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

Sad news... For every one they stop supporting, it'd be nice to have them release new benchmarks that are relavent for today's hardware.

Long live W7 and other good legacy like softwares.

 

I need to get some of those $400 wheels so I can roll out my massive TR testbench outside. Current temps are 15 degrees, aka -9.4C.

Edit: Tonight will be even lower but I have to get my pipes unfrozen. No hot water, hahaha.

 

Epic battle for those with deep (DEEP) pockets. If price was no object, I expected the 3175 to fare worse TBH:

 

I was secretly building a 3175X rig. I actually had the 3175X in hand and was waiting for EVGA to release their dark board. After seeing how well mainstream Ryzen 3000 did, I sold it. I honestly didn't think TRX40 would be this good. I am, however, still going to get a 10980XE when available for MSRP. In addition, I am searching for the original Intel IHS for my 7980XE so I can trade with @Mr. Fox (I didn't forget you brother).

After the 10980XE, I am holding off any more personal major purchases on PC hardware for awhile.

EDIT: I will be placing my 9900K with Asus Maximus XI Extreme Z390 mobo for sale on this forum's marketplace soon. I haven't used it in the past few months.

 

Unless grabbing a Zen 4 in 2021, I would wait for 2022. If Intel's back port of golden cover to 10nm+++ is any good, even with a low to mid-4GHz frequency, then that would be a back and forth in between Zen 4 and Zen 5. Also pcie 5 and ddr5.

Graphics cards would be good to buy in 2020, but waiting for 2022/23 after that would likely be worth it. 2022 would be 5nm, but doesn't have much node improvement over 7nm as some may want. If they get multi-die ready for then, then it would be worth the dive.

Otherwise, multi-die 3nm will be a great upgrade time on graphics, although that is 2023/24 time frame.

Cores and IPC seem it will be the race moving forward, not frequency. This is based on the tiger lake frequencies, which the ES is running slightly lower than the ice lake u chips. That should be higher in final silicon. But it being 3.6-3.8GHz says Intel isn't getting some monstrous 5GHz speed on 10nm. This does not mean their 7nm cannot go higher, but that will not be used for mainstream until 2022 at the earliest, but probably closer to 2023, which is when TSMC will have 3nm in full swing which will have more density than Intel's 7nm, even if it is on 7nm+ or ++ by that point, if they can deliver on their process roadmap, something still not proven.

 

I also forgot to mention Via's joint venture with the Chinese government. China is replacing all government hardware with Chinese made hardware over the next three years. X86 patents expire, including the 64-bit patent and SSE patents, next year. Via is working on a 32-core chiplet CPU on 12nm TSMC in 2021 (with possible I/O die) and a 7nm 8 core with pcie 4.0 support in 2021. So via may get into the race, along with others, by 2023 for cutting edge. Something to keep our eyes out for (especially with China's big pockets helping).

 

I saw the writing on the wall on frequency rates since the 6700k/7700k but I kept going. When I started overclocking, I didn't think I would be battling some of the best in the world. I didn't think I would have about six rigs (most of which are in pieces). As long as Intel/AMD have unlocked SKU's, I'll keep overclocking, even if the gains aren't as great. It's an art and I love it. I am done with mainstream products though.

 

I'm in the same boat... did not think so either. Crazy times in the PC world right now. Good call on getting rid of the 3175X when you did. Looking forward to the 10980XE despite being in the middle of a 3970x build. I have no doubts the 3970x will produce nice numbers, but it's expected given the cores... now as far as the fun factor goes it's going to be hard to beat my 7980XE... Perhaps you feel the same way and get where I'm coming from yea?

I'm thinking of going direct with my 7980XE ...hmmm... might as well.

The crazy nature of technology eh? It's exciting to see into the future, but at the same time there is a value to the cost of ownership. Especially for those that make money with their PC's / hardware like some of us here. Waiting several years is a long time for folks that actually utilize the hardware, whereas holding off wouldn't really affect the majority as their current hardware would likely get them by just fine. It's the whole "wants vs needs," battle and I often find that the 'wants' out weighs the actual 'needs' by a large margin.

I agree...going forward it's going to be the battle of cores and ipc. It's the very weapon AMD used against Intel and they've succeeded in holding Intel's head down in the ground hog hole. The only thing escaping for Intel is the fast rabbits (9900K's) that still out paces AMD's elephants. - Intel isn't dead, but the spot lights have shifted....for now.

 

There's also another chip maker emerging called: Zhaoxin? Currently they're on the 16nm, but releasing chips with 7nm soon... Intel and AMD may not be the only names in the industry in the very near future when it comes to CPU's... it'll be very interesting...

