Threadripper BUILDERS thread

Page 22 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

VirtualLarry

No Lifer
Aug 25, 2001
56,551
10,171
126
My Ryzen 5 1600 rig, seems to only hit 800Mbit/sec on speedtest.net. Maybe I need a ThreadRipper to hit 1Gbit/sec?

J/K, but I'm looking for an excuse to convince myself to consider getting one. No way I can afford that much, already spent too much on Ryzen.

Edit: 850/850 (almost)!
http://beta.speedtest.net/result/6556395600

Still not sure if it's my router, or the site, or if my connection is slightly bogging (unlikely at 4am EST though, I would think).
 
Last edited:

eek2121

Diamond Member
Aug 2, 2005
3,100
4,398
136
My Ryzen 5 1600 rig, seems to only hit 800Mbit/sec on speedtest.net. Maybe I need a ThreadRipper to hit 1Gbit/sec?

J/K, but I'm looking for an excuse to convince myself to consider getting one. No way I can afford that much, already spent too much on Ryzen.

Hah, I can get an 1800X locally for $369. I was a bit torn, since TR is out of stock along with the board I want...but I think I will wait.
 

ub4ty

Senior member
Jun 21, 2017
749
898
96
All set up! I need to finish installing software and downloading some benchmarks. For anyone who has trouble installing Windows 10, make sure to disable IOMMU in the BIOS!
Also, the HyperX Predator 4x8GB memory kit seems to be running at its advertised speed without issues (3200, 16-18-18). I just used the Memory Try It option in the MSI BIOS. These are Hynix chips so maybe someone finds that info useful.

I like your photos. What camera did you use to take em? What are your plans for your build and what is the tiny card in the x1 slot? Also, why do you have to fudge w/ IOMMU? I haven't heard this mentioned on any review sites.
 
Reactions: Drazick

Dave Null

Junior Member
Aug 21, 2017
2
0
6
So I was about to pull the trigger on a Threadripper build. However, my primary use case is audio with lots of virtual instruments, and both Scan Pro Audio and TechReport portray Threadripper doing poorly in realtime audio tasks. TechReport in particular shows ThreadRipper failing pretty hard:



However, I'm wondering if something is amiss with these results. While TechReport shows the 1950x not doing great against the 7900x at 64 samples, they also show the 6950x and the 7740x getting annihilated by the Ryzen 1800x. Can a 7740x really only achieve 14% as much polyphony as an 1800x? Why is the 6950x only getting 28% of the polyphony of an 1800x? And is Threadripper really only capable of 14% more polyphony than an 1800x? Something just seems wrong there. I find the entire benchmark to be suspect.

So here I am with a Threadripper build in my cart at NewEgg, but I've got benchmarks saying "buy Intel." And yet...the benchmarks just don't pass the smell test. Pretty frustrating.
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
26,129
15,274
136
So I was about to pull the trigger on a Threadripper build. However, my primary use case is audio with lots of virtual instruments, and both Scan Pro Audio and TechReport portray Threadripper doing poorly in realtime audio tasks. TechReport in particular shows ThreadRipper failing pretty hard:



However, I'm wondering if something is amiss with these results. While TechReport shows the 1950x not doing great against the 7900x at 64 samples, they also show the 6950x and the 7740x getting annihilated by the Ryzen 1800x. Can a 7740x really only achieve 14% as much polyphony as an 1800x? Why is the 6950x only getting 28% of the polyphony of an 1800x? And is Threadripper really only capable of 14% more polyphony than an 1800x? Something just seems wrong there. I find the entire benchmark to be suspect.

So here I am with a Threadripper build in my cart at NewEgg, but I've got benchmarks saying "buy Intel." And yet...the benchmarks just don't pass the smell test. Pretty frustrating.
What I can say is that a threadripper really is 2 1800X's in one chip, so if you believe that bench, then threadripper has twice that power. I agree that it smells fisky. Also what ram are they using on reach ? TR LOVES fast memory, the faster the better. I got 3600 cas 16, not cheap, but well worth it.
 

scannall

Golden Member
Jan 1, 2012
1,960
1,678
136
What I can say is that a threadripper really is 2 1800X's in one chip, so if you believe that bench, then threadripper has twice that power. I agree that it smells fisky. Also what ram are they using on reach ? TR LOVES fast memory, the faster the better. I got 3600 cas 16, not cheap, but well worth it.
I'm with you. Charts are meaningless without the details. Details add context to the charts, and make the information actually useful.
 

