Overclock.net › Forums › Industry News › Rumors and Unconfirmed Articles › [HeXus]Legendary CPU architect Jim Keller leaves AMD
New Posts  All Forums:Forum Nav:

[HeXus]Legendary CPU architect Jim Keller leaves AMD - Page 28

post #271 of 284
Quote:
Originally Posted by Sanmayce View Post

Oh, fellow countryman or rather cityman, zdrasti.
Yes, those little scamps (Intel Silvermont, AMD Carizzo) are the near future of the laptops and desktops, just in latter case clustered heavily.
As for the dirty games, don't know, you seem even more angry than me.

On a similar note, few days ago Google announced their compression library Brotli targeting tight textual compression & fast decompression - yes, thus boosting the whole Internet browsing, but guess what, as in the case with the poor companies, inhere people, the little man (LzTurbo) has little chance of survival, simply the megacorporations dictate the course of the user-experience by imposing their policy. I am AMD fan for a reason, always supporting the underdog, there is no style&honor to go with the upperdogs charging with inflated price-tags mad.gif.

I wonder how AMD power-efficient chips execute Brotli&Shifune:
c1.png

I asked on guy from Thailand to run it:
http://www.cnx-software.com/2015/09/24/brotli-compression-algorithm-combines-high-compression-ratio-and-fast-decompression/#comment-455825

The thing is that Google made their speed tests on Xeon E5-1650 v2, no idea wwhy when most people browse from laptops, naturally the most used platform should be Carizzo & Atom, anyway, my point is that there are always better&cheaper alternatives - LzTurbo vs Brotli and Zen vs Haswell.

Eager to see how AMD fares in such a tightly optimized decompression etude:
Code:
; 'Nakamichi_Shifune_branchfull' decompression loop:
; Size in bytes: 93-15+2=128
; Size in instructions: 38
; mark_description "Intel(R) C++ Intel(R) 64 Compiler XE for applications running on Intel(R) 64, Version 15.0.0.108 Build 20140";
; mark_description "-O3 -QxSSE2 -D_N_XMM -D_N_prefetch_4096 -D_N_Branchfull -FAcs";

.B7.3::                         
  00015 45 8b 1a         mov r11d, DWORD PTR [r10]              
  00018 41 f7 c3 0f 00 
        00 00            test r11d, 15                          
  0001f 74 4d            je .B7.5 
.B7.4::                         
  00021 44 89 d9         mov ecx, r11d                          
  00024 ba ff ff ff ff   mov edx, -1                            
  00029 83 f1 03         xor ecx, 3                             
  0002c c1 e1 03         shl ecx, 3                             
  0002f d3 ea            shr edx, cl                            
  00031 44 23 da         and r11d, edx                          
  00034 44 89 da         mov edx, r11d                          
  00037 c1 ea 04         shr edx, 4                             
  0003a 48 f7 da         neg rdx                                
  0003d 48 03 d0         add rdx, rax                           
  00040 48 8b 0a         mov rcx, QWORD PTR [rdx]               
  00043 48 89 08         mov QWORD PTR [rax], rcx               
  00046 48 8b 4a 08      mov rcx, QWORD PTR [8+rdx]             
  0004a 48 89 48 08      mov QWORD PTR [8+rax], rcx             
  0004e 44 89 d9         mov ecx, r11d                          
  00051 83 e1 03         and ecx, 3                             
  00054 41 83 e3 0f      and r11d, 15                           
  00058 ff c1            inc ecx                                
  0005a 41 83 c3 03      add r11d, 3                            
  0005e 48 8b 52 10      mov rdx, QWORD PTR [16+rdx]            
  00062 4c 03 d1         add r10, rcx                           
  00065 48 89 50 10      mov QWORD PTR [16+rax], rdx            
  00069 49 03 c3         add rax, r11                           
  0006c eb 22            jmp .B7.6 
.B7.5::                         
  0006e 41 c1 eb 04      shr r11d, 4                            
  00072 44 89 da         mov edx, r11d                          
  00075 41 83 e3 0f      and r11d, 15                           
  00079 f3 41 0f 6f 42 
        01               movdqu xmm0, XMMWORD PTR [1+r10]       
  0007f 41 ff c3         inc r11d                               
  00082 48 83 e2 0f      and rdx, 15                            
  00086 f3 0f 7f 00      movdqu XMMWORD PTR [rax], xmm0         
  0008a 48 03 c2         add rax, rdx                           
  0008d 4d 03 d3         add r10, r11                           
.B7.6::                         
  00090 4d 3b d0         cmp r10, r8                            
  00093 72 80            jb .B7.3 
*/

Well, zdrasti as well smile.gif

I am not really angry, but the amount of $/performance for the Intel is .... disappointing.