Amen to that lol...I totally understand where you're coming from.

 

Even though it is going to sound like I'm contradicting myself (in a way I am), I am curious about the last major ringbus based 10 core from Intel coming next year. That processor compared to my 6950X. The 6950X hits a wall at 4.6/4.7 without exotic cooling. I wonder if that 10 core will hit 5.0 much more easily on H2O. I love my 7980XE, one of the few processors I can say that about. The 10980XE will overclock better but again those baked-in security migrations holds it back. You guys will see how well my 7980XE performs when Mr. Fox has it under chilled water.

 

Zhaoxin is the joint venture between Via and the Chinese government. Via has some licensed to X86. A year ago, on a dirty move, some were saying AMD transferred x86 tech to China during their JV, which was a lie. Considering Via is working with them AND the patents run out next year, get ready for plenty chomping at the bit to enter the market.

So Intel's high margin days, already ended by AMD, will be buried in the post patent period. Even Apple will be able to design their own x86 chips. So the monopoly period is ending.

Also, RISC-V is on the rise as far as adoption across multiple industries. It will not be an x86 competitor any time soon, but being open source and less legacy instructions being required, it could have a quick rise.

 

BAHH! I would have been number one if Stuart_GGF didn't post his 6802. No worries, I am going to keep working on the RAM speed. I'm going to take that number one spot!!!

https://hwbot.org/submission/4308280_

15119cb on Cine-20 with Benchmate.

https://hwbot.org/submission/4308290_

 

Ha, my plan's coming together then! New GPU in 2020 and then the next one on 3nm multi-die in 2023/24!

 

That's all I really care about, that people know approximate timelines for releases, even if a year or two out. If buying toward the top end, or just trying to save to get a tier higher on your next build, planning ahead matters.

For me, 2020 is GPUs and cell phones (specifically not 5G; check out the plethora of articles on safety including the one in scientific America or the recent article that FCC isn't testing properly and phones were putting out 8-10x the legal signal to connect to towers 3 miles away). GPU makes sense with Ampere and big Navi coming.

Zen 4 seems like my CPU update. Probably will upgrade the house server as well. And either 2021 or 2022 will be a great year for TVs. We have microLEDs now, but nanoLEDs are coming. HDMI 2.1 should be more pervasive with more features implemented. And HDR coverage may further expand to get beyond just DCI-P3, all while marching toward the 2000-4000 nit range.

 

Very serious scores, that #1 is within reach!

 

I got a Sony BRAVIA (kdl-42w653a) in 2013, it'll be nearly 10yrs old come your 2022 you mention, could be time for an upgrade at that point, but this TV got great reviews and it's been great...I don't really need a new one (and I calibrated it with colorimeter too). I think it will be time to upgrade it when free TV here shows at resolutions over 1080p as standard (when 4K is standard I guess).

 

Ok, I'm pretty excited about this, I put an NVMe heatsink on my Sabrent Rocket 1TB drive and it's dropped temperatures by 25 degC and completely eliminated what I had previously thought was a TLC write hole - now the drive can write pretty much continuously at over 3GB/s.
This is the heatsink:


Have a look at the following screenshots of CrystalDiskMark tested at a massive 32GB file size:

Before installing NVMe Heatsink:


After Installing NVMe Heatsink:

As you can see the max temperature before applying the heatsink was 59 degC and after applying the heatsink it's only 34 degC. You can also see the massive speed up in sequential writes from 1GB/s to 3GB/s. What I find strange is that the drive wasn't reporting that it was thermally throttling before adding the heatsink, and 59 degC is not the throttling point for this drive, so I find it really quite strange that a heatsink made such a big difference. Basically applying this heatsink has given this drive the same/ or better sustained writes over large file sizes than say the Samsung 970 Evo which is many times more expensive. I recommend slapping a heatsink on your Phison E12 drive if you have one even if it's not running hot....brands that are Phison E12 are Sabrent Rocket / Corsair MP510 amoungst a number of others too.

I find it amazing that I thought this drive had TLC write hole problems with files larger than 27GB, but it turned out it was thermally throttling without even telling me at only 59 degC - applying a heatsink completely eliminated this 'problem' and now there is no perceived TLC write hole and only a max temp of 34 degC during that Crystal Disk Mark test!

Thinking back on it, previous testing without heatsink during the height of the Summer showed that it wouldn't go over 59 degC, which is the same temperature without a heatsink now in Winter....so my drive is obviously slowing down performance when it reaches 59 degC, but God knows why it doesn't report that it is throttling when it does so. Well, applying this heatsink completely solved that as you can see!


