Replacing dual X5670?

chx

Gawd
Joined
Jun 21, 2011
Messages
698
This platform might be aging but the video transcoding performance of it is really nice. However, the single thread performance is becoming harder and harder to tolerate by each passing day. What to replace with without losing video transcoding performance and gaining on single thread performance while keeping an eye on the price, too? I posted elsewhere and people recommended the Threadripper 1950X... I bet! I also bet that quad Xeon Platinums would be faster too! Please mind the price. Thanks :)

Ps. I am posting in the Intel forum because Hackintoshing might be in the cards. So Intel is preferred for that reason alone.

Ps2. userbenchmark is useless in here, look at this http://cpu.userbenchmark.com/Compare/Intel-Xeon-X5660-vs-Intel-Xeon-E5-2680-v2/m17750vsm17083 this claims a 6 Cores, 12 Threads @2.8 GHz Westmere EP is significantly faster than a 10 Cores, 20 Threads @2.8 GHz Ivy Bridge EP and the 2690 v2 is only 4% faster.
 
Last edited:
No. I am looking for a price efficient solution and then I will figure out whether I could afford it. The big question is: does this need a HEDT or are we there yet where a single desktop CPU can replace two very old hexacore CPUs?
 
You said you don’t have a budget, but nixed the Threadripper. So clearly you do have a budget. 2700(x)?
 
Is the 2700/2700x as fast as a dual X5670 in video transcode aka multicore?
 
Is the 2700/2700x as fast as a dual X5670 in video transcode aka multicore?

Ryzen 2700/2700x are faster than dual X5670 even factoring overclock.. and it will smoke it in single thread. but if you wanna go big leagues without breaking the bank, go with a threadripper 2950X when available or a cheaper 1950X
 
I also vote 2700x for bang for your buck horsepower. I have an overclocked ryzen 1700 and it does what I need for video work. If you have a project you'd like me to run through media encoder or handbrake to comp times let me know.
 
https://hackintosher.com/guides/amd-ryzen-hackintosh-guide-installing-macos-high-sierra-10-13/

HO-HUMM!

320 USD? That's more reasonable than a Threadripper, not to mention the motherboard. And the RX 580 is slowly returning to saner price levels... Thanks, there's some beancountering to be done :)

Yeah, Ryzen 2700x seems to BARELY match the performance of your existing dual-CPU in pure multi-threaded performance while offering much better single-threaded performance for newer games.

The way I did the comparison:

an i7 980x is clocked 15% faster than your Xeon.


The 5820k is 30% faster than he 980x in Handbrake

The Ryzen 2700x is 30% faster than a 5820k in Handbrake.

1.15 the performance of your Zeon to get to the 980X, then 1.3 to get form 980x to 5820, then 1.3 to get from 5820 to 2700x:

1.15 * 1.3 * 1.3 = 1.95 time the performance of one of your Xeons.

This link shows me that the Coffee Lake Core i7 8700kl is slightly faster than the Ryzen 2700x at Handbrake, and is a little faster at games as well.

https://www.anandtech.com/bench/product/2109?vs=2110
 
Last edited:
  • Like
Reactions: chx
like this
Thanks for the amazing work! This is a very exciting time then -- it took this long but finally the top end of the desktop CPUs can match the video transcoding performance of that old Mac Pro while delivering a much better experience in almost everything else (almost everything is still mostly single threaded).
 
Yeah, the factory-overclocked i7 8700k or the Ryzen 2700x were made for exactly users like you. Have you decided which one you will go with?
 
I do video work too....the 4k shit was killing my old rig.

Here's what I have done : (so far)

https://hardforum.com/threads/project-snowballed.1950723/

--------
Dual 8 core Xeon Golds (skylake)
Cost (as of right now) : 16,000$ CDN
I gotta tell yah it SMOKES the 4k.(Hardware HEVC)

Cost a lot , but it's better than I hoped so ,

(y)
One of my 1080ti can encode h265 4k faster by like 90% than 16 thousand dollar rig using NvEnc with nearly the same quality.

CPU vs. GPU in the quality arena is old news. NVidia has really caught up in that department.
 
Oh sure.
For me 'rendering speed' doesn't matter most of the time. I hit the go button and goto bed.
It's not about rendering speed, it's about "workflow" speed.
When I am actively editing , my video cards are busy running the 4k monitors.I want the cpu/s to handle on the fly decode.
I'll probably have Premier Pro , Photoshop and aftereffects all running at the same time,along with several misc programs.