The overall information on the library seems good, but i wasn't able to understand the results, can you explain them ? Also, a comparison would be good, even if its a server chip vs laptop one, we can still get some information out of it.
Desktop PC
(22 items)
 
ASUS ROG G72GX
(6 items)
 
 
CPUMotherboardGraphicsGraphics
AMD FX-8320 ASRock Fatal1ty 990FX Killer Sapphire Nitro+ RX480 Sapphire R9 290 Tri-X 
GraphicsRAMHard DriveHard Drive
XFX RX470 Singlefan Mushkin Redline 996996 2x4GB 2133Mhz Maxtor 6Y080L0 80GB 7200 RPM 8MB Western Digital 160GB 7200RPM 8MB 
Hard DriveHard DriveHard DriveOptical Drive
Maxtor 250GB 7200RPM 8MB Corsair Force LS WesternDigital Blue 500GB 7200RPM 16MB ASUS DVD-RW 
CoolingOSOSOS
ThermalTake Frio Silent 14 Windows 10 Enterprise Linux Mint 17.3 Rosa OphCrack 
MonitorKeyboardPowerCase
ASUS VS228HR Logitech K120 Corsair VS650 ThermalTake View 27 
MouseAudio
Bloody V5 Corsair HS30 Raptor 
CPUMotherboardGraphicsRAM
Intel Mobile Core 2 Duo P8700 G72GX NVIDIA GeForce GTX 260M  Hyundai Electronics  
RAMRAM
Hyundai Electronics  Hyundai Electronics  
  hide details  
Reply
Desktop PC
(22 items)
 
ASUS ROG G72GX
(6 items)
 
 
CPUMotherboardGraphicsGraphics
AMD FX-8320 ASRock Fatal1ty 990FX Killer Sapphire Nitro+ RX480 Sapphire R9 290 Tri-X 
GraphicsRAMHard DriveHard Drive
XFX RX470 Singlefan Mushkin Redline 996996 2x4GB 2133Mhz Maxtor 6Y080L0 80GB 7200 RPM 8MB Western Digital 160GB 7200RPM 8MB 
Hard DriveHard DriveHard DriveOptical Drive
Maxtor 250GB 7200RPM 8MB Corsair Force LS WesternDigital Blue 500GB 7200RPM 16MB ASUS DVD-RW 
CoolingOSOSOS
ThermalTake Frio Silent 14 Windows 10 Enterprise Linux Mint 17.3 Rosa OphCrack 
MonitorKeyboardPowerCase
ASUS VS228HR Logitech K120 Corsair VS650 ThermalTake View 27 
MouseAudio
Bloody V5 Corsair HS30 Raptor 
CPUMotherboardGraphicsRAM
Intel Mobile Core 2 Duo P8700 G72GX NVIDIA GeForce GTX 260M  Hyundai Electronics  
RAMRAM
Hyundai Electronics  Hyundai Electronics  
  hide details  
Reply
post #272 of 284
Quote:
Originally Posted by ku4eto View Post

So we can't even expect any performance increase, only power consumption ? Well, this doesn't sound hyping :/

If the new manufacturing process has completely different design targets (low power vs. high performance) and target devices (high performance CPUs vs. ARM SoCs & ASICs) , then absolutely not. The nominal size of the process doesn´t define how high Fmax it is able to reach, the other characteristics of the process define that.

The trend pretty much shows that the Fmax reached it´s peak on the 32nm SOI / P1268 HKMG, super / high performance processes.
Every node after that has produced at least some decline in the Fmax, despite more advanced manufacturing methods used. The effect has obviously been more grave for AMD than for Intel, since Intel moved to FinFet technology and never abandoned the high performance targeted manufacturing processes.
post #273 of 284
>I am not really angry, but the amount of $/performance for the Intel is .... disappointing.
Sure.

>The overall information on the library seems good, but i wasn't able to understand the results, can you explain them ?

Gladly, but maybe later tonight, right now I have to bail out, yet a quick shot:

Brotli_vs_lzma_vs_zopfli.png

The Google Switzerland team developing Brotli has done some good work, but as I told them not mentioning the current fastest decompressor on Internet (that is LzTurbo) is disrespecting, they did draw a line and discarded all compressors giving less compression ratio than the GZIP, so the superb Zstd was put outside the above picture as well, grmbl.

