLCC.

Asus TUF $699 has 6 MLCC as does the premium Strix line.

All Nvidia ref cards have 2 MLCC and 4 SP.

I believe MSI went 1 and 5 route.

but possibly rip those other brands until they swap/replace the faulty cards.

 

And the question is what spec did nvidia give them...

 

Yes definitely a question that will likely get answered in time. For now..

Hi all,

Recently there has been some discussion about the EVGA GeForce RTX 3080 series.

During our mass production QC testing we discovered a full 6 POSCAPs solution cannot pass the real world applications testing. It took almost a week of R&D effort to find the cause and reduce the POSCAPs to 4 and add 20 MLCC caps prior to shipping production boards, this is why the EVGA GeForce RTX 3080 FTW3 series was delayed at launch. There were no 6 POSCAP production EVGA GeForce RTX 3080 FTW3 boards shipped.

But, due to the time crunch, some of the reviewers were sent a pre-production version with 6 POSCAP’s, we are working with those reviewers directly to replace their boards with production versions.
EVGA GeForce RTX 3080 XC3 series with 5 POSCAPs + 10 MLCC solution is matched with the XC3 spec without issues.

Also note that we have updated the product pictures at EVGA.com to reflect the production components that shipped to gamers and enthusiasts since day 1 of product launch.
Once you receive the card you can compare for yourself, EVGA stands behind its products!

Thanks
EVGA

https://forums.evga.com/Message-about-EVGA-GeForce-RTX-3080-POSCAPs-m3095238.aspx

 

I run my set at 14-15-15-23-250-1T W/ 1.550 volts.

I am happy with 47.9 NS latency, and 124GB read bandwidth.

At only 4,000Mhz DDR4 speed.

I love love love this memory! It runs at literally whatever I type in, 14-14-14? No problem! 13-14-14? Let’s do it! No more BLACK SCREENS! And I haven’t even seen a single BSOD yet. I have not seen a gain beyond 14-15-15 timmings. I am still tinkering, and learning.

No heavy “Designated memory stability testing” has been done. But, I have run some strenuous benchmarks and stress test without issues. So, I guess it’s stable?

 

Buildzoid coverage of the nvidia issue

 

How’s the Aida 64 memory stress test good enough? How long 3 minutes good? Lol jk.

2 hours enough?

 

4700 CL18...

 

I let my DDR4 4000 memory run this overnight. Just to make sure it is stable this is at the default XMP profile. Almost 12 hours.

 

Memory stability testing is horrendous guys! It is so time consuming. This is gonna take a few days to tighten all of my timmings for maximum performance, with maximum stability.

I do like how the gold matches the X299 Dark theme.

These are some really flashy memory sticks, but they blend right in to the X299 Dark motherboards due to the “Gold motherboard theme already”

The RGB is blinding bright. I leave the RGB off, because I feel like it could maybe cause extra heat? Maybe not, I dunno! I know that with the RGB off, the memory sticks look like they are made this way, and they don’t even have RGB as an option. I really like the look, and they have grown on me.

I especially like the gold mixed with the EVGA Dark board with memory RGB off. Oh and, I am running another stress test. My idle package temps are not 50C lol.

 

So, I guess 1.6 Volts to the DDR4 is totally fine then.

 

Well, I wouldn't necessarily say that you should use it for your "daily driver" settings, although I do not believe that doing so is harmful to the RAM sticks or anything else. Serious memory overclocking usually occurs at a voltage level much higher than 1.600V. But, the value in running your memory clocks and voltage that far beyond more conservative specifications may not have a practical benefit.

 

Well, it runs 1.508V to both sticks default XMP profile. So it must be ok or G.Skill certainly wouldn’t warranty it. These sticks get pretty toasty during heavy loads too. This is the most stable ram I have ever owned in my life. It just runs! Maybe having matching sticks for quad channel is important after all. Because each ram stick is even serialized in perfect order like 0078,0079,0080,0081.

