• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

FRONTPAGE Threadripper 2 Review Compilation (2990WX and 2950X)

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
Right! This CPU should have a space reserved in Ripley's Believe it or not. Such a monstrosity of a CPU it's almost cartoon like. Wonder what the CPU landscape is going to look like in 20 years if AMD keeps going at this rate. 256 cores? 512 Threads? 128Mb cache?

And Intel's 10 nm will be just around the corner then...LOL
 
For the 2990WX in the video he was speculating that maybe this time next year AMD will have a significantly better processor using the 7nm process. However, 7nm node has nothing to do with processor architecture. A shrinking node allows more transistors to fit in the same prior die size and use less watts also allow higher clock speeds.

He was also speculating that if windows scheduler was improved that would help the poor performance levels with the 2990WX.

Yeah, I get what a node is/does. :)

Node size has little to do with it (gaming performance). Sure, clockspeeds and voltages we all (should) know can effect performance, but, a node itself doesn't really hold anything back directly. TR2 is reaching similar clocks to previous gen and getting raked over the coals in some titles (its not a gaming CPU really...can't hold that against it). This in and of itself should lend users to believe it isn't in the clocks, but in the arcitecture (CCX and intercommunication between them/windows scheduling, etc). I have never heard a process node being attributed to holding back gaming performance before. Maybe someone wrote that... I dont know.

Feels like ganged and unganged HDDs honestly. :p

Where did I say node size has anything to do with application performance.;) I said 7nm node has nothing to do with processor architecture. Architecture is physical structures of the processor like X86, cores, cache, memory controller and interconnects. AMD dropped the ball on the 2990WX architecture they could have made it do well in all applications including gaming and I'm betting in the future 32c 64t will do better in all applications like the 2950X. I don't believe folks that have the knowledge want to purchase two High end PCs one for gaming and good productivity and one for improved productively depending of the program or programs running. The software is so selective on the 2990WX for performance improvement and reductions compared to the i9-7980XE that does most everything well. The Video pointed out the 2990WX for the public has mixed performance results advised purchase 2950X. In reality what choices of software running is going to make 2990WX worth it?
 
The question above your early post (from gin) references node size. You replied, and since nothing was quoted, i thought your post was supporting his assertion on node size and you referenced gaming. ;)
 
Last edited:
How true AMD is doing much better. Intel only doped median target price from $61 in May to $58 in August.
AMD stock rose due to strong earnings growth. The earnings growth was driven by the demand for the company’s Radeon GPUs (graphics processing units) and Ryzen CPUs (central processing units). The processors gained popularity because they offered competitive performance at a relatively lower cost.
 
Last edited:
Intel ain't going broke. LOL I was just pointing out how fickle perception can be. There is nothing inherently "wrong" with Intel's product right now. There are a couple things AMD chips do better, though not necessarily faster overall. They're just stuck living my personal philosophy for the time being:

Every day above ground is a win. Some days, that's all you get.

Intel has a remarkably large and talented pool of Very Smart People working to keep them firmly in the game, and they're now very motivated. I expect that to work in their favor eventually. LOL
 
In one of those reviews and I can't wait for ours :) they had a CB-15 test. I think the 2990WX scored 85xx or some insane number. I remember when we where Benching the 14c/28t ES Intel CPU's in Philly. It looked like it took the CPU less than 3 seconds to finish the test. Just a simple 1..2..3.. Done. So does this now mean that CB-18 will be a timed score?? So when benching the 2990WX with CB-15 would you even get to 1 second???
This is a question on power. How much power does an 8 pin AUX CPU connector handle?? I know the spec on an 6/8 pin PCIe power connector, that's 75w/150w.
 
The 8 pin EPS +12 volt power cable 336 watts + The 24 pin +12 volts main power connector 144 watts = 480 watts.
 
I'm almost tempted to do my bottle neck test in linux now. But with different games.

And to make sure, the scheduler is basically what is responsible for how the OS utilizes the CPU cores right?


 
I would like to see how linux compares to windows 10 in gaming benchmarks. Thanks for the Video link.
 
i was going to post this as its own thread but then saw this one.
A Look At The Windows vs. Linux Scaling Performance Up To 64 Threads With The AMD 2990WX

I'm almost tempted to do my bottle neck test in linux now. But with different games.