The "usual suspects" bzip2, gzip are chosen for the showdown, but they use compression algorithms which are not competitive, the current best is ANS developed by one super open guy from Poland, in its essence it kills the 60+ years old Huffman by simplifying it, by the way, my amateurish Nakamichi decompressor is doing the same with LZSS - just quantifying the match lengths to a fixed "blocks" moveable with a single register, more later...

>Also, a comparison would be good, even if its a server chip vs laptop one, we can still get some information out of it.
Sure sure, just wanted a more natural environment for the decompressors, for example Yann (the developer of Zstd) is so well versed in what has to be done that he enforced very tight resource utilization in Zstd (something like 512KB) and in my opinion is with the best view and potential to deliver (next year) something amazing - the killer of this outdated (GZIP) [de]compressor. It is not good that they disregarded his work.
post #274 of 284
Quote:
Originally Posted by geoxile View Post

Pretty sure Keller initially said they weren't doing everything from scratch to save time. Then again, since Skybridge (also his project) was cancelled and K12 has gone silent maybe they dropped those projects to give him more time to work on Zen.

Yes, initially they said he would be taking the best parts from what AMD had to build Zen. However, the most recent statement is that he started from scratch. As I said a few posts back, he isn't going to reinvent the wheel, but "from scratch" and using parts of what they had are two different things to me.

Although this is AMD management we are talking about here, they aren't known for being the most clear.
    
CPUMotherboardGraphicsRAM
Intel i7 5820K AsRock Extreme6 X99 Gigabyte GTX 980 Ti Windforce OC 16 GB Corsair Vengeance LPX 
Hard DriveHard DriveCoolingOS
Samsung 840 EVO 250GB - HDD Speed Edtition Samsung SM951 512 GB - I still hate Samsung!  Noctua NHD14 Windows 10 
MonitorMonitorMonitorKeyboard
Achieva Shimian QH270-Lite Overlord Computer Tempest X27OC  Acer Predator XB270HU Filco Majestouch 2 Ninja 
PowerCaseMouseMouse Pad
Seasonic X-1250 Fractal Design R5 Razer Naga Razer Goliathus Alpha 
AudioAudio
AKG K702 65th Anniversary Edition Creative Sound Blaster Zx 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
Intel i7 5820K AsRock Extreme6 X99 Gigabyte GTX 980 Ti Windforce OC 16 GB Corsair Vengeance LPX 
Hard DriveHard DriveCoolingOS
Samsung 840 EVO 250GB - HDD Speed Edtition Samsung SM951 512 GB - I still hate Samsung!  Noctua NHD14 Windows 10 
MonitorMonitorMonitorKeyboard
Achieva Shimian QH270-Lite Overlord Computer Tempest X27OC  Acer Predator XB270HU Filco Majestouch 2 Ninja 
PowerCaseMouseMouse Pad
Seasonic X-1250 Fractal Design R5 Razer Naga Razer Goliathus Alpha 
AudioAudio
AKG K702 65th Anniversary Edition Creative Sound Blaster Zx 
  hide details  
Reply
post #275 of 284
Quote:
Originally Posted by PostalTwinkie View Post

Yes, initially they said he would be taking the best parts from what AMD had to build Zen. However, the most recent statement is that he started from scratch. As I said a few posts back, he isn't going to reinvent the wheel, but "from scratch" and using parts of what they had are two different things to me.

Although this is AMD management we are talking about here, they aren't known for being the most clear.


"Clean Sheet" has been said several times including by Keller at the Core Innovations event on May 5, 2014. Keller quotes from that event: "New from scratch x86 core", "A lot of clean pieces of paper in that plan", "High level design as a clean piece paper", "We know how to build lots of components and we brought that in and filled in that plan fast"

It isnt really a complete clean sheet because it would be monumentally stupid and time consuming to ignore the bits and pieces that they know work well. That is probably what he means when he says that they used components or parts of what they had.
post #276 of 284
@ku4eto

Okay, here comes one quick shootout between Google's Brotli, GZIP, Hamid's LzTurbo, Yann's Zstd, Kaze's Shifune.

For those who still wonder what is that about, it is about fighting to promote some of the best candidates for transparent decompression (Brotli is all about to replace ZIP's Deflate) either in Internet http transfers or as a textual decoder.

So, the name of the game is to achieve high compression ratio and high decompression speed. Those two when balanced well are the answer for many I/O bottlenecks.

