Intel Core i9-9980XE vs AMD Ryzen Threadripper @ [H]

Interesting review, thanks Kyle!
Will be very interesting to know this. Funny seeing nearly no OC left on both teams at this level.

There is OC left with the Intels and you can get a 4.5Ghz on all cores with the 7980XE or go per core and adjust. You NEED to delid and void warranty though and need Liquid Metal (not paste) or get rid of the IHS completely + very good custom watercooling.

The 9980XE seems to be worse temp wise compared to a delided 7980XE so the so called "solder" is only good enough for the stock clock increases. And yet, deliding the 9980XE is riskier than the previous SKUs.

It looks like I am staying with the 7940X (no point getting a new SKU at a loss) and will upgrade again once PCI-E 4/5.0 , DDR5 + Ice Lake is out.
 
I'd need like access to a swiss bank account to order any more than that at intel's prices lol
 
There is OC left with the Intels and you can get a 4.5Ghz on all cores with the 7980XE or go per core and adjust. You NEED to delid and void warranty though and need Liquid Metal (not paste) or get rid of the IHS completely + very good custom watercooling.

The 9980XE seems to be worse temp wise compared to a delided 7980XE so the so called "solder" is only good enough for the stock clock increases. And yet, deliding the 9980XE is riskier than the previous SKUs.

It looks like I am staying with the 7940X (no point getting a new SKU at a loss) and will upgrade again once PCI-E 4/5.0 , DDR5 + Ice Lake is out.

I can't see any enthusiast getting the 9xxx over the 7xxx SKLx chips. The 7980xe has way more thermal headroom when overclocked.

I am thinking your 7940x will achieve a very similar multicore performance when deluded and overclock when compared to a 9980xe with max overclock using reasonable cooling.
Of course your 7940x will be faster in lower thread count loads.

Pretty bad when these products are not only competing performance wise with AMD, but also older version of themselves.
 
Thanks Kyle for that registry screenshot. It shows that you have the Microsoft patch which now includes the microcode updates (something people were manually doing with a vmware utility back in January), but also shows that the ASUS firmware is out of date. I doubt most of the boards out there even have the second set of spectre fixes yet...

But this does mean that any performance impacts of those mitigation's are already a part of the benchmark results.
I can only use what is available to me. I did check with ASUS for that board before we started. That said a new update just dropped for AMD so all my AMD benchmarks from this review are now out of date....just found that out getting ready to run DLM TR tests. See we do actually check. :) You would not actually believe how many sites do NOT update EFI or reuse benchmarks from previous articles without even checking so they don't have to do as much work and make deadlines easy for them.
 
I can't see any enthusiast getting the 9xxx over the 7xxx SKLx chips. The 7980xe has way more thermal headroom when overclocked.

I am thinking your 7940x will achieve a very similar multicore performance when deluded and overclock when compared to a 9980xe with max overclock using reasonable cooling.
Of course your 7940x will be faster in lower thread count loads.

Pretty bad when these products are not only competing performance wise with AMD, but also older version of themselves.

Absolutely no reason to "upgrade" since there is no upgrade in practice - only the name is upgraded and in the end you might end up worse or with high risk for delidding them if it is even possible. It's the end of the road for Intel's current architectures it seems.

I am also working on a 9900K build for the fastest single core performance possible but this will be after Christmas.
 
Judging by the power to performance numbers, it looks like these are still 14 nm and not something more advanced like 14++. Am I correct on this?
 
If you are a content creator that also uses the same machine to game on and want the highest FPS possible then get the Intel chip.

If you are only content creating/workstation type workloads and looking to save a few bucks get an AMD chip.

If you are a youtuber and want fastest rendering times get the Intel chip.

I would actually argue against this. If you want fastest render times I would get the 32 core AMD chip. Why? Because I think the results are so askew as to be an anomoly and the issue will be patched.
 
Judging by the power to performance numbers, it looks like these are still 14 nm and not something more advanced like 14++. Am I correct on this?

Another review pointed out this is the exact same chip as previous generations they just did a solder on heat spreader as opposed to the TIM they were using before and OC'd it a few more MHZ. Not even a newer generation CPU in truth.
 
I'm not sure about y'all. But my budget usually targets the low end, not the top of the stack. If you look at the costs, the basin falls CPUs on the low end are competitive with AMD's low end HPC. Lower cores but faster clocks and better IPC. I'm waiting to see some benchmarks.

View attachment 119789
Even ignoring for a moment that "budget" and "HEDT" don't go well together:

The lower core count SKL-X (up to 10-core) have been made obsolete by the 9900K, unless you need the PCIe lanes.
If you are on a budget, consider the TR 1920X as it is very cheap right now. Newegg has the 1920X / ASUS X399 Prime combo for $679 currently. Amazon.com has the 1920X for $414.99 on back-order.
 
As an Amazon Associate, HardForum may earn from qualifying purchases.
Well going to say the same thing I've been saying recently. This is a smoke and mirrors CPU built on 14mm process and Intel is blowing smoke up people's asses. Come on man If you already have the Intel 7980XE well the 9980XE is its brother. there is little point of getting an expensive CPU when you already have one... oh about 2 or so years ago. Smoke and Mirrors... pure bullshit.

Note: I normally do not agree with Linus Tech Tips but there are cases that he does make some sense. Video below.

 
I love the 1440p game test to show something more real world.

I wish I had a reason for one of these so I can have an excuse for exotic cooling. 100C at overclock. Sounds like we need to sand a few mm of silicon off the die. ;)

It’s a nice side effect for TR that the dies are spread out decreasing the heat density. I look forward to the 7nm products.
 
