• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."

Review Threadripper 3rd Gen Review and Availability Thread

Hitman928

Platinum Member
Apr 15, 2012
2,595
1,833
136
Since the other thread got merged and I am not the OP as well as the announcement thread being so long, I figured it would be better to have a separate review and availability thread.

Once reviews are posted I will update this post. If you see a review you would like added to this post, please PM me.

Video reviews
Hardware Unboxed
Gamers Nexus
Linus Tech Tips

Print reviews
Phoronix
PCWorld
PCMag
Guru3d
Tweaktown
Anandtech
Computerbase (translated)
Servethehome
Phoronix (3990x)
For sale
3960x, 3970x, and 3990x listed for sale on Newegg.
 
Last edited:
  • Like
Reactions: Tlh97 and Mopetar

EXCellR8

Diamond Member
Sep 1, 2010
3,735
669
126
Phoronix has some Linux stuff published, including how to boot with TR, if anyone is interested in that.
 

Panino Manino

Member
Jan 28, 2017
183
138
86
Oh, it's already up?I don't know that impress me the most, if the difficulties Intel is having or how well AMD is performing.
The jump from the last generation is something that we don't see in a long time!
 

DisEnchantment

Senior member
Mar 3, 2017
578
1,079
106
phoronix has by far the best review - many uses cases for such a powerful machines
the standard review sites need to upgrade their methodics for such core count
I am not so sure most of the Tech Tubers are aware/have had experience of other actual uses for such machines besides content production.
Not that most engineers, designers, devs etc needed much guidance anyway.
 
  • Like
Reactions: Tlh97

StefanR5R

Diamond Member
Dec 10, 2016
3,496
3,767
106
the standard review sites need to upgrade their methodics for such core count
Many reviewers get some of the samples only on short term loan. (This is getting increasingly common between those reviewers who are enabled to publish "0 day" reviews.) They are therefore unable to run new tests weeks/ months/ years later on the same CPU. This is certainly a major reason of their reluctance to evolve their tests.

In my opinion, reviewers should disclose to their audience whether or not they had to return the sample after test. After all it reduces the possibilities to reproduce (and thus to confirm or update) their results. Same as they should disclose whether they are testing an engineering sample or retail sample, on a beta BIOS or retail BIOS, and so on. Vendors love this sampling policy, and only more openness about it can remedy this.
 

moinmoin

Golden Member
Jun 1, 2017
1,664
1,610
106
In my opinion, reviewers should disclose to their audience whether or not they had to return the sample after test. After all it reduces the possibilities to reproduce (and thus to confirm or update) their results. Same as they should disclose whether they are testing an engineering sample or retail sample, on a beta BIOS or retail BIOS, and so on. Vendors love this sampling policy, and only more openness about it can remedy this.
computerbase.de writes some of this info at the end of each article, whether an NDA was involved, what it enforced, and which parts were loaned. The other info is there but hidden within the review text and screenshots, so there's still room for improvement even there.
 

TheGiant

Senior member
Jun 12, 2017
667
283
106
Many reviewers get some of the samples only on short term loan. (This is getting increasingly common between those reviewers who are enabled to publish "0 day" reviews.) They are therefore unable to run new tests weeks/ months/ years later on the same CPU. This is certainly a major reason of their reluctance to evolve their tests.

In my opinion, reviewers should disclose to their audience whether or not they had to return the sample after test. After all it reduces the possibilities to reproduce (and thus to confirm or update) their results. Same as they should disclose whether they are testing an engineering sample or retail sample, on a beta BIOS or retail BIOS, and so on. Vendors love this sampling policy, and only more openness about it can remedy this.
Sure
But that is not the whole story
I am not sure the typical reviewer has enough knowledge
running cinebench and be done with it is not enough here
ofc except the forum warriors
 

beginner99

Diamond Member
Jun 2, 2009
4,548
890
126
Indirectly about availability. AMD still has to fight uphill battle. A big one. Want to get a new workstation at work. However they say "policy is HP". HP however doesn't offer any workstation with AMD whatsoever. Intel only. Does anyone wonder why intel still has the marketshare it has? It's the big OEMs and big stupid company polices. I could get a very nice threadripper workstation for like $7000 with 32-cores and 128gb RAM. With HP? A 28-core xeon based one costs $25000 because that xeon alone is 15000.

EDIT: I hope that price difference will matter more than policy but not holding my breath. But just winning benchmarks is not enough for AMD.
 