Dave Null

Junior Member
Aug 21, 2017
2
0
6
I'm with you. Charts are meaningless without the details. Details add context to the charts, and make the information actually useful.

I completely agree. Unfortunately, Tech Report didn't dive very deep into their testing methodology. I don't think they themselves are certain about the "why" of what they're seeing. From the article:

The VI test isn't as rosy for the AMD corner. While it's hard to say for sure, we suspect this test tends to favor low memory latency, high cache bandwidth, single-threaded performance, and raw clock speed at our test settings. Skylake-X chips have higher performance in all of those areas, and it seems to show in our results. The dominating performance of the Core i9-7900X is no fluke in these tests, either. We retested it several times to be sure, but the gap remained.

Readers will rightly perk up at a gap like this one, and when we see such a gap between two otherwise closely-matched parts, we do our best to get to the bottom of it. We tried overclocking the Threadrippers, to no effect. We flipped SMT off, to no effect. At the suggestion of one reader, we even played with the number of active cores in Reaper, also to no effect.​

This explanation does not account for the 7740x's poor showing here. Tech Report goes on to suggest their results are similar to those obtained by Scan Pro Audio, but a closer analysis shows that in fact their results are quite different (Scan Pro does show an Intel advantage at low buffers, but Threadripper pulls ahead at higher buffers, whereas Tech Report shows the gap becoming even worse).

I realize it seems unwise to put so much stock into a single benchmark, but not many publications are testing realtime audio performance of CPUs, and that is my primary requirement in a workstation.
 

Atari2600

Golden Member
Nov 22, 2016
1,409
1,655
136
However, I'm wondering if something is amiss with these results.

IMO at least one of those platforms must be un-optimised, or else another is running with incorrect settings. Its just too inconsistent to be right.

The 7820X has twice the cores and twice the bandwidth as the 7740X. Unless this bench uses AVX-512 in a big way, then it stands to reason the 7820X can be no more than twice as fast as the 7740X - and that assumes identical IPC & clock (and I suppose, memory speeds).

If I were writing that review, I would made note of the benchmark in the text, but not included the results due to high degree of uncertainty over accuracy.
 

StefanR5R

Elite Member
Dec 10, 2016
5,913
8,818
136
not many publications are testing realtime audio performance of CPUs
Is there really anything interesting to test in that field? I thought that this is determined by the latency of kernel mode drivers, and hardly anything else.

What I gathered from this test:
  • tested with 1950X at stock clocks
  • tested at the open bench, not in a case
  • tester applied a liberal amount of thermal paste to ensure full coverage, but I wonder about the resulting thickness of the application
  • temperatures under load:
    NH-U14S TR4-SP3 < Floe Riing 360 < NH-U12S TR4-SP3 < NH-U9S TR4-SP3
  • noise under load:
    Floe Riing 360 < NH-U12S TR4-SP3 < NH-U14S TR4-SP3 < NH-U9S TR4-SP3
  • mounting system is easy to handle, but you cannot turn these coolers by 90°
  • For compatibility, the Noctuas can be mounted with a little shift to the side. That way even NH-U14S TR4-SP3 does not interefere with the first PCIe slot on most mainboards. Exception: Asus Zenith Extreme's first slot is too close.
 

kaporkle

Junior Member
Aug 2, 2017
19
33
61
I'd move the 2nd card down to the next X16 slot and get a 2-slot space SLI bridge.

My thoughts exactly. Has to be horribly noisy the way its set up on the picture.

Right now I'm using the little wifi/bluetooth card that ships with the motherboard, and moving the second 1070 down a slot would evict that card. Eventually I'll remove the wifi card and add two more GPUs, at which point they will all be ass-to-ass anyways. I'm not finding it loud whatsoever, but then I'm not super noise-sensitive. The Quadro 4000 that was in my previous computer is a different story. Those are single-slot blower-style Fermy cards, and that thing would rev up at the slightest provocation and run very loud and hot.


