Amd Ryzen Threadripper & X399 MegaThread! FAQ & Resources

Page 3 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.

jdwii

Splendid


Take a look at the gaming results only same story as ryzen except not as much as ryzen due to the fact that its running in quad channel vs dual
 

liberty610

Distinguished
Oct 31, 2012
466
5
18,815
I myself am waiting on my Kraken X62 from NZXT for my Threadripper setup. I have everything else here, so the CPU is already seated in the socket on the board. I went with the Gigabyte Aorus Gaming 7 board, since I have had disastrous results with the MSI boards, and the Asus was just a little to much price wise. Plus, I have had great success with Gigabyte boards.

I'll be upgrading my current build to Threadripper hopefully tomorrow. Pretty excited to get this thing built. Have head headaches with the ordering process since Amazon dropped the ball on my pre-order. Had to make a rush check on Newegg to make sure I could get everything in time.

This is the current system I will be putting it in:
https://pcpartpicker.com/b/BqHhP6
 

liberty610

Distinguished
Oct 31, 2012
466
5
18,815
Oh, and sorry for breaking up the tech junkie debate :)

I just thought I would throw out there that I liked what I saw in threadripper, and went for it. I do a lot of video with Vegas Pro/Handbrake, and I do a lot of digital audio stuff as I ruin a small project studio. Kinda excited about getting this machine up and running, despite the Intel/AMD debates.
 

juanrga

Distinguished
BANNED
Mar 19, 2013
5,278
0
17,790


If we consider only gaming results the performance increases 6% on average. Zen gets higher gains because doesn't have the power trotting mechanism than ThreadRipper has.
 

jdwii

Splendid


Not breaking up anything this is a threadripper thread after all haha, tell us how you like it and how much faster it is then what you are currently using.
 

Solarion

Prominent
Jul 6, 2017
66
0
660


At the risk of feeding an obvious troll. I don't suppose you have any evidence of this alleged "power trotting" mechanism at work do you? Is this going to be like the last time, in this very thread, I requested proof of one of your wild claims and you went silent? Don't get me wrong, I find your posts to be quite comical, just curious if you have anything at all to substantiate your position...this time. If you mean simply that a particular board is set to bone stock settings and is enforcing a 180w TDP, then that's fine, but if you're trying to suggest that there's some AMD hocus pocus going on at the chip level then...yeah...gonna need some proof bud.
 

Solarion

Prominent
Jul 6, 2017
66
0
660


Debates are good though! Even when obvious trolls are obvious. lol

I have my own x399 gear on the way too and looking forward to my own testing...particularly with regard to encoding, rendering, and cryptography...where threadripper is an absolute monster. It'll be a week or so for me though, at least beyond basic power on testing, because I ain't pushing it too hard without a real water block and I haven't chosen one yet.

Cheers.
 

liberty610

Distinguished
Oct 31, 2012
466
5
18,815


Well, if you remember from my other thread lol... water blocks and I have not been on good terms this week lol. Damn Corsair reps not telling me the bracket on the H115i was removable... I'd be up and running right now if it wasn't for that haha. Oh well. I learned something new for sure, as it was my first water cooler.

I too am really excited to see what kind of numbers I can get on the Threadripper setup. I am currently using a 6800k Braodwell-E with 32gigs of G.Skill, but my current ram is slower timing then the new kit I just got.

When rendering at stock speeds on the 6800k and just using the XMP profile on the ram, I was able to render down a 30 minute full HD video with a second audio source, mild color grading. several cross-fade transitions, and a few audio mastering plugins in on it in about 1hr 14 minuets. It rocks about every core at 90 to 100% the entire time. The video is 1920 x 1080 60fps @ 50 mbps. So they are good sized files. I shot them on my Sony NX100 camera.

So now that I have ram render timings on it, I can see what Threadripper will do to it. I also plan on running Handbrake tests today to compare it to. I am hoping my Kraken x62 gets here tomorrow... I have seen solid reports about it's cooling for Threadripper, and I really don't wanna wait for the full TR4 blocks to release.
 

jdwii

Splendid


Hey as a question did you go with air cooling? I'm wondering as I really like the noctua options where it actually covers the whole IHS.
 

liberty610

Distinguished
Oct 31, 2012
466
5
18,815


It's mine understanding that Solarion does not have his Threadripper build done yet. Having said that, he's waiting for a water block for it, so I guess he is not doing air cooling.

There are water blocks coming out that cover the whole IHS, but they have not been released yet I don't think. In the video I posted above, Steve (gamer nexus) has a whole video about Threadripper, thermal past application for it, and what is the best coverage for current water blocks that do not cover the whole IHS.

From the research I have done, air cooling MUST be done with a TR4 designed fan cooling kit, but a lot of the AIO water coolers are good to go as is. I am assuming that a full water block that is designed for TR4 sockets that will cover the whole IHS will be better then the current ones that do not. But, the current water coolers do keep it in check as far as temps go.

