Jump to content
Sign in to follow this  
Windsurf48

Finally got Overdrive to finish

Recommended Posts

I finally got the Overdrive test to finish almost successfully, but the scores are disappointing, with everything in the top 4 percent except Internet, which didn't run: http://www.pcpitstop.com/betapit/sec.asp?conid=26073238

 

ASUS Zenith II Extreme

AMD Threadripper 3960x

Corsair H150i

4 x 8 GB Corsair Dominator 4266 at 3200 MHz

Patriot Viper VP4100 1T PCIe4 M.2 drive

 

Share this post


Link to post
Share on other sites

Holy moly, that is a beastly machine! Whatcha gonna do with all those glorious cores?

 

Tx 🤠

Share this post


Link to post
Share on other sites

Mostly run benchmarks and experiment.  I'm still going through a learning experience with some of its quirks.  The biggest issue is keeping it cool because the H150i with six push-pull ML120 fans isn't enough to handle 48 Prime95 workers at default settings, so overclocking is out the question for now.   Eventually, I want to experiment with  PCIe4 M.2 RAID, which should in interesting since the bandwidth limits are gone.

Share this post


Link to post
Share on other sites

By the way, I got messages about IE scripting not being enabled after the first test run.  I went through the IE options and it was enabled already, but then got another message about cookies or something similar.  I went through the IE options and the second item was also already enabled.  The third time, the test finished.  I don't know if the process of going through the IE options somehow reset something that let the test finish, or whether it just was a coincidence that the test finished successfully on the third attempt.  I may be able to find out after I figure out how to get the Corsair memory stable at higher speeds and I retest.

 

Share this post


Link to post
Share on other sites

That is quite the benchmark queen! I bet it's a treat to use. One thing I'd recommend if the ram is on the mobo qvl is to set it to 3600Mhz, it puts the Infinity Fabic at its max clock speed and will enable optimal bandwidth for RAM and IF.  ;)

 

One other thing of note, the hotplate on that AiO likely is not large enough to fully cover the iihs of that massive cpu package. If it doesn't explicitly state designed for TRX40 or even TR4 socket, it's not gonna be up to the task. https://www.cgdirector.com/best-cpu-coolers-ryzen-threadripper/ as a reference. Notice how large the contact plate is on the Noctua cooler compared to the one on your AiO.

 

 

Tx 🤠

 

 

Share this post


Link to post
Share on other sites

I've tried the RAM at 3600 MHz, but the system is not stable.  I probably need to increase the power, which defaults to 1.2 but goes up as high as 1.45 for 4266 MHz.  That's one of the things I haven't gotten around to trying yet.

 

Thanks for the link to CPU coolers.  I already had the H150i and some of the 3960x reviews were done using the same cooler, but I assumed that some of the heat issues were related to the size of the pump's contact area and was planning on looking for one designed for Threadrippers.

 

By the way, I completed another run today on my other hard drive's Windows 10 system and it reported that my 85708 score was in the top 20%, with each of the individual scores in the top 4% except for Internet, which had "soon!" instead of a score. My last previous score was 45583 from 2017, and that was in the top 5% which was expected because it would have been my Core i7-6700k or 8700k.  What kind of systems are getting the top scores these days? http://www.pcpitstop.com/betapit/sec.asp?conid=26073290&report=Summary

Share this post


Link to post
Share on other sites

I checked the link for Threadripper coolers.  The highest-rated one is the Thermaltake Water 3.0, but it has the same size pump and contact surface as the H150i.  The be quiet! Dark Rock Pro TR4 is the highest rated air cooler and has a base which looks as though it would cover the whole CPU, but I don't know if the better coverage with air cooling would be an improvement over the H150i.  I'm certainly going to watch for an AIO designed for Threadrippers, but the article says that the one currently available is too unreliable to be useful.

 

I'll also see if I can find any Threadripper 3960x and 3970x stats comparing coolers which include the Corsair H150i and the Dark Rock Pro, but I think I remember seeing some where the liquid cooling trumped the better base.

Share this post


Link to post
Share on other sites

If your system is unstable at the XMP 3600Mhz, double check that the ram is on the QVL for the mobo. If it's not, I'm 99.9% sure this is the cause based off my experience with Ryzen starting with Gen1. 

 

As for your question about what systems are being tested that are raising the bar so high, your guess is as good as mine. I'll ask the developers and see if they have any clues. I've built several TR4 systems and used the Noctua TR4 spec'd cooler with fabulous results. It's right there with the best AiO setups. I've learned though that if you're not running full bore for extended an duration of time, you're better off w/a known effective air cooler like the Noctua.