I like your photos. What camera did you use to take em? What are your plans for your build and what is the tiny card in the x1 slot? Also, why do you have to fudge w/ IOMMU? I haven't heard this mentioned on any review sites.

Sony a6000 with an old Canon film lens, but the camera/lens don't matter so much as having manual control (so you can control ISO/aperture/shutter speed) and a tripod. You can do this with any entry level Nikon/Canon/Sony/etc. Couple more pics:


The build is primarily for 3D work/content creation. The two 1070s are for GPU rendering.
IOMMU was preventing me from installing Windows 10 (but I've re-enabled it since). Apparently a problem on more than just the MSI board: https://www.youtube.com/watch?v=mMt-Hqf76LI
 

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106
I'm getting close to a point where I'll be throwing the platform into a garbage bin or giving it few blows with a hammer.

The 3D printed mounting bracket failed miserably (incompatible due to the curvature of the heatpipes) and I still have no way to cool the CPU.
None of the resellers have any ETA for Enermax TR4 coolers or Noctua U14S-TR4 either (or anything else for that matter). The only thing available for this platform is still those insufficient POS AIOs made by Asetek (20+ different variants of them).
Few different waterblocks will be available shortly, but a custom loop is not an option for me due to the nature of the setup.

I'm amazed by the level of incompetence AMD has illustrated with this platform. It is basically no different than releasing a new CPU with no motherboards available for it, utterly useless.
Normally I would blame it on the cooler manufacturers, in this case the final thermal design specifications for this platform (SP3) has been available since September 2016. Therefore the only explanation I can come up with, is that AMD didn't inform the cooler manufacturers
about this platform until the release was at hand.

/Rant
 

dnavas

Senior member
Feb 25, 2017
355
190
116
None of the resellers have any ETA for Enermax TR4 coolers or Noctua U14S-TR4 either (or anything else for that matter).

I liked Noctua's statement which I read ... somewhere ... where they claimed that their cooler was (not word for word here, obviously) knocked together because it was so last minute*. I immediately wondered the same thing you ranted about. Someone pointed out the heat pipe distribution, and together this forces me to consider the idea that they have a longer-term, better design that might happen. Dunno. I hope so. I prefer air :shrug:

Also, finally did some research on liquid coolers, and I'm not thrilled by the copper + aluminum cooler setup that liqtech apparently has. Just spend the extra dough and do a proper copper radiator! Grr. Which forces me to consider a custom loop. Which makes me just want to wait.

I'm amazed by the level of incompetence AMD has illustrated with this platform. It is basically no different than releasing a new CPU with no motherboards available for it, utterly useless.

I agree, especially a platform like this. I'm definitely waiting to see what Intel has to offer for their 14 core, and then I'll have to weigh that (likely better performance) against the likelihood that I'll be able to drop-in better CPUs into TR4 in the future. Their current level of 3rd party coordination does not instill confidence.

*: Ah, it's the German review linked a couple posts above. The translated text read "... Noctua in the discussion with ComputerBase said that the coolers for the new base are rather an emergency solution because of the short period of development..."
 
Last edited:

IEC

Elite Member
Super Moderator
Jun 10, 2004
14,440
5,429
136
I'm getting close to a point where I'll be throwing the platform into a garbage bin or giving it few blows with a hammer.

The 3D printed mounting bracket failed miserably (incompatible due to the curvature of the heatpipes) and I still have no way to cool the CPU.
None of the resellers have any ETA for Enermax TR4 coolers or Noctua U14S-TR4 either (or anything else for that matter). The only thing available for this platform is still those insufficient POS AIOs made by Asetek (20+ different variants of them).
Few different waterblocks will be available shortly, but a custom loop is not an option for me due to the nature of the setup.

I'm amazed by the level of incompetence AMD has illustrated with this platform. It is basically no different than releasing a new CPU with no motherboards available for it, utterly useless.
Normally I would blame it on the cooler manufacturers, in this case the final thermal design specifications for this platform (SP3) has been available since September 2016. Therefore the only explanation I can come up with, is that AMD didn't inform the cooler manufacturers
about this platform until the release was at hand.