You can see Gamer Nexus's fuill video about the thermal past application here:
https://www.youtube.com/watch?v=QpvGYxaMLc0&t=212s

Skip ahead to 3:35 and he talks about coverage on the entire IHS plate.
 

jdwii

Splendid
You might find 3:25 in the video steve does this and spreads the thermal compound around the AIO that is probably what I would do not sure in this video or another video I watched they showed like 5C drop from just doing that, which is why I personally think a NH-U14S TR4-SP3 would beat a Kraken X62. But I do understand people may want to get something better looking.

https://www.youtube.com/watch?v=Pxeaamdwty4
 

liberty610

Distinguished
Oct 31, 2012
466
5
18,815
On the topic of this with the AIO and Threadripper.... I have the Corsair Graphite series 760t, can be seen here:
https://www.newegg.com/Product/Product.aspx?Item=N82E16811139055&cm_re=760t-_-11-139-055-_-Product

Now, with Threadripper not getting 100% coverage with any AIO cooler, including my Kraken x62 that is coming, I am kinda wondering if I would benefit a pull method for the fans instead of a push?

When i had a Corsair h115i for this case, it was mounted at the top, as I already have 2 good fan coolers in the front for my drives and what not. There is no way I could mouth a rad in the front.... that entire area is all hard drives (I have 5 total).

So, I mounted the Corsair to the top and had the fans pushing air out off the CPU. With Threadripper not getting full coverage over the IHS, should I reverse the fans and pull air IN from outside the case?
 

letsrun4it

Prominent
Jul 25, 2017
76
0
660
I just want to commend the bro who poster the Threadripper RAM speed that's really good info we did not know before the release. I bought 32GB of Corsair Platnium Dominator Ram 3600 and I'll be trading that into my Ryzen 1800x build and take the 32GB of Corsair Vengence 2400 and put those in the Threadripper.
 

juanrga

Distinguished
BANNED
Mar 19, 2013
5,278
0
17,790


AT review:

The key element to this graph is the 1950X running at DDR4-3200. Because the faster DRAM requires the memory controller to draw more power, it leaves less power for the CPU cores, potentially resulting in a lower turbo core frequency. So while the faster memory might guarantee faster performance in memory limited scenarios, the core frequency might end up lower given worse performance overall. It’s an interesting thought, so we plotted the per-core power for the 1950X at DDR4-2400 and DDR4-3200.

HFR review:

We could see such a throttling under several applications on the 1950X although it is far from systematic, applications using AVX seem the most affected, logically. The limitation in our case, after many checks was not thermal, but related to consumption. AMD confirmed to us that such a throttling was theoretically possible without confirming completely. The 1920X did not exhibit such behavior in our tests, logically enough.

[...]

Moving under the base frequency is however something annoying, even if it is not the first time that we see this behavior at AMD.

This power throttling mechanism makes sense considering this TR4 socket has been derived from the SP3 socket used in EPYC. The internal codename for the TR4 socket is "SP3r2".
 

liberty610

Distinguished
Oct 31, 2012
466
5
18,815
Ah. I stand corrected than. I haven't gotten a chance to read the entire review. I just skimmed over it.

I just got this ram kit to go into my gigabyte aorus x399 Gaming 7 motherboard. It's 3200kHz with xmp enabled. I'll see what kinda hundreds I get today hopefully.

In rubbing an Intel 6800k right now, and I've done rendering in handbrake and Vegas Pro. I'll see what the numbers look like with Theater and post the results this week.just for a real world use kinda thing.
 

liberty610

Distinguished
Oct 31, 2012
466
5
18,815
Something to note if any Threadripper builders pan on using the NZXT Kraken X62 and the Gigabyte Aorus x399 Gaming 7 board... The pump hoses/block sticks out, and you could potentially loose a ram slot, preventing you using all 8 dimms. Unless you rotate the block, making the NZXT logo go side ways.
 

jdwii

Splendid
A review from newegg said this on threadripper 16 core 32 threads which again makes me say Handbrake doesn't scale perfectly beyond a certin amount of cores i know i've been using this software for years

"Applications need better scaling with this many cores. Handbrake at highest settings still only use 40-69% utilization. That being said, I never have to worry about processing power being available."

If its only at 70% usage or so that means no Handbrake isn't taking full advantage of all the cores and threads that threadripper has


That is why the 10 core Intel CPU is as close as it is to the 16 core threadripper same with adobe
 


Not shocking. As I've noted before, we (Software Engineers) figured out scaling starts to fall off a cliff after eight cores. You quickly run into diminishing returns due to various overheads.

And we figured this out in the EIGHTIES.

