• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD Threadripper 8Pack Elite Tier bundles and Threadripper testing.

OcUK Staff
Joined
20 Feb 2012
Posts
10,163
Location
John Smiths Stadium
Hi all,

This thread is to show some Threadripper Performance testing, some temperature and stability testing and to show how different memory configs vastly impact on Threadrippers efficiency and performance. Its also here to show 24-7 overclocking and indeed promote our range of excellent upgrade bundles designed to get superb performance in a plug and play upgrade solution.

Much of the information contained in this post is available in video form by following this link:
https://www.youtube.com/watch?v=LrG30TZMQS8


and how to install an AIO cooler by Asetek on the TR4 Threadripper socket is covered by the following video: https://www.youtube.com/watch?v=FpfZvk9lkEU

Before we go into my bundles lets examine the impact of higher speed memory and dual channel vs quad channel on Threadripper Performance.

Firstly we have several benchmarks with the memory running in Dual Channel mode (Only two sticks installed) C14 3200mhz With the CPU overclocked to 3.8ghz.







Next we have the same benchmarks with the memory this time in quad channel 2666mhz C14 and the CPU at 3.8ghz







and finally we have the CPU at 3.8ghz the memory in quad channel at 3200mhz C14






What can be summarised from the above is that firstly high speed memory is more important across a wide range of benches all of which simulate different tasks a CPU may be used for. The high speed memory is more important than using 4 sticks in quad channel infact and dual 3200 beats quad 2666 across all benches. In my own view the main reason for this is the higher effective cache frequency of using high speed mems. On Threadripper just as Ryzen previously apart from core speed this cache frequency is the most important factor for getting good IPC out of the CPU. We see in all benches the quad channel option with high memory speed (3200mhz) is the best for performance but quad channel showed the biggest gain in performance if your using your PC for rendering type tasks as opposed to gaming.

3200mhz CL14 - 3333mhz CL14 on Samsung B die like the 8Pack dark pro for me was most compatible across all boards I tried as well as being the best performer.

The performance pre-overclocked bundles I have put together for AMD Threadripper launch are as follows:

Threadripper 1950X or 1920X CPU overclocked to at least 3.7ghz on all cores (I will optimise this for each individual CPU for the customer so if the CPU is capable of more stable it will ship with higher OC)
ASUS Zenith X399 MB or AsRock X399 Taichi MB
32gb 3200mhz 8Pack Team Dark Pro memory with 64GB option.
Dual Asetek cooler with mounting bracket with triple AIO and custom water as further options.
Silverstone water cooling fans with great static pressure.
Thermal Grizzly Kryonaut thermal paste.

So here is some stability and temperature testing of the Asus bundle with the 1950X CPU on the included cooler.



And here is the performance you can expect for the 1950X 3.7ghz 3200mhz CL14 ASUS combination.






Here we have some stability testing of the 3.7ghz 1950X AsRock Bundle.




1950X has a built in offset of 27C which is similar to that of Ryzen where we saw 20C. ASUS bios is already compensating for when reading but AsRock is not so the 110C seen here is really an OK 83c.

Here is the performance of the AsRock based bundle using a 1950X at 3.7ghz.







Threadripper scales with both cooling and voltage but in my mind for 24-7 use on anything but custom water 1.35v should be your highest Vcore used to stabilize any OC. All chips I tried could do 3.7ghz hence why these bundles start at that but I am overclocking them so I will max out everything on the given CPU you receive so if it can do 3.9ghz or 4ghz stable that is that you will receive. The 4ghz stable chips are very very rare. Almost all are 3.7 or 3.8 max stable clocks with reasonable voltage. With our bundles we obviously give the customer the options for better cooling and with it the chance of a better overclocking result.

I am currently testing 4x16gb configuration for 64gb and 8x 16gb for 128gb which I will add to these bundles in the very near future once they are fully validated.

Any questions about the bundles or testing do let me know.

Buy the bundles here:
ASUS:https://www.overclockers.co.uk/detail/index/sArticle/69042

AsRock: https://www.overclockers.co.uk/detail/index/sArticle/69043
 
64gb GSKILL testing 4x16 3066mhz C16 any higher needs tuning and is not easy for the standard end user. Only limited coverage I know but will leave running overnight along with 128gb Testing of the same mems.
 
Thanks for the test results!!
No problem. a lot of reviews I see posting 4ghz etc. I think maybe some cherry picking has gone on here. None of our stock samples tested went this high without going for high volts and temps 90C+. Like I said realistically 3.7+ is what people can expect at around 1.3-1.35v for Prime and Real bench stability testing. Many CPUs could also hit 3.8 but not all within what I call acceptable thermals. Obviously custom loops with triple rad configs will get that extra 100mhz or so from the CPU.
 
No problem. a lot of reviews I see posting 4ghz etc. I think maybe some cherry picking has gone on here. None of our stock samples tested went this high without going for high volts and temps 90C+. Like I said realistically 3.7+ is what people can expect at around 1.3-1.35v for Prime and Real bench stability testing.

A lot of reviews might also be ramping up the voltage and doing a suicide run on their test CPUs,as we all know after proper stress testing it might not actually pass.
 
AMD naming their HEDT platform X299.... sorry X399 is a real tongue-twister isn't it? :D

Its good to see an OCUK'ers face and some reviews, keep it up :)
 
4x16gb testing now................


No all core are at the same speed. I estimate we ship almost all these bundles at 3.8ghz.

I also have 128gb running now for overnight stability.
 


128gb on the test!!! I will do some performance stuff tomorrow with OC CPU and 128gb config providing its tested as stable overnight.
 
Not really a gaming or low threaded application CPU though is it?

But you dont solely run highly threaded applications even on a HEDT system either. It's not simply one or the other.

Are you saying someone who both plays games and does heavy CPU tasks should buy 2 systems, each suited the task?
 
Last edited:
But you dont solely run highly threaded applications even on a HEDT system either. It's not simply one or the other.

There is a reason for CPU's with a lot of cores, people buy these CPU's for Multithreaded performance rather than single threaded performance.

Few applications they might use that are not multi-threaded the performance difference is about 10%, thats nothing like enough to give up the lower power consumption, lower heat and upto 40% higher performance they get in their primary use of the CPU.

Seriously arguing the core i9 is better because it has a bit batter performance in just that one aspect out of all the others where its often far worse is straw clutching of epic proportions.
 
But you dont solely run highly threaded applications even on a HEDT system either. It's not simply one or the other.

Are you saying someone who both plays games and does heavy CPU tasks should buy 2 systems, each suited the task?
How many games are are so CPU dependant that the difference in clockspeed will significantly affect fps but that also only support up to 4 threads?

Also, if you're spending £1000 on just a CPU, then maybe you can afford a separate 7700k gaming system.
 
How many games are are so CPU dependant that the difference in clockspeed will significantly affect fps but that also only support up to 4 threads?

Also, if you're spending £1000 on just a CPU, then maybe you can afford a separate 7700k gaming system.

HEDT systems are fine for gaming. Not everyone is trying to run at 1080p with less than maximum settings and trying to achieve 144fps, in fact I'd say very few are. Like probably 90% of gamers I'm happy with 60fps at 1440p and above at maximum settings.

Why waste another £1k on a "gaming" system which will in reality give no difference in gaming experience over the HEDT system you already have.
 
Back
Top Bottom