/Rant

Yes, the cooler situation is less than desirable. Considering EK was claiming they had final specs for the socket many months prior to launch, I'm thinking the lack of air coolers is probably due to so many new product launches close together (Ryzen, Kaby Lake, Skylake-X, etc) stretching manufacturers thin. Not that it excuses the lack of proper air cooling.

My EK waterblock is arriving today, so I hope to be operational no later than tomorrow.
 
Reactions: Drazick

hasta666

Junior Member
Aug 18, 2017
17
8
81
Asetek cooling solutions should be sufficient to run at stock speed. But I'm also waiting for a dedicated solution. I'm not gonna go cheap for the cooler with such a high-end platform nor buy something in a few months when more products are released. Not having received the motherboard helps ;-)
 
Last edited:

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
26,129
15,274
136
I'm getting close to a point where I'll be throwing the platform into a garbage bin or giving it few blows with a hammer.

The 3D printed mounting bracket failed miserably (incompatible due to the curvature of the heatpipes) and I still have no way to cool the CPU.
None of the resellers have any ETA for Enermax TR4 coolers or Noctua U14S-TR4 either (or anything else for that matter). The only thing available for this platform is still those insufficient POS AIOs made by Asetek (20+ different variants of them).
Few different waterblocks will be available shortly, but a custom loop is not an option for me due to the nature of the setup.

I'm amazed by the level of incompetence AMD has illustrated with this platform. It is basically no different than releasing a new CPU with no motherboards available for it, utterly useless.
Normally I would blame it on the cooler manufacturers, in this case the final thermal design specifications for this platform (SP3) has been available since September 2016. Therefore the only explanation I can come up with, is that AMD didn't inform the cooler manufacturers
about this platform until the release was at hand.

/Rant
I have been running an AIO for a week now. At 4.1 ghz it ran very hot (hit 97c max), but I only ran that for an hour for some benchmarks, Been at 100%load@3.9 the rest of the week@65c, no worries, and that the hottest of all cores per coretemp.

I know its not ideal, but just get one so you can play with it, and when good ones come out, then you can really OC and have fun.
 

MarkPost

Senior member
Mar 1, 2017
308
575
136
I'm getting close to a point where I'll be throwing the platform into a garbage bin or giving it few blows with a hammer.

The 3D printed mounting bracket failed miserably (incompatible due to the curvature of the heatpipes) and I still have no way to cool the CPU.
None of the resellers have any ETA for Enermax TR4 coolers or Noctua U14S-TR4 either (or anything else for that matter). The only thing available for this platform is still those insufficient POS AIOs made by Asetek (20+ different variants of them).
Few different waterblocks will be available shortly, but a custom loop is not an option for me due to the nature of the setup.

I'm amazed by the level of incompetence AMD has illustrated with this platform. It is basically no different than releasing a new CPU with no motherboards available for it, utterly useless.
Normally I would blame it on the cooler manufacturers, in this case the final thermal design specifications for this platform (SP3) has been available since September 2016. Therefore the only explanation I can come up with, is that AMD didn't inform the cooler manufacturers
about this platform until the release was at hand.

/Rant
About the Noctua coolers, are in stock at Noctua Shop:

http://stores.ebay.de/NOCTUA-Designed-in-Austria
 

ddogg

Golden Member
May 4, 2005
1,864
361
136
Looking forward to seeing temps with the EK.
The lack of proper waterblocks is really frustrating but I can live with it.
I think the XSPC block is very close to release so I'm anxiously waiting to pull the trigger on this build.
 

ub4ty

Senior member
Jun 21, 2017
749
898
96
I'm getting close to a point where I'll be throwing the platform into a garbage bin or giving it few blows with a hammer.

The 3D printed mounting bracket failed miserably (incompatible due to the curvature of the heatpipes) and I still have no way to cool the CPU.
None of the resellers have any ETA for Enermax TR4 coolers or Noctua U14S-TR4 either (or anything else for that matter). The only thing available for this platform is still those insufficient POS AIOs made by Asetek (20+ different variants of them).
Few different waterblocks will be available shortly, but a custom loop is not an option for me due to the nature of the setup.