Brotli uses LZ77 with several tricks and supporting algorithms;
GZIP uses deflate algorithm;
LzTurbo uses ANS;
Zstd uses FSE;
Shifune uses simplified LZSS;
Igor Pavlov's LZMA is awesome but speedwise it fails behind in this task.

For a long time I witness author's discontention when [de]compressors are compared, too often they say unfair this unfair that, however I am immunized against that passion therefor I put Shifune right into the fire to see how it fares against PRO tools.
To avoid complains I used 4MB dictionary with LzTurbo, that's the default dictionary within Brotli.
Zstd uses very tight structures, AFAIK <1MB, so it is unfair to it, in near future to be boosted more.
GZIP uses 32KB - the walking dead.
Shifune uses 256MB window - the bad guy here.

Options used (compression):

FOR /F %%G IN (DIRLIST) DO bro_Intel15 -i %%G -o %%G.brotli
FOR /F %%G IN (DIRLIST) DO 7za a -tgzip -mx9 %%G.zip %%G
FOR /F %G IN (DIRLIST) DO lzturbo.exe -39 -b4 -p0 %G %G.4MB.lzturbo12-39.lzt
FOR /F %%G IN (DIRLIST) DO zstd.exe %%G
FOR /F %%G IN (DIRLIST) DO Nakamichi_Shifune_branchfull.exe %%G

So, the package is downloadable here: https://mega.nz/#!c0hBkCYQ!U2irEQPbiig6xcUD0mXIVn7P-suzTcku6COKtkvKHmo
It contains 6 testdatafiles:
Code:
D:\Showdown_Brotli_vs_Zstd_vs_GZIP_vs_Shifune>dir

09/26/2015  10:08 PM    13,713,275 Complete_Works_of_Fyodor_Dostoyevsky.txt
09/26/2015  10:35 PM     3,717,191 Complete_Works_of_Fyodor_Dostoyevsky.txt.4MB.lzturbo12-39.lzt
09/26/2015  10:16 PM     3,717,583 Complete_Works_of_Fyodor_Dostoyevsky.txt.brotli                         ! 153.228 MB/s !
09/08/2015  02:33 AM     4,582,363 Complete_Works_of_Fyodor_Dostoyevsky.txt.Nakamichi                      ! 448 MB/s !
09/26/2015  10:19 PM     4,617,360 Complete_Works_of_Fyodor_Dostoyevsky.txt.zip
09/26/2015  10:11 PM     5,209,670 Complete_Works_of_Fyodor_Dostoyevsky.txt.zst                            ! 302.5 MB/s !

09/26/2015  10:08 PM    10,192,446 dickens
09/26/2015  10:35 PM     2,976,910 dickens.4MB.lzturbo12-39.lzt
09/26/2015  10:17 PM     2,962,118 dickens.brotli                                                          ! 146.61 MB/s !
09/08/2015  02:33 AM     3,740,418 dickens.Nakamichi                                                       ! 448 MB/s !
09/26/2015  10:19 PM     3,681,828 dickens.zip
09/26/2015  10:11 PM     4,134,924 dickens.zst                                                             ! 298.9 MB/s !

09/26/2015  10:09 PM   100,000,000 enwik8
09/26/2015  10:37 PM    29,148,393 enwik8.4MB.lzturbo12-39.lzt
09/26/2015  10:24 PM    27,722,164 enwik8.brotli                                                           ! 134.094 MB/s !
09/08/2015  02:33 AM    34,218,460 enwik8.Nakamichi                                                        ! 256 MB/s !
09/26/2015  10:21 PM    35,102,891 enwik8.zip
09/26/2015  10:11 PM    40,024,854 enwik8.zst                                                              ! 325.0 MB/s !

09/26/2015  10:09 PM    14,613,183 The_Book_of_The_Thousand_Nights_and_a_Night.txt
09/26/2015  10:37 PM     4,241,855 The_Book_of_The_Thousand_Nights_and_a_Night.txt.4MB.lzturbo12-39.lzt
09/26/2015  10:19 PM     4,163,630 The_Book_of_The_Thousand_Nights_and_a_Night.txt.brotli                  ! 144.867 MB/s !
09/08/2015  02:33 AM     5,293,102 The_Book_of_The_Thousand_Nights_and_a_Night.txt.Nakamichi               ! 384 MB/s !
09/26/2015  10:22 PM     5,198,949 The_Book_of_The_Thousand_Nights_and_a_Night.txt.zip
09/26/2015  10:11 PM     5,932,453 The_Book_of_The_Thousand_Nights_and_a_Night.txt.zst                     ! 305.8 MB/s !