 

Yes, that is all true. I run my RAM at 1.500V basically perpetually with no issues and I have benched the crap out of it at 1.650V with no issues. And, people like Lumi and Kingpin run it a 2.100V even. G.Skill wouldn't be selling you a product that is going to kill itself running exactly as expected. Neither would their competitors. I think the really crazy high memory voltage actually puts more stress on the CPU and memory controller than it does the memory sticks.

And, yes... having matched stick is always advisable. If you buy sticks individually they can have differences in batch quality and SPD firmware. Unmatched sticks are not necessarily a guarantee of having problems, but avoiding problems by not doing things to create them is a smarter approach.

 

Anyone familiar with Raid 0 VROC performance? I am running (2) Intel 660P's in VROC Raid. But, I feel like my performance is similar to one drive? Maybe my write speeds are better. I am not sure.

They are at about 42% capacity.

Either way, I am ok with this performance. I really only wanted VROC RAID 0 to double my capacity.

 

I've never used it. I used to run everything in RAID until NVMe became the thing. RAID0 made things faster and it was awesome to create a larger storage volume for things like my massive Steam library. The advent of NVMe speed (even the slowest NVMe drives are very fast in comparison) and 1TB being like totally normal/nothing special drive size changed the need for it because the reasons kind of went away. NVMe and 1TB being the norm were also close in time to the introduction of Windows 10. (Funny how great NVMe and large SSD drives are, and how not-so-great Windows 10 is.)

Plus, the kind of overclocking I do tears up Windows real bad and dealing with RAID just got to be too much of an extra burden when I no longer had a good reason to use RAID. I'm frequently restoring a Macrium Reflect image because of overclocking-induced OS corruption, especially Windows 10. It tends to be more fragile than Windows 7. The way I multi-boot 3 versions of Windows 10, Windows 7 and Linux also creates some extra issues and RAID makes that more complicated and fragile. And, another thing that really sucks is having Intel bloat like RST and XTU and IME components installed. All they do is bog down Windows with extra trash Services and background process bloat that lowers benchmark scores. They're also an added layer of stupid digital crap that is constantly nagging you about updates and other similar nonsense. Life is just so much better without all the extra junk running in the background, simply not having RAID is a more than welcomed trade-off to me.

Do I like RAID0? Yup, I love it. I just don't love all the other silly stuff that I have to put up with. Kind of like a sassy-mouth rebellious punk teenager that knows more than mom and dad and is always getting into some kind of mischief or causing somebody some grief... you just wish they would go away and stay away until they can get their act together and act like a normal person. You still love them to pieces, probably even willing to die for them if you needed to. You know (because you used to be one) that someday their stupidity will eventually wear off, but you as sure as hell don't like putting up with their idiotic behavior, LOL.

So, long story short... nope.

 

Well, I’ll say this. VROC is easy. You don’t need any IRSTe or ME this or that. It’s a virtual raid on the CPU, and it’s running straight of the CPU lanes so a lot less latency VS. running off of the PCH chipset. I only downloaded IRST to show someone that it was indeed running VROC and it was running “Without the required $150 dollar intel VROC key” IRSTe is already gone before posting this lol.

Anyways, I got these drives for dirt cheap, and brand new. Cheaper than a single 1TB 660P. And it seems to be faster than a single 1TB 660P.

I am new to raid. And it is silly raiding NVME’s lol. It just isn’t necessary at all. I just wanted the capacity. I do like it so far. I think it is cool.

One thing about VROC Raid is, due to my overclocking nature and constant tinkering, I reset my bios quite a bit. Which means the VROC virtual CPU raid disappears. So, I have to turn it back on in the bios after a bios reset. Not a huge hassle, but just an extra step. Kinda like having to go back and set your fans and pump to MAX again.

 

So, it works similar (in some respects) to a RAM drive?

 

Depends what you paid. You can get a single 1TB drive for $119 with higher perfomance than your VROC Raid0 setup. And please, don't look you blind on the highest numbers in the benchmarks Not so sure you will see them in your day to day tasks

 

I got (3) 512GB 660P’s for $120, which is decent. nothing crazy. I know there are faster drives though for sure. These are working pretty good so far though. Sustained write performance seems to be much better than a single M.2, although read speeds may be slower. Since I am on X299, I will eventually try (4) in Raid 0. Just for fun of course. This requires (16) pci-e lanes, something I couldn’t do on my last platform.