And to make sure, the scheduler is basically what is responsible for how the OS utilizes the CPU cores right?


yea that phoronix article is saying windows scheduler is not properly setup for a high amount of cores.
This past week we looked at the Windows 10 vs. Linux performance for AMD's just-launched Ryzen Threadripper 2990WX and given the interest from that then ran some Windows Server benchmarks to see if the performance of this 64-thread CPU would be more competitive to Linux. From those Windows vs. Linux tests there has been much speculation that the performance disparity is due to Windows scheduler being less optimized for high core/thread count processors and its NUMA awareness being less vetted than the Linux kernel. For getting a better idea, here are benchmarks of Windows Server 2019 preview versus Ubuntu Linux when testing varying thread/core counts for the AMD Threadripper 2990WX.
 
We can throw **** at AMD all day for the ridiculousness of a 32c/64t CPU today, but isn't that where CPUs are heading?
Dennard scaling seems to have broken down since about 2006. It's my belief that as a method of getting more powerful processors, we won't see single core performance doing much until the next paradigm shift. I think more cores are the way to go; so.. more cores = better? it might take a while for software to catch up, but I think we're finally seeing concurrency taken more seriously, and some 32core processors on the market are just what we need to give the consumer software multi-threading snowball a push down the hill.

On a side note, while overclocking much past 4GHz is going to be tough, what's to stop someone disabling half or 3/4 of the CPU and just overclocking 16 or 8 cores to get the best of both worlds only a reboot away?

I think this thing is absolutely fantastic on *almost* all fronts.

Edit: Never mind about disabling cores. I googled it. I think no is the answer.
 
Last edited:
We can throw **** at AMD all day for the ridiculousness of a 32c/64t CPU today, but isn't that where CPUs are heading?
Dennard scaling seems to have broken down since about 2006. It's my belief that as a method of getting more powerful processors, we won't see single core performance doing much until the next paradigm shift. I think more cores are the way to go; so.. more cores = better? it might take a while for software to catch up, but I think we're finally seeing concurrency taken more seriously, and some 32core processors on the market are just what we need to give the consumer software multi-threading snowball a push down the hill.

On a side note, while overclocking much past 4GHz is going to be tough, what's to stop someone disabling half or 3/4 of the CPU and just overclocking 16 or 8 cores to get the best of both worlds only a reboot away?

I think this thing is absolutely fantastic on *almost* all fronts.

Edit: Never mind about disabling cores. I googled it. I think no is the answer.

well from the link i posted in post #55, it looks like downcore control in the Asus UEFI allows you do that. my take away from that picture is you can choose cores or core+smt, but not being able to see the bios my self. i hope those are not the only options, that would be really limited if the highest selection was 3c+3smt for 6 threads. i know someone is going to correct me..
 
well from the link i posted in post #55, it looks like downcore control in the Asus UEFI allows you do that. my take away from that picture is you can choose cores or core+smt, but not being able to see the bios my self. i hope those are not the only options, that would be really limited if the highest selection was 3c+3smt for 6 threads. i know someone is going to correct me..

This is how I understand the article

[oldtable]Setting|Threads w/ SMT|Cores;Threads wo/ SMT|Cores/die
Stock|64|32|8 Cores/die
3+3|48|24|6 Cores/die
4+0|32|16|8 Cores/die, 2 dies disabled
2+2|32|16|4 Cores/die
3+0|24|12|6 Cores/die, 2 dies disabled
2+0|16|8|4 Cores/die, 2 dies disabled
1+1|16|8|2 Cores/die
1+0|8|4|2 Cores/die, 2 dies disabled
[/table]

I think the written number in the BIOS refers to how many half-dies? are enabled. If each die is essentially made of two smaller 4 core dies put together it makes sense, or if there's a 16c/32t CPU in the socket and they don't change the list entry strings for the 2990WX; or can't and expect most people to be running 16c/32t.

Edit: I think single core over clocking is limited more by the architecture than anything that disabling other cores could help based on some google searches. I think you're up against a wall with single core performance, not to say I think it's at all bad.
 
Last edited:
Back