The Ryzen "ThreadRipper"... 16 cores of awesome

[urvile]

So much for all the FUD. Hopefully the 1920X won't be to far behind it's big brother......

 

[Ajay]

So much FUD every which way lately. Can't wait till TR reviews Thursday to shut some of it down.

 

[IndyColtsFan]

I'm optimistic for sure, but that IS a Wccftech link, so I'll wait for some reputable reviews first.

 

[Osjur]

I'm optimistic for sure, but that IS a Wccftech link, so I'll wait for some reputable reviews first.

Which took the pics from reddit, where one guy (callingthewolf) actually has the CPU and ran the benches. That is 4.1Ghz OC on TR with shitty 3066mhz ram (8 slots populated though) with even shittier timings.

Its legit.

 

[texasti89]

Simply WOW! . Designing a brand new x86 architecture from scratch while achieving this high performance objective at one go is marvelous. Hat's off to AMD's engineers. Can't wait to see what Zen-2 iteration has to offer.

 

[jmelgaard]

OMG, this is one thing that is almost unbelieveable: "Not even the world record Cinebench R15 run of the 7900X at 5.96GHz with liquid nitrogen cooling is enough to surpass Threadripper’s score"(from the link above). And thats with 3066 memory, not 3600, and only at 4.1 ghz !!!!

Really?... Is that so unbelievable?... ...

 

[IndyColtsFan]

Which took the pics from reddit, where one guy (callingthewolf) actually has the CPU and ran the benches. That is 4.1Ghz OC on TR with shitty 3066mhz ram (8 slots populated though) with even shittier timings.

Its legit.

Thanks, but I am well aware of where they got the source material. I'll wait for other reviews before celebrating.

 

[ub4ty]

@Pandora's Box
Do I see that you are only running about 3000 on the ram ? If so image what my 3600 will do. !!!

Given what I observe on x370 boards w.r.t to when you fully populate all slots, I doubt you can push such clocks with all 8 slots populated without significantly diminished CL settings. Per memory controller, things seem best suited for 2 sticks. We'll see shortly I'm sure and it's one of the things I am looking forward to seeing proofed in reviews before I make a RAM purchase decision.

 

[SPBHM]

as far as I know Cinebench is typically not very affected by memory speed.

 

[ub4ty]

Interesting article from Nexus about thermal paste application methods on 1950X. It seems its not a trivial thing with actual "compatble" AIO. OC is at 4.0ghz all cores btw

http://www.gamersnexus.net/guides/3013-amd-threadripper-thermalpaste-application-methods-benchmarked
https://www.youtube.com/watch?v=nWu2tcm4wL8

So these "compatible" AIO make the job decently (they use a Kraken x62). But I think that to push these babies to the limit, new and specific coolers are a need. I am interested to see too the Noctua TR air coolers perfomance. IMO it will be better (at least with the 140 fan model) than with these AIO.

Yep, it's exactly as I stated and warned about. Furthermore, I'm aware of the exact diameter a circular contact plate has to be to cover the active dies completely. Thus, I have a standard to measure the coolers by. Given that a rectangle doesn't optimally fit inside a circle, I am sure the more sound cooling solutions will have rectangular contact patch to match the active die configuration. At a minimum, it needs to be sized across the active dies with more optimality if they cover the whole heatspreader. As such, i find all current offerings to be temporary solutions until a quality contact plate is specifically engineered for Threadripper.

And hot off the presses we have :
https://www.techpowerup.com/235950/...ch-tr4-aio-liquid-cooler-for-amd-threadripper

Others less heard of names are also getting in on the game :

I feel a little better now that my commentary is clearly shown to be sound and also that I will get a purpose built and uncompromised cooler for my new Threadripper. I am very thankfully for Steve at GamerNexus for being critical and thorough in his reviews and critiques showing the obvious problem with the recommended compatible coolers. Looks like we'll be getting proper AIOs quite soon as competition is no doubt going to drum up now that there are proper offerings

 

[ddogg]

That Bykski waterblock looks really nice. I've never used their products before though.

 

[ub4ty]

That Bykski waterblock looks really nice. I've never used their products before though.

It does but be careful with off brand wares.There were comments about the unspecified metals they use which could wreak havoc w.r.t to metal mixing (degradation/corrosion) w/ other components in your cooling loop.
Interesting to see Enermax beat others to the punch. Time for others in the cooling scene to get on the ball !

 

[ddogg]

It does but be careful with off brand wares.There were comments about the unspecified metals they use which could wreak havoc w.r.t to metal mixing (degradation/corrosion) w/ other components in your cooling loop.
Interesting to see Enermax beat others to the punch. Time for others in the cooling scene to get on the ball !

Someone on the overclock.net forum got one so I'm curious to see his thoughts on it. I have mostly barrow hardline fittings in my loop and they are very high quality so I'm open to trying stuff but "unspecified metals" made me cringe a little I will likely pick up an EK block if it goes on sale tomorrow.

 

[dlerious]

Bitspower and EK should have waterblocks. I tried the configurator on EK and it had the Zenith Extreme listed under motherboards, but it didn't show any blocks yet. Don't know if they're waiting until tomorrow or not.