My old OC'd 3930k/32 gb ram used to choke on it.The new Xeon rig is butter smooth regardless of what I throw at it.

(y)
 
Oh sure.
For me 'rendering speed' doesn't matter most of the time. I hit the go button and goto bed.
It's not about rendering speed, it's about "workflow" speed.
When I am actively editing , my video cards are busy running the 4k monitors.I want the cpu/s to handle on the fly decode.
I'll probably have Premier Pro , Photoshop and aftereffects all running at the same time,along with several misc programs.

My old OC'd 3930k/32 gb ram used to choke on it.The new Xeon rig is butter smooth regardless of what I throw at it.

(y)

Buttery smooth is always super nice. But man did you really spend 16k on that machine.. ouchie
 
Buttery smooth is always super nice. But man did you really spend 16k on that machine.. ouchie

Yep.
Now , That's 16,000$ CDN (call it 10 grand US), and I'm well over $18k at this point....I still have 7 SSD's to add , a pair of 1180's lets say.....and that's not counting all the gear to make my shiny new socket P waterblocks work.
:D

A LOT of cash, yes. I'm very happy with my purchase though. I have never built/setup/used a nicer machine !!
-

All that being said :
If you *kinda* have a budget for a video editing rig , I think a threadripper or even an Epyc, is where you want to go.Personally, I don't like the current Intel single chip solutions.

-Don't buy anything with less than say 48 PCIe lanes....I run 2 separate NVMe ssd's as source drives in addition to c: with 2 video cards, that's 44 without even trying.
Threadripper does 128, my Xeons do 88 + NVME with separate lanes.
-Don't do less than 4Gb memory per core.

AMD or Intel , spend the cash for a new, professional rig.....you get to play with all the current tech.Xeons/Epyc's use stuff you will never see in prosumer gear.

(y)
 
Yep.
Now , That's 16,000$ CDN (call it 10 grand US), and I'm well over $18k at this point....I still have 7 SSD's to add , a pair of 1180's lets say.....and that's not counting all the gear to make my shiny new socket P waterblocks work.
:D

A LOT of cash, yes. I'm very happy with my purchase though. I have never built/setup/used a nicer machine !!
-

All that being said :
If you *kinda* have a budget for a video editing rig , I think a threadripper or even an Epyc, is where you want to go.Personally, I don't like the current Intel single chip solutions.

-Don't buy anything with less than say 48 PCIe lanes....I run 2 separate NVMe ssd's as source drives in addition to c: with 2 video cards, that's 44 without even trying.
Threadripper does 128, my Xeons do 88 + NVME with separate lanes.
-Don't do less than 4Gb memory per core.

AMD or Intel , spend the cash for a new, professional rig.....you get to play with all the current tech.Xeons/Epyc's use stuff you will never see in prosumer gear.

(y)
Definitely I agree on your recommendations. I'm toying with a 32core TR. Dont want to spend epyc.
 
I have the CPU's you're talking about in my server. I am presently shopping for the HPDL380 take a look at the ebay server search https://www.labgopher.com/ is your best bet if you're really wanting cheap performance. I'm shooting for dual 8 or 10 core chips but may want to go up another gen.
 
I know this post is a little late, but thought you might find this interesting. I transcoded a 1 hour 55 minute movie to 1080p mp4 using the High Profile, 2-pass at 3000 kbit/s with Handbrake 0.9.9.5530 on my new i7-8700K set-up. I used this version of Handbrake because it's the version I used when I had my dual Xeon X5650 set-up. The i7-8700K took about 1 hour and 15 minutes to complete the transcode whereas the dual X5650 took 1 hour and 31 minutes to finish. I know you have X5670 CPUs, but this comparison should still help some.

hb1.png
 
  • Like
Reactions: chx
like this
The i7-8700K took about 1 hour and 15 minutes to complete the transcode whereas the dual X5650 took 1 hour and 31 minutes to finish.

This shows a single i7-8700k is basically on par with a dual X5670. Nice. Thanks! And it's not late. I am working for a friend here.
 
Depending on how far you want and can push the x56xx series, there are x5690s out there. They run a little hotter but should wake up some performance based on the single thread number:
https://www.cpubenchmark.net/compare/Intel-Xeon-X5670-vs-Intel-Xeon-X5690/1307vs1314

That would probably put you right there to the 8700k encode time in the test above if not better. And it seems a pair of x5690s runs about half the cost of a new 8700k. (y)
 
Back
Top