EDIT: Ok, some more testing using other 'benchmarking software' is showing that the TLC write hole is not removed, but the heatsink does seem to have helped in some situations like this CrystalDiskMark test that I outlined above. I'm continuing some other testing to see what's going on........

EDIT #2: Yes, the same TLC write hole even with the heatsink attached. I re-ran CrystalDiskMark at 32GB file size and this time it reported only 1GB/s write on sequentials - that earlier test must have been some kind of 'blip' that enabled it to get to 3GB/s, however sequential reads are still a bit faster than before (which can also been seen in the screenshots above), so there does look like there is some benefit to having a heatsink. And given my previous observations that without a heatsink it always topped out at 59 degC regardless of large changes in room temperature suggests it was doing some kind of temperature management or throttling before - and this is now definitely gone with the heatsink attached, which I think is showing in increased Read performance on that CrystalDiskMark test. I don't know how it managed to hit that 3GB/s write for 32GB file size...that was just a one off blip not related to the heatsink, it's gone back down to the usual 1GB/s. In conclusion, I think there is some benefit to have an NVMe heatsink, but the benefit is not as great as I thought it was when I first started writing this post!

 

That's pretty interesting. I'm sure the heat sink can do nothing but help. Cooler is always better.

I am wondering if maybe your drive is reporting wrong temperatures. It is interesting that my Crucial MX500 M.2 SATA SSD is reporting temps that are much higher than any other SATA or NVMe drives in my system. I wonder if that drive temperature is also inaccurate. I have 5 NVMe drives installed and all of them are much lower temperatures. The MX500 is used for Zorin OS (Linux) so it's not even being used at the moment. Makes little sense to me that it would be running hotter. All but the two Samsung 960 Pro are mounted to PCIe cards and are open air (including the Crucial M.2 SATA SSD). The two 960's are on the motherboard under the plastic shroud.

 

I think my drive under reports temperature by about 8 degC. I know this because after installing the heatsink & turning the PC on for the first time I immediately loaded up HWInfo Sensors and it was reading 11 degC when it's at 21 degC room temperature at waist level - lets say it's 2 deg C cooler near the floor which is where my PC is, then that makes for about 8 degC under reporting of temperature. Without heatsink my drive would always top out at 59 degC regardless of room temperature changes so maybe it was really 67 degC which is kinda close the supposed 70 degC throttling temperature. Yeah, so I think my drive was under reporting temperature to external sources, but I wonder if it knew internally that it was running hotter than what it was reporting (maybe it's a 'trick' to make it look like it's running cooler to please consumers). Yeah, so I think the heat sink has definitely helped because by bringing max temps down by 25 degC there is zero chance of this drive ever getting close to that throttling point (even if it's externally under reporting by that 8 degC).

As a side note, you don't want your NAND running too cold because it reduces the life expectancy of the NAND - during write actions it creates some kind of 'breakdown' if the temperatures are too low (40 degC is the ideal target for NAND writes). I suppose I'm not far off that with a 34 degC max temperature during the most intensive writes - at idle it's 16 degC it's telling me now so that's less than ideal but at idle it's not doing much writing so the 'damage' is minimal. Well with the controller, you want that running as cool as possible, just the NAND doesn't like being too cold.

 

Fired up the 3970x and it went for Gold. This CPU is a monster... I've barely got it set up, but ran some benches and I was surprised to hit an OC of 4.6GHz (All-Core).

I haven't really tuned it yet. Looking forward to some more benching... stay tuned.




As you can see it beat one of the fellow OC'ers (keeph8n's) score who was on LN2 @ 5122MHz.


@Papusan @Mr. Fox @Rage Set @Raiderman @JoeT44 @Talon @ssj92 @electrosoft @ajc9988

Here is the wPrime 1024M run:

Everyone above is all LN2. I'm using bloated W10 so I might be able to pass keeph8n with some tuning..

 

Cinebench R20: 3970x @ 4.5GHz

Haven't touched the RAM yet...


@Rage Set if it wasn't for the LN2 guy lol...that feeling eh? (..then again he's within reach so we'll see...)

 

Congrats on the scores! I was watching Joe's stream the other night and must say that thing is an absolute beast.

 

It does a fantastic job in Cinebench. Nice! Congrats on the new toy.

The wPrime results are kind of disappointing, even for the LN2 guy (substantially slower than 7960X in wPrime). It must be a bit sluggish to initially react for those speed tests, especially the 32m test.

 

Thanks Brother. Yea it did seem a little off for sure... So far I've noticed that the cpu is a little clunky. It has a lot of torque, but compared to the 9900K with a lot of top end or even the 7980XE with a strong low end and a faster top end (than the 3970x), the difference is noticeable.