09/26/2015  10:09 PM     4,445,260 The_Project_Gutenberg_EBook_of_The_King_James_Bible_kjv10.txt
09/26/2015  10:37 PM     1,089,279 The_Project_Gutenberg_EBook_of_The_King_James_Bible_kjv10.txt.4MB.lzturbo12-39.lzt
09/26/2015  10:19 PM     1,087,439 The_Project_Gutenberg_EBook_of_The_King_James_Bible_kjv10.txt.brotli    ! 160.277 MB/s !
09/08/2015  02:33 AM     1,441,679 The_Project_Gutenberg_EBook_of_The_King_James_Bible_kjv10.txt.Nakamichi ! 704 MB/s !
09/26/2015  10:22 PM     1,320,100 The_Project_Gutenberg_EBook_of_The_King_James_Bible_kjv10.txt.zip
09/26/2015  10:11 PM     1,537,047 The_Project_Gutenberg_EBook_of_The_King_James_Bible_kjv10.txt.zst       ! 320.7 MB/s !

09/26/2015  10:09 PM     3,265,536 University_of_Canterbury_The_Calgary_Corpus.tar
09/26/2015  10:37 PM       921,958 University_of_Canterbury_The_Calgary_Corpus.tar.4MB.lzturbo12-39.lzt
09/26/2015  10:19 PM       867,503 University_of_Canterbury_The_Calgary_Corpus.tar.brotli                  ! 144.849 MB/s !
09/08/2015  02:33 AM     1,319,701 University_of_Canterbury_The_Calgary_Corpus.tar.Nakamichi               ! 576 MB/s !
09/26/2015  10:22 PM     1,017,658 University_of_Canterbury_The_Calgary_Corpus.tar.zip
09/26/2015  10:11 PM     1,174,349 University_of_Canterbury_The_Calgary_Corpus.tar.zst                     ! 367.3 MB/s !

D:\Showdown_Brotli_vs_Zstd_vs_GZIP_vs_Shifune>

To see the decompression speed by Brotli, Zstd and shifune just run RUNME.BAT and the outcome will be autoloaded into NOTEPAD.
My old Core 2 Q9550s @2.83GHz laptop 'Results.txt' has:
Code:
Nakamichi 'Shifune-Totenschiff', written by Kaze, based on Nobuo Ito's LZSS source, babealicious suggestion by m^2 enforced, muffinesque suggestion by Jim Dempsey enforced.
Note: This compile can handle files up to 1711MB.
Decompressing 4582363 bytes ...
RAM-to-RAM performance: 448 MB/s.
Compression Ratio (bigger-the-better): 2.99:1
Nakamichi 'Shifune-Totenschiff', written by Kaze, based on Nobuo Ito's LZSS source, babealicious suggestion by m^2 enforced, muffinesque suggestion by Jim Dempsey enforced.
Note: This compile can handle files up to 1711MB.
Decompressing 3740418 bytes ...
RAM-to-RAM performance: 448 MB/s.
Compression Ratio (bigger-the-better): 2.72:1
Nakamichi 'Shifune-Totenschiff', written by Kaze, based on Nobuo Ito's LZSS source, babealicious suggestion by m^2 enforced, muffinesque suggestion by Jim Dempsey enforced.
Note: This compile can handle files up to 1711MB.
Decompressing 34218460 bytes ...
RAM-to-RAM performance: 256 MB/s.
Compression Ratio (bigger-the-better): 2.92:1
Nakamichi 'Shifune-Totenschiff', written by Kaze, based on Nobuo Ito's LZSS source, babealicious suggestion by m^2 enforced, muffinesque suggestion by Jim Dempsey enforced.
Note: This compile can handle files up to 1711MB.
Decompressing 5293102 bytes ...
RAM-to-RAM performance: 384 MB/s.
Compression Ratio (bigger-the-better): 2.76:1
Nakamichi 'Shifune-Totenschiff', written by Kaze, based on Nobuo Ito's LZSS source, babealicious suggestion by m^2 enforced, muffinesque suggestion by Jim Dempsey enforced.
Note: This compile can handle files up to 1711MB.
Decompressing 1441679 bytes ...
RAM-to-RAM performance: 704 MB/s.
Compression Ratio (bigger-the-better): 3.08:1
Nakamichi 'Shifune-Totenschiff', written by Kaze, based on Nobuo Ito's LZSS source, babealicious suggestion by m^2 enforced, muffinesque suggestion by Jim Dempsey enforced.
Note: This compile can handle files up to 1711MB.
Decompressing 1319701 bytes ...
RAM-to-RAM performance: 576 MB/s.
Compression Ratio (bigger-the-better): 2.47:1

Loading Complete_Works_of_Fyodor_Dostoyevsky.txt...       