 

I used to be a firm believer in the AiO coolers, but the results of testing just doesn't support their superiority or cost. It's just simpler to use an air cooler and if you need the extra cooling capacity, build a custom water loop.

 

Tx 🤠 

Share this post


Link to post
Share on other sites

I checked the QVL before ordering the RAM.  That's not the problem.  TRX40 doesn't have XMP, and the equivalent DOCP only has an option for 4266,.  The problem is that selecting DOCP automatically locks the cores at 4.3 Ghz, which is too much for the H150i.  It took me a couple of weeks to deal with driver issues in Windows 10 x64 and I reseated the pump with better thermal paste last weekend, so I'm just reaching a point where I'm ready to experiment with RAM settings.  

 

I need to check to see if either the Dark Rock or Noctua's TRX40 cooler will clear the RAM.  I've had good results with Noctua heatsinks in the past; the only downside is Noctua's preference for really ugly fan colors.  I'm not sure I want to overclock badly enough to set up a custom water loop.

Share this post


Link to post
Share on other sites

You've more than done your homework is sounds, but there is good news on the Noctua front, they started making all black stuff. https://noctua.at/en/noctua-introduces-chromax-black-cpu-coolers

 

Here's one build I did with the Noctua cooler, granted I use Corsair LPX ram, the cooler i specifically designed to not obstruct dimm slots. JHV1lUD_zI79EAETx17ldDgooqs-mjRplztwxAEH

Share this post


Link to post
Share on other sites

I've seen the black Noctua coolers.  Unfortunately, the RAM is too high for either the Noctua or be quiet! coolers.  I'll just run the PC without overclocking the CPU and play around with other variables until someone other than Enermax comes out with an AiO pump designed for Threadrippers.

 

My i7-8700k and Maximus X Apex was using RAM designed to fit under heatsinks. I haven't used an AiO since my first one leaked and fried my motherboard three or four years ago.

 

Edited by Windsurf48

Share this post


Link to post
Share on other sites

I forgot the mention that I noticed something about the percents on the Ranking tab.  The five individual scores showed 4% each, with no score or percent for Internet bandwidth, and the total ranking showed 20%.  Is there any possibility that Overdrive is adding the individual percents to get the overall percent instead of calculating it separately?  The 4% calculations are suspect, too, because it seems unlikely that the CPU, RAM, hard drive, and video scores would all fall in exactly the same percent range.

Share this post


Link to post
Share on other sites

I've no clue how that number is derived. I did ask the developers and they'd said the scores were off for a bit due to a table not populating(I think I remember that correctly, but don't quote me), it's since been fixed.

 

 

Share this post


Link to post
Share on other sites

I'm just curious about what kind of rigs are getting the top scores.  If they are using LN cooling or are running two Xeons and three video cards, I don't worry about it, but if they are comparable to mine, then it's interesting to try to figure out what's holding mine back.

 

I was trying to figure why the 3960x and TRX40 chipset were getting 10% lower scores than my Core i7-8700k and Z370 in Passmark Performance Test with both running at default settings, and found out that temporarily disabling the Meltdown and Spectre patches resulted in scores 25 percent higher.  It didn't make a difference in Overdrive, however.

 

Edited by Windsurf48

Share this post


Link to post
Share on other sites

The memory has been running at 3600 18-19-19-39 1.35 since noon with no problems, including completing Windows Memory Diagnostics.  Unfortunately, I haven't been able to complete Overdrive to see what difference it makes, but the AIDA64 memory scores are encouraging.

 

ASUS's RealBench has similar problems, displaying a message about being unable to parse system specs which blocks uploads after completing the tests.  This seems to have been going on for several years with no resolution.  It's frustrating because the score looks good compared to the online database, but that might be partly because other high scores wouldn't upload.

 

At least one of SiSoft Sandra's full system benchmark tests chokes on TRX40 systems, but hopefully this is a recent problem which will be fixed soon.

Edited by Windsurf48

Share this post


Link to post
Share on other sites

I replaced the Corsair Hydro H150i with a Noctua NH-U14S TR4-SP3.  Idle temperatures are slightly lower, but initial testing indicates that temperatures under load are higher.  However, I'm waiting for an RMA for the memory certified for the Zenith II Extreme and using a pair of older memory sticks until it arrives, so there's no point in doing much at this time.  The NH-U14S does fit easily on the TRX4 motherboard.  The fans have to be raised slightly for taller memory, but with four sticks, the front fan is clear of the memory anyway.

 

AIDA64 shows the fans speeding up under load as expected, so if the load temperatures are still high when I can resume testing, I'll take the heat sink off to verify that the thermal paste was applied correctly and, possibly, try different thermal paste.

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Sign in to follow this  

×
×
  • Create New...