It almost feels like a Quadro card...which it is a workstation cpu after all. It's a heavy duty diesel truck for sure, excelling at what its made for, but other than that I actually enjoy the 7980XE more.

I'm going to see how far I can push the clocks on water, tune the RAM a bit and see how it does...

 

Great analogy as it really does fit. How's the infinity fabric on your sample? And surely there are gains to be made from tightening up the current CL26-25-25-58 timings.

 

Absolutely. I haven't gotten to the ram yet, however, I'd be happy to get 1:1 or close to it. With Christmas right around the corner, I'm sneaking in as much benching as possible without being noticeably MIA from everyone lol. I'll dab more into it tonight after dinner.

Stay tuned...

 

Oh man, lol, you gotta stop those car analogies, keep benching though! Looking forward to hearing about your RTX 2080ti under water advantage, and yep that CPU of yours is crazy in that wprime1024 with 17secs!

 

Haha I love cars too and it just so happens to fit the comparison well between the cpu's, it's hard not to use it. I'm getting base line and bench data first then moving onto the 2080ti on water. Stay tuned.

 

Yeah, I know the feeling. That's one of the reasons why i didn't get the 70x....yet. I knew the majority of the extreme overclockers are going with the 70X and not the 60X. It's within reach but not sure on normal water.

I'm currently investigating a burning smell coming from my board. It's operational but I'm noticing that it isn't overclocking as well anymore. If it were cpu degradation, I wouldn't be getting a smell from it and I haven't pushed the voltage above 1.35 at all.

 

How are your VRMs looking? I'm going to use the car analogy here, CB is like putting the HW onto a dyno. Doing several pulls does a number on the VRMs.

 

I would have performed a full diagnosis by tomorrow evening on the board and I'll share my findings. I do know the VRM fans have not spun yet during an overclocking run, only when I boot the system do they spin. Also, I have not enabled LN2 mode, which would cause issues with the VRM's if I were stupid enough to run that mode in ambient temps.

 

I think both the 60x and 70x are about the same in terms of xoc adopters so it doesn't matter much either way as it'll eventually even out. If you look on hwbot, they're both pretty much the same as far as activity goes. I'm also on a chiller now as of last night, not normal water.

Hmm that's odd... did you take the vrm heat sink apart to examine it? Perhaps it could be a piece of plastic or something that got stuck somewhere. I'm really curious now... hope everything is okay.

Haha bingo... the dyno analogy fits the bill perfectly...

Sounds good.. keep us posted Brother...

 

I fear it is my VRM's based off my initial testing. CB15 and CB20 are crashing on the same exact settings I used just a few days ago. While using HWINFO and running a Time Spy Extreme CPU bench, my VRM temps stay strangely at 24C. I know Time Spy Extreme doesn't pull the same amount of power as CB15/CB20, so perhaps that is why it doesn't crash. I may be seeking a replacement from Newegg before it is too late. I don't want to go through Asus at this point and be without a mobo for a few weeks.

I am going to continue to test the mobo.

 

Which mobo do you have? The Strix TRX40 board that I have feels cheap and flimsy. My GB Aorus Z390 Master is better built than this Asus... pretty disappointed in their quality lately. I also have the Zennith II so hopefully that's better...

If that fails too... I may just go with the Gigabyte board, which is better quality than the Asus boards.

 

Oh man, that sucks Did you visually inspect the VRMs under the heatsink? If you smelled smoke, chances are you can see what component failed.

 

I have the Zenith Extreme II. It appeared to have better build quality than the original I have but that remains to be seen if the VRMs are bad.

I am not sure if I'm going to go that route, as I don't want to give Newegg any excuse to deny a replacement. I'm just going to send it back and get a replacement. I love Asus's BIOS but I'm done with Asus after this.

 

Ah ok ...darn ... we'll I guess we'll see how my Zenith Extreme II holds up.

About being done with Asus...yea I agree, I was done with them after Z370 ... Their bios intuitiveness is top notch, but others have improved a lot to the point where it's not even a real issue anymore. The GB bios isn't bad at all... feature rich, just takes a little to get used to from those coming from asus. I find myself zipping through GB bios actually faster than asus's bios...

 

~Merry Christmas~

I was able to get in some benching last night and here are the results. Hanging out with the LN2'ers with my water scores... pretty happy about the results .

----------------3970x Threadripper----------------

Hall of Fame:

https://hwbot.org/submission/431107...threadripper_3970x_19825_pts?recalculate=true

Here are both wPrime v1.55 runs at 4.6GHz and 4.5GHz.


https://hwbot.org/benchmark/wprime_-_32m_with_benchmate/
https://hwbot.org/benchmark/wprime_-_1024m_with_benchmate/

 

I have not benched with benchmate yet, but I can see there is a performance hit when using it. A pretty significant one as well. At least in the wprime numbers.