                                                                               
1-Complete_Works :  13713275 ->
1-Complete_Works :  13713275 ->   5127549 (37.39%),  104.5 MB/s
1-Complete_Works :  13713275 ->   5127549 (37.39%),  104.5 MB/s ,  302.5 MB/s
2-Complete_Works :  13713275 ->
2-Complete_Works :  13713275 ->   5127549 (37.39%),  104.5 MB/s
2-Complete_Works :  13713275 ->   5127549 (37.39%),  104.5 MB/s ,  302.5 MB/s
3-Complete_Works :  13713275 ->
3-Complete_Works :  13713275 ->   5127549 (37.39%),  104.5 MB/s
3-Complete_Works :  13713275 ->   5127549 (37.39%),  104.5 MB/s ,  302.5 MB/s
Complete_Works_o :  13713275 ->   5127549 (37.39%),  104.5 MB/s ,  302.5 MB/s

Loading dickens...       

                                                                               
1-dickens        :  10192446 ->
1-dickens        :  10192446 ->   4070941 (39.94%),   99.7 MB/s
1-dickens        :  10192446 ->   4070941 (39.94%),   99.7 MB/s ,  298.9 MB/s
2-dickens        :  10192446 ->
2-dickens        :  10192446 ->   4070941 (39.94%),   99.7 MB/s
2-dickens        :  10192446 ->   4070941 (39.94%),   99.7 MB/s ,  298.9 MB/s
3-dickens        :  10192446 ->
3-dickens        :  10192446 ->   4070941 (39.94%),   99.8 MB/s
3-dickens        :  10192446 ->   4070941 (39.94%),   99.8 MB/s ,  298.9 MB/s
dickens          :  10192446 ->   4070941 (39.94%),   99.8 MB/s ,  298.9 MB/s

Loading enwik8...       

                                                                               
1-enwik8         : 100000000 ->
1-enwik8         : 100000000 ->  39573323 (39.57%),  108.2 MB/s
1-enwik8         : 100000000 ->  39573323 (39.57%),  108.2 MB/s ,  324.8 MB/s
2-enwik8         : 100000000 ->
2-enwik8         : 100000000 ->  39573323 (39.57%),  108.2 MB/s
2-enwik8         : 100000000 ->  39573323 (39.57%),  108.2 MB/s ,  324.8 MB/s
3-enwik8         : 100000000 ->
3-enwik8         : 100000000 ->  39573323 (39.57%),  108.2 MB/s
3-enwik8         : 100000000 ->  39573323 (39.57%),  108.2 MB/s ,  325.0 MB/s
enwik8           : 100000000 ->  39573323 (39.57%),  108.2 MB/s ,  325.0 MB/s

Loading The_Book_of_The_Thousand_Nights_and_a_Night.txt...       

                                                                               
1-The_Book_of_Th :  14613183 ->
1-The_Book_of_Th :  14613183 ->   5855516 (40.07%),  102.2 MB/s
1-The_Book_of_Th :  14613183 ->   5855516 (40.07%),  102.2 MB/s ,  305.8 MB/s
2-The_Book_of_Th :  14613183 ->
2-The_Book_of_Th :  14613183 ->   5855516 (40.07%),  102.2 MB/s
2-The_Book_of_Th :  14613183 ->   5855516 (40.07%),  102.2 MB/s ,  305.8 MB/s
3-The_Book_of_Th :  14613183 ->
3-The_Book_of_Th :  14613183 ->   5855516 (40.07%),  102.2 MB/s
3-The_Book_of_Th :  14613183 ->   5855516 (40.07%),  102.2 MB/s ,  305.8 MB/s
The_Book_of_The_ :  14613183 ->   5855516 (40.07%),  102.2 MB/s ,  305.8 MB/s

Loading The_Project_Gutenberg_EBook_of_The_King_James_Bible_kjv10.txt...       