 

Honestly, I am not sure. Not very familiar with the ram drives. From what I understand you can only use Intel SSD’s. And it requires a special hardware VROC key that plugs directly in to the motherboard. I didn’t need it. And somehow the X299 Dark allowed me to set it up either way.

 

https://www.anandtech.com/show/1243...part-1-virtual-raid-on-cpu-vroc-scalability/2

Intel's Virtual RAID on CPU (VROC) feature is a software RAID system that allows for bootable NVMe RAID arrays.

Does using Windows RAID affect CPU utilization?
There is some CPU overhead involved in using Windows RAID, but in our testing we found it to put a smaller load on the processor than Intel's VROC. With most modern CPUs it should have very little if any discernable impact on overall performance.
https://www.pugetsystems.com/labs/a...erformance-in-Windows-10-Pro-1369/#Conclusion

Not a solution you want if you benching after points.

 

Roman and his results of shunt modding and volt mod.

 

Perfect. Now we just need to figure out which vendor and model vBIOS is going to be the one that allows the GPU to get a lot more voltage than the rest. 2100MHz on air at 50°C is pretty decent and very similar to Turing. Core overclocking doesn't seem like it is going to be remarkably better than 2080 Ti, but performance gains could be more significant from improvements in things other than core clock speed. Too bad the voltage boosting can't be as ridiculously simple as the shunt mod. A shunt mod without a voltage increases doesn't do much in the grand scheme of things. Both are equally important.

I may wait until more info/detail is available on the voltage control before jumping in on this generation. 3090 is too expensive to buy with the idea to use it with stock voltage and power limits and 3080 doesn't really interest me very much. I also want to see if the Green Goblin is holding back something for later. It would really suck to drop a stupid amount of money on a 3090 only to find out that a 3090 Ti was lurking in the shadows that nobody knew about.

 

If the rumor and leaked screens screenshot are true regarding the Quadro 6000 ampere, the 3090 is the highest. In that leak, the supposed Cuda core count was around 2.x% higher (i mentioned this in a post above, cannot remember the exact percent off the top of my head). So, if that is true, there is no more room for a 3090 Ti on core count. That card did, however, have 48GB of memory. The Quadro 8000 would likely be a GA100 die, which we haven't had a 100 die for consumers since the 580 days (unless you count a 110 die like on the 780 Ti as a 100 die).

But I hear ya on waiting to see...

 

That's the problem with these cards that are reaching super high clocks, whilst a 100Mhz increase might just about see like a good overclock on the surface of things it's only a 5% increase in core clock if you go from 2000Mhz to 2100Mhz, so these are only minimal gains, but it could be fun to get there. I mean once stock cards can hit 2000Mhz with an easy/basic software overclock and no mods, then I think you really need to be hitting 2200Mhz+ t0 start thinking about significant gains, and even then sometimes performance won't scale linearly with clock speed depending what else is happening in the pipeline of the card in terms of bottlenecks. Back in the day overclockers could get significant performance gains, but it's decreasing all the time now.....I think I managed 14% on my GTX 1070 (IIRC), 80% IIRC on my laptop 670MX, and about 70% IIRC on my 8600M GT which was my first 'recent' GPU going back to a laptop in 2008 which was my first time I overclocked! (I've had computers since BBC Micro times, but only started overclocking them in 2008).

 

I wish there were an easy way to push 1.250V-1.300V to the core of this 2080 Ti. Then I'd probably have no interest in wasting money on something newer because it wouldn't be as good or better (at least for overclocking). I can only get to 2250MHz stable (when I can keep it cold enough) with the chintzy 1.125V cap using the highest available voltage from any vBIOS. I'm thinking I could likely wring at least another 150-200MHz from it. I'm just sick of everything being so mickey mouse and half-assed, even in the enthusiast realm.