CPU architecture is really not designed with massively parallel workloads in mind; too much scheduling overhead, memory access contention, and other issues to overcome. That's why GPUs are more ideally suited for those types of workloads, since you can get the same benefits of parallelization without crushing the CPU/OS with more threads than they can handle (as few as one thread can achieve better utilization).
 

Solarion

Prominent
Jul 6, 2017
66
0
660


Nope, I don't think air cooling will be the way to go with threadripper. As was mentioned I'm assembling a custom loop with an Enermax Neochanger pump/res, 420mm radiator, and (3)140mm Fractal high pressure Venturi fans. Should end up being nice and quiet. Leaning toward the XSPC Raystorm NEO sTR4 block at this point, but it won't be available till the end of the month. Till then I have a woefully inadequate Cooler Master LM 120 AIO to at least check out the parts for DOA as they come in. The board and processor arrive Thursday.

11411_xspc%20raystorm%20neo%20threadripper.jpg


Handbrake only uses about 70% of the 24 cores on my current 2P rig as well. Premiere/Media Encoder does a little better, but doesn't use all cores either, so I usually just launch a parallel task to chew up the rest of the clock cycles.
 


Gamerk, please, don't confuse programs-scalability with algorithm-scalability. It's a bad thing to mix them up.

Two counterpoints: weather simulation and web request processing.

The software has to be designed around scaling across multiple CPUs at the higher level as well as the lower level. That is to say, you can make software parallel in different parts of its design. I don't think I'm talking something alien here by saying you can use "heavy threads" and "soft threads" wherever you want and however you want in your design. Monolithic designs for the industry are long gone and you have clustered (for the lack of a better term) software pieces that do usually scale up to 100s of CPUs with no problem. I know in my world they do just fine for ETL tasks when you're processing millions of millions of records per day.

Cheers!
 

Solarion

Prominent
Jul 6, 2017
66
0
660


No links? ...again? Where's the context to descriptions of this alleged "problem"? Why do you keep posting quotes without links? Now to expose your slanted nonsense one must waste time figuring out where you assembled your cherry picked quotes before picking your poorly researched positions apart. I don't even know what an "HFR", is, so to figure out where you ganked your likely out of context quote, I have to go waste time searching for where you claim to have received your info.

Again, are you talking about a specific motherboard throttling to maintain a TDP spec? That's nothing new and isn't fundamental to the CPU architecture. Intel platforms do precisely the same thing...unless you think an AVX offset is there to sound kewl. Spoiler alert...AVX offset is a throttling mechanism due to the huge power draw with AVX-512. If all you're talking about is motherboard specific implementation to maintain a specific TDP then it's a huge nothing burger, but your posts are long on innuendo, speculation, hyperbole and short on specifics and context. LINKS man. See...like so...

I won't lie, the past two weeks with the Core i7 7900X and the X299 boards I have tested have been a challenge. Memory XMP profiles would not stick, power consumption with one BIOS was OK, the other through the roof. But, most of all, the processor performance was all over the place. We've seen perf differences of up-tp 20% in-between just different motherboards.

MSI however, with their latest BIOS seem to have found a state of equilibrium. My biggest fight over the two weeks was ironically game performance, it was severely lacking. Example: a platform like this should run Rise of the Tomb Raider at ~140 FPS at 1080p on a GeForce GTX 1080. We'd end up at 90~100 FPS. And that problem occurred with pretty much all games. I have been discussing this with the motherboard partners (as yes, it is widespread) and we all agreed, it has everything to do with 'hardware P states' that Intel recommends to leave enabled. Intel recommends certain power states to keep the TDP in line. For most overall tests that worked fine, but the toll on game performance was abysmal. Days before this launch, MSI however released and provided a new BIOS, this restored the performance to what it needs to be. But as you have been able to see, the power consumption is certainly a bit on the high-side. In the end though, the performance is there, but we do expect several BIOS updates that will have an effect on performance overall, in gaming and on power-consumption.

http://www.guru3d.com/articles_pages/msi_x299_gaming_pro_carbon_ac_review,29.html

Let me know if you need any number of examples of Intel sky-x chips exceeding their TDP or being choked to maintain them. There are plenty of examples to choose from.

This wasn't even new architecture from Intel and had a ton of problems at launch. Heck, the bird poop TIM is still a problem for sky-x chips and that will only get worse when the rest of the chips finally appear. AMD's Zen on the other hand is in its infancy and while there have been some growing pains, particularly with Ryzen 7 at launch, the product is clearly capable of providing all the competition chipzilla can handle at a given price point.
 

juanrga

Distinguished
BANNED
Mar 19, 2013
5,278
0
17,790


HandBrake.png


The TR 1950X is noticeably faster than the TR 1920X. Both have the same IPC, and the 1950X has slightly lower clocks. The reason why the 1950 is faster is because Handbrake scales up above 12C.

The 10C SKL is so close to the 16C TR and beats often the 12C TR, because Intel has both IPC and clock advantage.
 
Status
Not open for further replies.