                                                                               
1-The_Project_Gu :   4445260 ->
1-The_Project_Gu :   4445260 ->   1512215 (34.02%),  115.2 MB/s
1-The_Project_Gu :   4445260 ->   1512215 (34.02%),  115.2 MB/s ,  308.0 MB/s
2-The_Project_Gu :   4445260 ->
2-The_Project_Gu :   4445260 ->   1512215 (34.02%),  115.2 MB/s
2-The_Project_Gu :   4445260 ->   1512215 (34.02%),  115.2 MB/s ,  320.2 MB/s
3-The_Project_Gu :   4445260 ->
3-The_Project_Gu :   4445260 ->   1512215 (34.02%),  115.2 MB/s
3-The_Project_Gu :   4445260 ->   1512215 (34.02%),  115.2 MB/s ,  320.7 MB/s
The_Project_Gute :   4445260 ->   1512215 (34.02%),  115.2 MB/s ,  320.7 MB/s

Loading University_of_Canterbury_The_Calgary_Corpus.tar...       

                                                                               
1-University_of_ :   3265536 ->
1-University_of_ :   3265536 ->   1164397 (35.66%),  124.8 MB/s
1-University_of_ :   3265536 ->   1164397 (35.66%),  124.8 MB/s ,  367.3 MB/s
2-University_of_ :   3265536 ->
2-University_of_ :   3265536 ->   1164397 (35.66%),  124.8 MB/s
2-University_of_ :   3265536 ->   1164397 (35.66%),  124.8 MB/s ,  367.3 MB/s
3-University_of_ :   3265536 ->
3-University_of_ :   3265536 ->   1164397 (35.66%),  124.8 MB/s
3-University_of_ :   3265536 ->   1164397 (35.66%),  124.8 MB/s ,  367.3 MB/s
University_of_Ca :   3265536 ->   1164397 (35.66%),  124.8 MB/s ,  367.3 MB/s


Brotli decompression speed: 153.228 MB/s
Brotli decompression speed: 146.61 MB/s
Brotli decompression speed: 134.094 MB/s
Brotli decompression speed: 144.867 MB/s
Brotli decompression speed: 160.277 MB/s
Brotli decompression speed: 144.849 MB/s

I very much like to see AMD Carrizo decompressing those files, as Yann smartly writes Zstd to use very scarce resources it opens unseen highlanes especially when Zstd is multi-threaded on Atom-like or ARM cpus.

If I had to decide what is most PRO I would instantly go with LzTurbo and Zstd - in my view they just warm up to kick asses.

Just imagine this scenario, the most printed and referenced book/text is that very King James Bible, let the user wants to load the whole Bible into the browser, instead of drawing 4,445,260 bytes when using
Code:
09/26/2015  10:09 PM         4,445,260 The_Project_Gutenberg_EBook_of_The_King_James_Bible_kjv10.txt
09/26/2015  10:37 PM         1,089,279 The_Project_Gutenberg_EBook_of_The_King_James_Bible_kjv10.txt.4MB.lzturbo12-39.lzt
09/26/2015  10:19 PM         1,087,439 The_Project_Gutenberg_EBook_of_The_King_James_Bible_kjv10.txt.brotli
09/08/2015  02:33 AM         1,441,679 The_Project_Gutenberg_EBook_of_The_King_James_Bible_kjv10.txt.Nakamichi
09/26/2015  10:22 PM         1,320,100 The_Project_Gutenberg_EBook_of_The_King_James_Bible_kjv10.txt.zip
09/26/2015  10:11 PM         1,537,047 The_Project_Gutenberg_EBook_of_The_King_James_Bible_kjv10.txt.zst

some of above decompressors, we will have to download only 1MB-1.4MB, if the connection is 10Mbps or roughly 1MB/s then Shifune saves 3 seconds, since the file is "small" decompression speed is lost in the ... translation.

Note that, by choosing this ~4MB file, I didn't hurt feelings of Brotli authors - their current dictionary is 4MB.
post #277 of 284
4 core no HT (6600K)

Brotli decompression speed: 397.508 MB/s
Brotli decompression speed: 374.577 MB/s
Brotli decompression speed: 305.225 MB/s
Brotli decompression speed: 372.129 MB/s
Brotli decompression speed: 370.247 MB/s
Brotli decompression speed: 283.114 MB/s
Results.txt 7k .txt file


will do a 5960X run, but I believe the program will not use 16 threads.
x99
(22 items)
 
Z370
(8 items)
 
x299
(18 items)
 