Zucker2k

Golden Member
Feb 15, 2006
1,160
520
136
I could get a very nice threadripper workstation for like $7000 with 32-cores and 128gb RAM. With HP? A 28-core xeon based one costs $25000 because that xeon alone is 15000.
That's quite brutal! I suppose you can't overcome ten years of stagnation in a couple of years.
 

maddie

Diamond Member
Jul 18, 2010
3,161
1,837
136
Indirectly about availability. AMD still has to fight uphill battle. A big one. Want to get a new workstation at work. However they say "policy is HP". HP however doesn't offer any workstation with AMD whatsoever. Intel only. Does anyone wonder why intel still has the marketshare it has? It's the big OEMs and big stupid company polices. I could get a very nice threadripper workstation for like $7000 with 32-cores and 128gb RAM. With HP? A 28-core xeon based one costs $25000 because that xeon alone is 15000.

EDIT: I hope that price difference will matter more than policy but not holding my breath. But just winning benchmarks is not enough for AMD.
Once a narrative is internally fixed, we (humans), have such a hard time changing. Of course, all the voices arguing for keeping it (Intel) dominant, also play a role.
 

TheGiant

Senior member
Jun 12, 2017
667
283
106
Indirectly about availability. AMD still has to fight uphill battle. A big one. Want to get a new workstation at work. However they say "policy is HP". HP however doesn't offer any workstation with AMD whatsoever. Intel only. Does anyone wonder why intel still has the marketshare it has? It's the big OEMs and big stupid company polices. I could get a very nice threadripper workstation for like $7000 with 32-cores and 128gb RAM. With HP? A 28-core xeon based one costs $25000 because that xeon alone is 15000.

EDIT: I hope that price difference will matter more than policy but not holding my breath. But just winning benchmarks is not enough for AMD.
you can try this one https://ark.intel.com/content/www/us/en/ark/products/193752/intel-xeon-w-3275-processor-38-5m-cache-2-50-ghz.html

its like with red bull- you say energy drink? ofc red bull, but there are many others which are good
you will find very small % of corporate managers really do know what they are buying and even then they have the incentive to fight with the corporate policy
if it fits into the budget, I dont care

dont fight with it, buy the 4k xeon- its not better, but its not bad
 

beginner99

Diamond Member
Jun 2, 2009
4,548
890
126
Once a narrative is internally fixed, we (humans), have such a hard time changing. Of course, all the voices arguing for keeping it (Intel) dominant, also play a role.
I suspect with OEMs it's also simply saving money. The design their workstation: case. motherboard etc. and then keep it for several years. If you only offer intel xeons, you only need 1 mobo. They are still offering 3 year old cpus with these workstations...With threadripper they would need a new mobo and possibly new case. So they don't do it or it takes months after release till you can buy it.
 

beginner99

Diamond Member
Jun 2, 2009
4,548
890
126
fair enough, that specfic workstation probably was multi-socket hence the CPU price. HPs shops doesn't help in finding a good solutions. It's terrible.

EDIT: I now checked again and the onyl 28-core option that can be chosen from is the 15k one. w-3275 is not an option, in fact the w-3xxx series is limited to 8-cores. Only sane option available is the 6252 with 24-cores. But base clock is terrible at 2.1ghz.
 
Last edited:

KentState

Diamond Member
Oct 19, 2001
7,853
194
106
fair enough, that specfic workstation probably was multi-socket hence the CPU price. HPs shops doesn't help in finding a good solutions. It's terrible.

EDIT: I now checked again and the onyl 28-core option that can be chosen from is the 15k one. w-3275 is not an option, in fact the w-3xxx series is limited to 8-cores. Only sane option available is the 6252 with 24-cores. But base clock is terrible at 2.1ghz.
If you are looking at the 28 core Z8, that's a dual processor configuration.
 

KentState

Diamond Member
Oct 19, 2001
7,853
194
106
Looks like resellers are already jacking the price up. Only $100 more, but hate seeing this behavior one day in.

1574888885143.png
 

beginner99

Diamond Member
Jun 2, 2009
4,548
890
126
f you are looking at the 28 core Z8, that's a dual processor configuration.
The price is for ONE 28-core. If you want dual 28-core, you can add another 15k on top. Of course this is the platinum xeon which no one really buys and HP doesn't officially* offer the lower-end high core count xeons. I wonder why? (I don't, it's free money from the clueless).

* maybe depends on country or maybe you can get it if you are big enough customer and ask for it.
 

Ottonomous

Senior member
May 15, 2014
552
287
136
Dumb question, is it a legal minefield or do some companies allow experienced employees to build or configure workstations themselves?
 

jpiniero

Diamond Member
Oct 1, 2010
7,955
1,216
126
Dumb question, is it a legal minefield or do some companies allow experienced employees to build or configure workstations themselves?
It isn't as much of a legal minefield as it is supporting it if something breaks.

There's also the question as to how much validation has been done on ECC with Threadripper even though it is enabled.
 

ASK THE COMMUNITY