I'm amazed by the level of incompetence AMD has illustrated with this platform. It is basically no different than releasing a new CPU with no motherboards available for it, utterly useless.
Normally I would blame it on the cooler manufacturers, in this case the final thermal design specifications for this platform (SP3) has been available since September 2016. Therefore the only explanation I can come up with, is that AMD didn't inform the cooler manufacturers
about this platform until the release was at hand.

/Rant
Don't let it get to you that much. Just get something to last you until proper cooling comes out. This is the lessor of the evils w.r.t to Intel's Skylake-X mayo paste under the heatspreader which can't be addressed. I've made the same observations and commentary before launch and it is being shown that I was correct. There seems to be a marriage of issues w.r.t to proper cooling solutions. That being said, Noctua was the only company that was 100% honest and frank w/ me. For that, even though I am not a fan of the non-contact heatpipe towards the outer edge of the contact plate, they will likely get my $$$. My e-mail exchanges with Noctua did in fact confirm that they extended the contact plate only and its not an ideal solution. Interestingly, they stated this very early on.
In further exchanges, they indicated that they haven't yet been sampled with a retail unit from AMD to finalize some figures :
http://noctua.at/en/tdp-guide?___from_store=de
Note the blank wattage fields for the threadripper units...
Noctua indicated that they'd be shipping units by the end of the month and it appears that is holding true. Be aware, if you're purchasing through their ebay page that it's shipping from overseas and you're paying a forex differential : $20 over retail.

As for AIOs on the market....
At best, they're temporary holdovers. Use them as such. They're using and screwing you after-all.I personally am opposed to them as I don't want a leak screwing up an expensive workstation and possibly critical data.

Custom loops ...
Price

IMO, what is the biggest revelation of doing a computer build in recent times is that America has extremely destroyed a significant aspect of the country/culture : It's manufacturing base/pride in quality production. Every single thing I touch is made in a foreign country w/ subsequent lead times, pricing, and quality issues. From power buttons, switches, wires, computer cases, water blocks, tubes, connectors, water pumps, fans, heatsinks... motherboard, AIB cards, electronics.

There used to be a thing called quality American Manufacturing and honest work and honest companies. Now, were left with questionable corner cutting, elaborate marketing lies, and an inability to access the means to produce better solutions in our own back yards. A highschool dropout who went to trade school can mill a custom water block CNC/CAD drawing from pushing a couple of buttons based on figures from a manufacturer provided schematic ... Thermal cooling capability is all done is software. Literal button pushing...

This crap isn't complicated nor is it expensive. Meanwhile, we have to rely on British/European parts that should be made in the U.S... Pay silly prices and wait weeks for the product to make it across the shores through retailers who add no value but markup. This mess has to stop. A whole generation is losing the ability to relate and contribute to a huge engine of productivity because of myopic/greedy business types who've sent all of the manufacturing overseas. What's left are a bunch of brain deficient marketing/business reps who know nothing about the product, how to serve their customers, and whom lie every chance they get with marketing speak (compatible coolers). We should have unique and quality computer parts made in the U.S. Someone who has an idea should be able to drive up the street and have it milled/made for them in a machine shop. Alas, instead, in the mass market we have to keep buying cut quality junk from overseas made by a generic shop who then stamps 20 different company's logos on it. It's likely that a large number of these AIO companies don't even have an engineering staff.

We need to start making things in america again... Engineered properly with care and sold via honest detailing. I've worked in the tech industry for some time and have seen the very foundation of it erode over the years as bean counters/business people sell this country and its future up river through offshoring/cut quality production aimed at doing nothing more than profiteering off of differentials. The results are :
> bad engineering (planned obsolescence) which leads to resources being wasted)
> Lies to consumers (marketing) to cover the truth about how the product is made and what its quality is
> And just all together b.s

You can't keep shoveling money to those who produce nothing of value and in some cases undermine it and expect to be able to build the future...

It's time for a shakeup among the lazy middle men skimming off the top and those who make a business from labeling overseas junk as a premium product.

There's my /rant.
 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |