• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Threadripper BUILDERS thread

Page 51 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
ddogg said:
Yeah, I was a little surprised at the orientation as well. They are still waiting to get a photographer to update their website. Really curious to see what kind of temp drops I see with this.
Click to expand...

If you can do a 1:1 replacement versus the EK TR4 block, I'd be curious to see how much improvement the change in water block alone gives to your temperatures. Because 2.2 lbs is pretty hefty for a water block.
 
I'll try to get it done tomorrow. And this is closer to 2.4lbs 😱
EWDxS7h.jpg
 
ub4ty said:
K, you have a solid understanding of this. Question : Every reviewer (when the CPU was at stock settings) noted poor performance in certain heightened RAM clocks when they started ratcheting up the RAM clock timings above 2133. This is likely due to the 180W ceiling the CPU imposes whereby at higher ram clocks, it uses more power and thus must cut power to cores and not clock them as high to stay within the stock 180W ceiling.

I think anandtech defined the phenomenon as operating exactly like this. Higher RAM clocks when the CPU is stock starves the cores for power. Thus, to keep things balanced, you essentially have to OC in order to get things back in balance. I stated prior w.r.t to boot loops and instability that my bios OC appeared not to be taking. I am not sure what allowed for it to take but it took after I did the 1.0 -> 1.3 update. When I didn't OC the CPU I got boot loops and 3600@CL16 XMP setting wouldn't take. After I applied a mild OC to the CPU 3700Mhz @1.21V the memory settings took, i booted to windows, ran cinebench, am now running burnintest and will switch to more extreme testing.

So, is the instability due to the CPU doing probing of the voltage range/clocks on Post and in windows with a 180W limit but being unable to supply sufficient settings to run the RAM at 3600Mhz while giving adequate power to the cores simultaneously? Trying to root cause why stock CPU + 3600RAM @CL16 is unstable.
Click to expand...


An inquiry for folks who have spent some time running their 1950X at stock speeds to satisfy my curiosity:
Background: When running the 1950X CPU at stock speeds but with RAM "overclocked" (e.g. running at 1.35V and DOCP speeds instead of "default" 1.2V JDEC speeds), there is a possibility of decreased performance. This is because the CPU needs to allocate more power (from its max TDP of 180W at stock) to the memory controller. Anandtech results suggested a possible 13W loss of power to the CPU cores under such a scenario. But they never actually tested the performance tradeoff of increased memory frequency versus more power available for to boost CPU core speeds. See: https://www.anandtech.com/show/11697/the-amd-ryzen-threadripper-1950x-and-1920x-review/19

Question: Has anyone here run benchmarks on a non-overclocked 1950X at different RAM speeds (e.g. 1.2V 2133 MHz vs 1.35V 3200 MHz) to see how big of a performance impact this has? And which is more advantageous (e.g. higher memory speeds/infinity fabric performance vs more watts available to CPU cores).

The Anandtech article talks about the possibility of decreased performance, but they don't actually test the impact. I presume that, unless your work load is mostly limited by memory speed/bandwidth, that overall system performance might take a hit when running RAM at faster speeds. Which would be the opposite of the desired outcome. But who knows? This infinity fabric business is new to me. You guys have done a lot of testing (in aggregate) so might be able to answer this question relatively easily.

[EDIT: By the way, I could not find the mentioned reviews that discussed this issue and showed considerably reduced performance at higher memory speeds. But maybe I didn't look hard enough. Most reviews showed higher performance at higher clocks, but it was not clear that they compared at 1.2V for stock memory speeds versus 1.35+V for higher memory speeds. If RAM voltage was kept relatively constant, that would reduce the power draw differences.]
 
Last edited:
I was wondering what default VDDCR_SoC values you guys are getting. I'm seeing higher voltages with my Taichi x399 than mentioned by the Stilt.

With 8 sticks of memory at @2800 MHz, the motherboard sets VDDCR_SoC to 1.136V. The voltage goes higher if I increase the RAM speed, as expected. Interestingly, if I manually set the voltage to 1.1V, the Taichi seems to ignore it and keeps the higher setting.

I read that the SoC voltage should not go above 1.1V, so I'm worried that the motherboard is setting the voltage too high at these faster RAM speeds.
 
Heads up, bios v2 is out for the x399 Taichi.

1.Sync the H/W Monitor and Dr.Debug code Temp source to Tctl.
2.Update AmdPackageString to "ThreadRipperPI-SP3r2-1.0.0.4"
3.Enhance SATA type M.2 compatibility.
4.Include "BT On/Off" item for Bluetooth control.
Click to expand...
 
Charlie22911 said:
Heads up, bios v2 is out for the x399 Taichi.
Click to expand...

Cool, thanks! I'm on 1.80, might try this to see if it changes anything with the voltages.

Edit:

I updated to the new 2.0 bios and (in my very limited testing) it looks like the default SOC voltages are the same. However, it is accepting my manual override, so at least I'm able to lower it a bit.
 
Last edited:
so......... Now that TR has been out for awhile... what's the best MB to get??

Thinking about this one... Gigabyte X399 DESIGNARE EX

Anyone got on feedback with this board?

Got the 1950X, 360 enermax, Case and P/S. Just need RAM and MB now and a 1080 ti. I'm taking my time. 🙂
 
ericlp said:
so......... Now that TR has been out for awhile... what's the best MB to get??

Thinking about this one... Gigabyte X399 DESIGNARE EX

Anyone got on feedback with this board?

Got the 1950X, 360 enermax, Case and P/S. Just need RAM and MB now and a 1080 ti. I'm taking my time. 🙂
Click to expand...
The Asrock Taichi seems to be the most stable, with the best ram support.
 
TrackSmart said:
An inquiry for folks who have spent some time running their 1950X at stock speeds to satisfy my curiosity:
Background: When running the 1950X CPU at stock speeds but with RAM "overclocked" (e.g. running at 1.35V and DOCP speeds instead of "default" 1.2V JDEC speeds), there is a possibility of decreased performance. This is because the CPU needs to allocate more power (from its max TDP of 180W at stock) to the memory controller. Anandtech results suggested a possible 13W loss of power to the CPU cores under such a scenario. But they never actually tested the performance tradeoff of increased memory frequency versus more power available for to boost CPU core speeds. See: https://www.anandtech.com/show/11697/the-amd-ryzen-threadripper-1950x-and-1920x-review/19

Question: Has anyone here run benchmarks on a non-overclocked 1950X at different RAM speeds (e.g. 1.2V 2133 MHz vs 1.35V 3200 MHz) to see how big of a performance impact this has? And which is more advantageous (e.g. higher memory speeds/infinity fabric performance vs more watts available to CPU cores).

The Anandtech article talks about the possibility of decreased performance, but they don't actually test the impact. I presume that, unless your work load is mostly limited by memory speed/bandwidth, that overall system performance might take a hit when running RAM at faster speeds. Which would be the opposite of the desired outcome. But who knows? This infinity fabric business is new to me. You guys have done a lot of testing (in aggregate) so might be able to answer this question relatively easily.

[EDIT: By the way, I could not find the mentioned reviews that discussed this issue and showed considerably reduced performance at higher memory speeds. But maybe I didn't look hard enough. Most reviews showed higher performance at higher clocks, but it was not clear that they compared at 1.2V for stock memory speeds versus 1.35+V for higher memory speeds. If RAM voltage was kept relatively constant, that would reduce the power draw differences.]
Click to expand...


I might be misinterpreting your post but you seem to be saying that you can’t find any evidence that proves this is an actual problem?
 
urvile said:
I might be misinterpreting your post but you seem to be saying that you can’t find any evidence that proves this is an actual problem?
Click to expand...

Indeed. It's a curiosity question with regard to threadripper and the relationship between package power and power allocated to the CPU cores themselves when operating at the default TDP limits.

The Anandtech review appears to show a potential performance issue caused by running at higher RAM voltages when not overclocked. And the poster on this board (ub4ty) seems to suggest that performance regression at higher RAM voltages/speeds was commonly found in reviews when tasks were not heavily bound by memory bandwidth -- that was news to me! I could not find such reviews.

So, the question is really whether there is any real evidence for a performance tradeoff at stock CPU settings when using higher RAM frequencies.

[For Reference: I was more likely to see examples like this (see second graph), where virtually all tests show higher performance with higher memory speed.
https://www.pcper.com/reviews/Proce...0X-Review/Memory-Mode-UMANUMA-and-Memory-Spee
That said, it is possible that these reviewers are running their RAM modules at 1.35V (standard voltage for higher clocked kits) at both 2400 and 3200 MHz, meaning the disadvantage of higher voltages is uniform. They do not specify.]
 
Howdy2u2 said:
FYI Threadripper is on sale @ microcenter..........$150-$300 off
Click to expand...
unfortunately most of the country or the world can go there...... The closest one for me is over 1000 miles away. Any those sale prices are always in-store only.

But thanks for the heads up for the small number that can make it there. (speaking world-wide)
 
ericlp said:
so......... Now that TR has been out for awhile... what's the best MB to get??

Thinking about this one... Gigabyte X399 DESIGNARE EX

Anyone got on feedback with this board?

Got the 1950X, 360 enermax, Case and P/S. Just need RAM and MB now and a 1080 ti. I'm taking my time. 🙂
Click to expand...
I have the Asus Zenith Extreme. I would've bought the Taichi if I didn't see the Asus board at $350 (also considered the MSI board). I have the X370 Taichi. I like Asrock and Asus BIOS - not too familiar with Gigabyte.

I grabbed my Gigabyte 1080 Ti from Massdrop ($679). It seems like Ram prices are still climbing. The 4x16GB that was around $750 when I bought it, is now almost $950. I'm hoping to see more reasonable prices around Christmas, but I'm not counting on it.
 
Markfw said:
unfortunately most of the country or the world can go there...... The closest one for me is over 1000 miles away. Any those sale prices are always in-store only.

But thanks for the heads up for the small number that can make it there. (speaking world-wide)
Click to expand...

Yeah I know, It's a haul for me to get to a store also. I should have added the "catch phrase" In store only. Amazon had them on sale last week for $799.99 but that ended quickly and now is back up to $900+ at most places you can buy on line. Alas, I was just putting it out there for those lucky enough to be near a Microcenter store.
 
TrackSmart said:
Indeed. It's a curiosity question with regard to threadripper and the relationship between package power and power allocated to the CPU cores themselves when operating at the default TDP limits.

The Anandtech review appears to show a potential performance issue caused by running at higher RAM voltages when not overclocked. And the poster on this board (ub4ty) seems to suggest that performance regression at higher RAM voltages/speeds was commonly found in reviews when tasks were not heavily bound by memory bandwidth -- that was news to me! I could not find such reviews.

So, the question is really whether there is any real evidence for a performance tradeoff at stock CPU settings when using higher RAM frequencies.

[For Reference: I was more likely to see examples like this (see second graph), where virtually all tests show higher performance with higher memory speed.
https://www.pcper.com/reviews/Proce...0X-Review/Memory-Mode-UMANUMA-and-Memory-Spee
That said, it is possible that these reviewers are running their RAM modules at 1.35V (standard voltage for higher clocked kits) at both 2400 and 3200 MHz, meaning the disadvantage of higher voltages is uniform. They do not specify.]
Click to expand...

Perhaps it will have to go down as one of life’s great mysteries?
 
Howdy2u2 said:
FYI Threadripper is on sale @ microcenter..........$150-$300 off
Click to expand...

Microcenter's pricing is... interesting. For Black Friday / Cyber Monday through the day after prices were fantastic. I think the 1900x was $350, the 1920x $500, and I can't remember what the 1950x was, but it was cheap. Then that Wednesday prices were still good with the 1900x at $370 if I remember correctly, and the 1920x at $550. I was working late and couldn't get to a store in time and then postponed my purchase. Then earlier this week the prices climbed back up for some weird reason, and the 1920x was up to $650, the 1900x $450. Now they're down again!

And interestingly it's worth doing the math on what you actually end up paying when considering taxes paid. In some cases you actually don't gain anything monetarily by going to a store versus buying from Newegg for example (although it's good to be able to physically go to a store (if it's nearby) to return an item or whatever if there's a problem).

Anyway, question:

Does anyone have any experience with Microcenter and Newegg / Amazon etc as far as Christmas sales go? Can we expect further price reductions before 2018?
 
depends... normally, newegg and amazon match each other pretty quickly. I use camel camel camel to alert me if prices start to dip.... esp, since.... amazon will match a lower price in 30 days or something like that. The only thing I know, is if Intel ever comes out with a 16 core cpu, it sure as hell won't be 799. Also, for those thinking that intel will be faster, guess again. While like everyone here.....I'm all for innovation and hope that Intel or AMD will bring something new to the table twice as fast while using less power, creating less heat etc...etc... I just don't see how we going to go past 16 cores and do it at 5Ghz and overclocking and still manage to keep the heat within below 70C.

I think the next step might be multiple CPU's on a motherboard. Like have a board where you got 4 CPU sockets across the top. Maybe populated with Ryzen 1700's ... Essentially a 32 core / 64 Threads system that is effient and can manage the heat. Or how about 4 TR 1950X? 64 Cores and 128 Threads anyone? LOL!
 
Amazon is down to $899, but I bet between now and Christmas it will hit $799 again, maybe even the $779 we saw once.
 
mattiasnyc said:
And interestingly it's worth doing the math on what you actually end up paying when considering taxes paid. In some cases you actually don't gain anything monetarily by going to a store versus buying from Newegg for example (although it's good to be able to physically go to a store (if it's nearby) to return an item or whatever if there's a problem).
Click to expand...
Unfortunately, I got an email from Newegg yesterday saying they're charging state tax now 🙁
 
mattiasnyc said:
Doesn't Newegg's tax charges depend on the state they're shipping to though?
Click to expand...
From their website:
Within the United States, Newegg.com charges tax on orders shipping to California, Indiana, New Jersey, Pennsylvania and Tennessee only.
I'll have to keep that in mind when I finally order items for a TR build... just moved to Indiana 🙁

In other news, my arms got tired playing a drum roll waiting for a report from:
ddogg said:
I'll try to get it done tomorrow. And this is closer to 2.4lbs 😱
EWDxS7h.jpg
Click to expand...
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top