Review: AMD Ryzen Threadripper 1920X

Jagman

2017-08-10 15:23

Both this and the 1950X are as expected....Awesome for the money but I do agree with you Hilbert this one needs to be 749 to be priced correctly as it were. Great reviews as always 🙂

#5461416

WhiteLightning

Moderator

2017-08-10 15:40

Its hammer time

Contentteller The MySQL database of this Contentteller Content Management System installation has encountered the following problem: The MySQL database server is not reachable Please try it in a few minutes again. We apologize for any inconvenience.

https://media.tenor.com/images/0fa00a479139acebfaf51c885db3db4e/tenor.gif Epic CPU's. These are great for streamers (who just want one rig to do it all).

#5461420

Aura89

2017-08-10 15:45

So, basically, for $1000, if you go AMD instead of Intel, you can get a much, much faster processor, or for $200 less, you can still get a decently faster processor then going Intel. Yup.

#5461442

Guru3dreader

2017-08-10 16:07

Yes you'll be able to perhaps gain 10% more perf at 1080p in a CPU bound situation from a faster clocked Intel counter-part. How important that is to you is totally up-to you. My advise, take the money you save on the AMD platform and invest that in a faster GPU.

I would love to see this. A 12 core Threadripper at $799 plus a GTX 1080 Ti, against an 7900X at $999 with a GTX 1080. Thanks for this review.

#5461444

chispy

2017-08-10 16:07

Awesome review as always Hilbert , thank you for the in depth review ! Absolutely amazing performance , great to see AMD back in the top. Finally the Intel monopoly has been broken and that is a good thing as everybody wins here.

#5461521

Agent-A01

2017-08-10 18:36

H, can you test numa vs the other mode in AIDAI64 cache/latency test? Curious how much of a difference it makes, Intel can be less than 40ns depending on ram.

#5461543

Hilbert Hagedoorn

Administrator

2017-08-10 19:04

H, can you test numa vs the other mode in AIDAI64 cache/latency test? Curious how much of a difference it makes, Intel can be less than 40ns depending on ram.

I already did? http://www.guru3d.com/articles_pages/amd_ryzen_threadripper_1950x_review,20.html But I also made the remark that AIDA needs to be updated. The numbers are not yet 100% right. UMA (Distributed) would site in the 90~100ns and NUMA (local) in the 60~70ns ranges.

#5461550

alanm

2017-08-10 19:22

Thanks HH, nice review. Is there an error in the Firestrike chart p27? The 7700k surely cant be doing that good vs all these high end multi-cores in this test?

#5461564

Hilbert Hagedoorn

Administrator

2017-08-10 20:30

Thanks HH, nice review. Is there an error in the Firestrike chart p27? The 7700k surely cant be doing that good vs all these high end multi-cores in this test?

Firestrike is pretty sensitive to high CPU frequencies (which the OC already shows, far more so than Time Spy. The 7700k is doing well there thanks to its 4.5 GHz boost.

#5461607

Agent-A01

2017-08-10 22:18

I already did? http://www.guru3d.com/articles_pages/amd_ryzen_threadripper_1950x_review,20.html But I also made the remark that AIDA needs to be updated. The numbers are not yet 100% right. UMA (Distributed) would site in the 90~100ns and NUMA (local) in the 60~70ns ranges.

Oops I'm blind. Thanks 🙂 Wonder if this would help those games that are poorly threaded, AKA old UE3 games and the like.

#5461722

The Edge

2017-08-11 08:59

I'm wondering why is there a 13 FPS difference between the 1950X and 1920X in Hitman 1080p in favor of 1950X when the stock 1920X has even higher base clock ?

#5462014

Aura89

2017-08-11 22:58

I'm wondering why is there a 13 FPS difference between the 1950X and 1920X in Hitman 1080p in favor of 1950X when the stock 1920X has even higher base clock ?

Additional cores? The 7900k wins out to a 7740k even though the 7740k has higher clocked speeds as well

#5462080

Taint3dBulge

2017-08-12 05:36

Thanks for posting that video of the 1920x running BF1. That is interesting that the game can handle up to 12 threads but seems like only 4 really get used used more then 25%, and 1 of the 4 is at 100%. But then again, that 1 core is prolly handling alot of other things on top of the game. So that makes me wonder, why doesnt it push more processing power to a core that is only being 30-40% utilized? Anyways thanks again for spending time doing that.

#5462135

The Edge

2017-08-12 12:19

Additional cores? The 7900k wins out to a 7740k even though the 7740k has higher clocked speeds as well

Makes no sense. It gets beaten by all those Ryzens 7 and 5 even overclocked to 4 GHz and has way more cores.

#5462146

Hilbert Hagedoorn

Administrator

2017-08-12 13:03

Yeah I remember running the test several times, each time with the same outcome. No clue, I can only note down what I measure. With upcoming X399 reviews I'll again revisit and re-check that though.

#5462170

The Edge

2017-08-12 14:48

Yeah I remember running the test several times, each time with the same outcome. No clue, I can only note down what I measure. With upcoming X399 reviews I'll again revisit and re-check that though.

I thought it might be the difference between distributed and local RAM mode. But then I thought you tested both threadrippers in the same mode. Strange...

#5462248

Aura89

2017-08-12 21:12

Makes no sense. It gets beaten by all those Ryzens 7 and 5 even overclocked to 4 GHz and has way more cores.

Makes perfect sense. The game apparently supports additional cores, but is not linear in terms of how much extra performance those cores give you.

#5462384

The Edge

2017-08-13 11:54

Makes perfect sense. The game apparently supports additional cores, but is not linear in terms of how much extra performance those cores give you.

Man are you looking at the same charts as I do ? Your theory makes no sense, which even Hilbert just confirmed above. The 1920X is loosing to fricking Ryzen 5 - 6 core and it is loosing to it even when overclocked.

#5462398

WaroDaBeast

2017-08-13 12:46

Nicely written and thoroughly tested. Thanks, Hilbert. Also, thank you for the feedback to other forum members. It's always nice to be able to discuss benchmark results.