 

And, depending on how you have configured and what you have disabled in the BIOS, Benchmate sometimes cannot work right because it is unable to read some things correctly and declares results invalid when they are not. I stopped trying to use it for both reasons. Waste of time.

 

I mean you are forced to use Benchmate with AMD due the never ending mess... Ticked Off: Clock Bug Leads to False Ryzen Benchmark Scores anandtech.com

 

Thats some serious click bait there. I read the forum comments as well, and the bug was introduced in Windows 8, and involves altering the BLCK, which nearly all AMDers never adjust.

I agree with jimmy smitty and Danthunder. This thing couldn't be written more "click-baitish " in favor of Intel processors than if it was written by the marketing people at Intel. Besides it sounds like thew whole thing was initiated by Intel/Microsoft effort when Windows 8 was introduced. Why does the headline read like all AMD Ryzen processors have questionable timing outcomes. Wait let me guess. Has AMD been getting some deservedly good press too long. Although the article thoroughly covered the topic. Shame on Tom's Hardware for running the article with that headline!!

https://forums.tomshardware.com/thr...o-false-ryzen-benchmark-scores.3520916/page-2

I am wondering if that is why some of my benches with similar systems seemed low. I wonder if some people were using this to gain an advantage? I bench primarily with Win7, and the bug doesnt seem to be an issue with it.

 

Although I have my Windows 10 LTSC looking almost as good as my 7

 

Just did a quick comparison in Windows 7 Ultimate, and wprime 1.55 with benchmate, and without. Unexpected results.

 

Windows 10 sucks

 

https://hwbot.org/submission/4311248_

There is a big difference with that particular benchmark, and win 7/10.
My guess is everyone is using win 10 except me. Since you are not supposed to be able to use Windows 7 on the 3000 series Ryzen, there are no official chipset or USB drivers.

 

Indeed... It's unfortunate that they're likely going to transition into using benchmate as a standard ...

I agree.. I don't want to use it, however, it is what it is... At least the benches will all be relative due to everyone having the same handicap so to speak.

True... It's just another unit of measure and looked at two different classes of its own. Hence, why hwbot has individual categories for both although they are the same benchmark.

Tom's Hardware has gone down hill so far it's comical. They've been ripped on by the industry for their over the top articles that are just hilariously contradicting. Actual Tom left / sold the site a long time ago and since then it has been a downward slide into candy land.

Ideally w10 isn't ideal. I don't prefer it at all.. I'm windows 7 all the way... A lot of these benchmarks would do better on W7 as you've shown .. But for giggles I've ran these on my general bench drive..

Aweosme score! Congrats.

 

Brother @Papusan created a new thread that should be fun to watch. Grab some popcorn and watch the kiddos have an emo meltdown. This Plundervolt drama couldn't have come at a better time for AMD. Won't matter that their CPUs suck at overclocking if Intel is going to commit hara kiri for the illusion of "security" LOL.

I bet the whole thing is an elaborate and carefully orchestrated scheme to get people to buy new computers that are "safe" from the boogieman. If it is not something they invented for that purpose, they'll certainly be able to milk it that way.

BLACK LIST. Adjustable voltage control/turbo ratio limits are locked out due latest Win Update/Bios

Windows 10 is another example of a scam to trick people into spending money on crap they don't need in the name of "security" (also a lie). It didn't work, so they've had to kick it into Nazi-mode and resort to digital genocide to get the job done. I'm still using Windows 7, and will continue using it until it stops working, so up their nose with a rubber hose.

 

Amen to that. The false sense of security with w10 users jumping onto kb updates labeled "security," is cringing. I just don't get it... To trust Microsoft or any of these corrupted organizations is voluntary submission.

Long long live windows 7 and don't worry brother, if that ever haults, there's always many flavor of open source Linux goodness to enjoy for computer eternity.

Have a Merry Christmas!

 

Thanks! Considering the true reason we celebrate Christmas, there is no way for it to turn out other than great.

I'm pretty sure the same flavor of stupid (illogical panic over security) has a similar stranglehold on the emo-cleptic clowns behind Linux development. I don't think any of this is going to have a happy or acceptable ending for hardware enthusiasts no matter what OS they're using. The latest kernel/Intel micro-code updates for Linux are causing random (although not frequent) hard freezes for my desktop on both Mint and POP!_OS. Both Linux OSes began having the same behavior after the latest security updates packing their latest and greatest cancer payload.