 

[ub4ty]

Bitspower and EK should have waterblocks. I tried the configurator on EK and it had the Zenith Extreme listed under motherboards, but it didn't show any blocks yet. Don't know if they're waiting until tomorrow or not.

That or end of the month seems to be the deal w/ coolers...
Lots of AIO/custom water blocks and a number of air coolers are slated for the end of the month as opposed to manana.

Seems this is potentially do to some kind of 2 week embargo placed on custom jobs vs. their (compatibility) list. The timing is just too perfectly synced.

Also, there's something about these full coverage custom blocks that get me just the slightest bit moist

 

[MarkPost]

glad to see that Enermax TR AIO, now we are talking. The rest of brands will have to offer similar specific products, no doubt

 

[Pandora's Box]

I really don't think that not covering the entire heat spreader is that big of a deal. AMD wouldn't have sent thermaltake 360mm aios with review kits if there was a problem

 

[StefanR5R]

Interesting article from Nexus about thermal paste application methods on 1950X. It seems its not a trivial thing with actual "compatble" AIO. OC is at 4.0ghz all cores btw

http://www.gamersnexus.net/guides/3013-amd-threadripper-thermalpaste-application-methods-benchmarked
https://www.youtube.com/watch?v=nWu2tcm4wL8

So these "compatible" AIO make the job decently (they use a Kraken x62). But I think that to push these babies to the limit, new and specific coolers are a need. I am interested to see too the Noctua TR air coolers perfomance. IMO it will be better (at least with the 140 fan model) than with these AIO.

Yep, it's exactly as I stated and warned about.

You warned that Asetek-type AIOs will perform decently?

Here is what I understood from GamersNexus' article and video:

• The Asetek preapplied TIM, which covers only half of the area of the dies, performs just as well as the various other methods of TIM application which were tried by GamersNexus, which cover almost all of the area. (Only his "big blob" method stood apart from the other methods.) This can only mean that Threadripper's heat spreader conducts horizontally very well.
• All methods of TIM application which GamersNexus tried, Asetek's preapplication included, resulted in workable, sustainable die temperatures when the CPU was not overclocked.
• When GamersNexus overclocked, all differences of thermal performance of the different TIM applications vanished because his slim 280 mm radiator became the bottleneck. Temperatures weren't good, but even this combination of imperfect cover by the cold plate + small radiator + OC'd 16 core CPU did not result in thermal throttling, according to GamersNexus.

BTW, the photos of GamersNexus' TIM imprints seem to me as if TIM layer thickness increased notably towards the periphery of the cold plate. This would be in line with [H]'s claims that a lapped cold plate would be better suited to Threadripper than Asetek's standard convex cold plate. I wonder if the curvature of the cold plate's contact area is actually of greater importance than its outline.

 

[.vodka]

I wonder if the curvature of the cold plate's contact area is actually of greater importance than its outline.

It seems to be the case, since the hotspot to cool isn't right in the middle of the plate, but in two opposite spread out points.

A waterblock with a bigger plate and the micro channels distributed along its entirety should be better than a regular one...

Still, it should get the job done.

 

[Markfw]

I will let you know Monday (later) if the AIO block on a 360 rad will cool it enough./

 

[csbin]

http://www.coolaler.com/threads/amd-ryzen-threadripper-1950xcineben-r15-4122.346417/

AIDA64 Memory Bandwidth

1950X@4.0GHz

128GB(8*16GB) DDR4 3200 19-19-19 QUAD Channel AIDA64 Bandwidth

 

[.vodka]

Nice numbers for vomit inducing 19-19-19-53, at least for DDR4-3200. Latency could be so much better with CAS16-14 memory there...

Insane performance at its worst.

 

[formulav8]

Yeah 105ns latency isn't too well for the imc. But most else is ok I guess. I got to 82.3ns before the program expired when on a 1.0.0.4 agesa bios. Its not worth it to register the program for my purposes to recheck my latest settings. I should decently be into the 70ns area's now with my Hynix @ 3066 14/16/15 timings. But not 100% sure because of needing to pay now for the bench and recheck. I wish they would do a cheapo lite version in the future.

 

[beginner99]

This will be an uphill battle for AMD. Local shop just released TR prices. While the Intel 7900x is priced like msrp $999, TR 1950x and 1920x both are priced roughly $200 higher than msrp meaning it's actually the 1920x going against the 7900x on price...add to that the motherboard costs of TR (Asus Zenith is $650) and the 7900x is the clear winner on performance/$

 

[ddogg]

This will be an uphill battle for AMD. Local shop just released TR prices. While the Intel 7900x is priced like msrp $999, TR 1950x and 1920x both are priced roughly $200 higher than msrp meaning it's actually the 1920x going against the 7900x on price...add to that the motherboard costs of TR (Asus Zenith is $650) and the 7900x is the clear winner on performance/$

Are you in the US? Because here it's 999 and the the Zenith is 550. Also you can purchase a taichi for $330 if you don't want to fork that much money on a motherboard. Not sure what uphill battle you're talking about.