Anyone have an idea how this generation of CPU (specifically the 9900x) would compare to a 6950x?

Little off topic: My Silicon Lottery 6950x currently overclocks to 4.7GHz on all cores. I also have a 7700K (de-lid, re-lid, + lapped die) that runs at 5.2GHz. Trying to figure out if it makes sense to make a move to this generation for gaming only system at 1440p....
 
Anyone have an idea how this generation of CPU (specifically the 9900x) would compare to a 6950x?

Little off topic: My Silicon Lottery 6950x currently overclocks to 4.7GHz on all cores. I also have a 7700K (de-lid, re-lid, + lapped die) that runs at 5.2GHz. Trying to figure out if it makes sense to make a move to this generation for gaming only system at 1440p....
I have a 6950X in my current box. It runs wide open encoding 4K video with ease at 4.3GHz. Given the two extra cores, in a content creation regard, the 6950X I think would still outperform it handily. That said, in CPU-limited gaming, the 9900K would win out due to its high boost clocks. I say stick with what you got.
 
If in a CPU limited gaming situation... any idea how close in clock speed the 6950x would need to be to 9900k to make performance equal? I assume gaming performance IPC in 9900k is still negatively affected vs. Broadwell-E like all Skylake-X processors:
https://digiworthy.com/2017/07/11/intel-skylake-x-gaming-broadwell-e/
I think you are really splitting hairs at that point. If you need the best possible gaming, get a new 9000 and clock it. I just got a 9600K in to play with also.
 
It's not the WX processors that are broken with certain workloads, but the Windows 10 scheduler. AMD's DLM tries to work around that (not sure how well), but if/when Microsoft fixes their scheduler to properly handle the different NUMA nodes, there should be a marked increase in performance with things like Premiere. Of course, it's also possible for Premiere to fix it at the application level, but it would take some work - essentially doing several things that should be done by the scheduler.
 
It's not the WX processors that are broken with certain workloads, but the Windows 10 scheduler. AMD's DLM tries to work around that (not sure how well), but if/when Microsoft fixes their scheduler to properly handle the different NUMA nodes, there should be a marked increase in performance with things like Premiere. Of course, it's also possible for Premiere to fix it at the application level, but it would take some work - essentially doing several things that should be done by the scheduler.
Do you have proof of this? It is not as if we have not discussed this with Adobe and AMD looking for answers.
 
Should be easy to test, just test it on linux? Assuming there is a ver of app on linux. That should tell us a little more. There were some linux kernel updates I hear for 32core amd, maybe there will be more. Or they all will ignore it now that the epyc7nm will have different IO system so maybe its an old stuff no1 care about.
So your answer is, "No, I really am just making up what I am saying, but really I know....in my world." Got it.
 
It's not the WX processors that are broken with certain workloads, but the Windows 10 scheduler. AMD's DLM tries to work around that (not sure how well), but if/when Microsoft fixes their scheduler to properly handle the different NUMA nodes, there should be a marked increase in performance with things like Premiere. Of course, it's also possible for Premiere to fix it at the application level, but it would take some work - essentially doing several things that should be done by the scheduler.

while the scheduler might be a slight problem it's not the entire problem.. the one advantage linux has is having far less overhead than windows does.. encoding while it is better in linux shows the exact same performance gains for the 7980XE so that's not specifically scheduler related that's just flat out OS/software differences. when specifically looking at adobe related tasks i think the problem revolves around all 3 of them(Adobe, Microsoft, AMD). that being said DLM doesn't fix everything, it improves a few things that are heavily memory dependent tasks and hurts the things that aren't.

i think the 2990WX is cool as hell and AMD has some massive cajones for releasing it at a consumer level but there are just some things the design just can't perform at the same level as other things and people just have to accept that. just my opinion though.
 
Should be easy to test, just test it on linux? Assuming there is a ver of app on linux. That should tell us a little more. There were some linux kernel updates I hear for 32core amd, maybe there will be more. Or they all will ignore it now that the epyc7nm will have different IO system so maybe its an old stuff no1 care about.

No, there is no premiere for linux
 
what Dariuz said make some sense. a simple test could overthrow it and know for sure. rather you could simply say hardocp isnt going to do the test but instead choose a phrase that to some people might be offensive to them. make sense why i unsubbed u guys and remove my pledge, its okay, you won't be missed as theres always anandtech and tomshardware.

now, before you go all rage, do you see how a silly statement like that can be offensive? and if you're not offended, thats good because you have more capacity as a person than i thought.

Hey man, how's the view from your high horse? Anyway, what the guy said he had 0 proof of, if you make a claim, at least substantiate it.

Enjoy Toms and Anandtech
 
Very nice comparison. Makes me realize that not upgrading my 1st Gen Core i7, and only buy new vid cards was a sound choice :giggle:
 
I see what you did there. I guess those of us hanging out in "Small Form Factor Systems" don't count, heh. But in all seriousness, nice review.
Hi ceski, sorry all for the necro/ offtopic post, speaking of "Small Form Factor Systems" .. what happened to you bro? PM me, I had some questions about some of your hosted images in that forum, they've been lost..! I was hoping to contact you about a re-upload of them :cry:
 
https://www.anandtech.com/show/1385...-2-performance-and-windows-scheduler#comments
The timeline for a fix will depend on a number of factors between AMD and Microsoft, however there will be announcements when the fix is ready and what exactly that fix will affect performance. Other improvements to help optimize performance will also be included.
I did speak with AMD at CES about this. They plainly stated that the ball was fully in Microsoft's court on this. This is the first time they have confirmed to me that it is not a problem they can solve on the AMD side.
 
Back
Top