CPUCPUMotherboardGraphics
i7 5960X i7 6950X Asus Rampage V Extreme 10 2 GTX Titan X Pascal 
RAMHard DriveHard DriveHard Drive
GSkill 3200c14 TZs 8x8GB @ 3400c13 Intel 750 NVMe 400GB 2x Plextor SSD 256 Raid 0 (Win 7) Samsung NVMe 950 Pro M.2 
Optical DriveCoolingCoolingCooling
Plextor 810 2x XSPC RX360 Koolance 380i D5 
CoolingOSOSOS
Aquaero 6 Windows 10 x64  Windows 7 Pro Linux Mint 
MonitorKeyboardPowerCase
Seiki 50" 4K monitor (720P-2160i) 240-30Hz Das Keyboard Model S Pro Corsair AX1200 Case Labs SM8 
MouseAudio
Steelseries Rival Ultimate Ears TF10 
CPUMotherboardGraphicsRAM
8700K ASUS Maximus X Apex GTX 1080 2x8 GB G.Skill 4400c19 
Hard DriveCoolingOSCase
Samsung 960 Pro 360Rad + Chiller Win 10 HWBOT OBT 
CPUCPUMotherboardGraphics
7740X 7980XE ASUS Rampage Vi Apex 2x Nvidia Titan Xp SLi 
RAMHard DriveHard DriveHard Drive
G.Skill 3600c15 2x8GB kits Intel 900p WD Raptor 1T Plextor SSDs Raid 0 
Optical DriveCoolingOSOS
Plextor DVD/BR R/W (very) Custom Water Windows 10 Pro Windows 7 Pro 
MonitorKeyboardPowerCase
ASUS 144Hz Ducky Corsair 1500i Microcool Benchetto 101 
MouseAudio
Steelseries Rival Ultimate ears TF10s (IEMs) 
  hide details  
Reply
x99
(22 items)
 
Z370
(8 items)
 
x299
(18 items)
 
CPUCPUMotherboardGraphics
i7 5960X i7 6950X Asus Rampage V Extreme 10 2 GTX Titan X Pascal 
RAMHard DriveHard DriveHard Drive
GSkill 3200c14 TZs 8x8GB @ 3400c13 Intel 750 NVMe 400GB 2x Plextor SSD 256 Raid 0 (Win 7) Samsung NVMe 950 Pro M.2 
Optical DriveCoolingCoolingCooling
Plextor 810 2x XSPC RX360 Koolance 380i D5 
CoolingOSOSOS
Aquaero 6 Windows 10 x64  Windows 7 Pro Linux Mint 
MonitorKeyboardPowerCase
Seiki 50" 4K monitor (720P-2160i) 240-30Hz Das Keyboard Model S Pro Corsair AX1200 Case Labs SM8 
MouseAudio
Steelseries Rival Ultimate Ears TF10 
CPUMotherboardGraphicsRAM
8700K ASUS Maximus X Apex GTX 1080 2x8 GB G.Skill 4400c19 
Hard DriveCoolingOSCase
Samsung 960 Pro 360Rad + Chiller Win 10 HWBOT OBT 
CPUCPUMotherboardGraphics
7740X 7980XE ASUS Rampage Vi Apex 2x Nvidia Titan Xp SLi 
RAMHard DriveHard DriveHard Drive
G.Skill 3600c15 2x8GB kits Intel 900p WD Raptor 1T Plextor SSDs Raid 0 
Optical DriveCoolingOSOS
Plextor DVD/BR R/W (very) Custom Water Windows 10 Pro Windows 7 Pro 
MonitorKeyboardPowerCase
ASUS 144Hz Ducky Corsair 1500i Microcool Benchetto 101 
MouseAudio
Steelseries Rival Ultimate ears TF10s (IEMs) 
  hide details  
Reply
post #278 of 284
4960X results

Brotli decompression speed: 346.897 MB/s
Brotli decompression speed: 313.557 MB/s
Brotli decompression speed: 266.911 MB/s
Brotli decompression speed: 321.111 MB/s
Brotli decompression speed: 304.988 MB/s
Brotli decompression speed: 245.217 MB/s
Results.txt 7k .txt file
x99
(22 items)
 
Z370
(8 items)
 
x299
(18 items)
 
CPUCPUMotherboardGraphics
i7 5960X i7 6950X Asus Rampage V Extreme 10 2 GTX Titan X Pascal 
RAMHard DriveHard DriveHard Drive
GSkill 3200c14 TZs 8x8GB @ 3400c13 Intel 750 NVMe 400GB 2x Plextor SSD 256 Raid 0 (Win 7) Samsung NVMe 950 Pro M.2 
Optical DriveCoolingCoolingCooling
Plextor 810 2x XSPC RX360 Koolance 380i D5 
CoolingOSOSOS
Aquaero 6 Windows 10 x64  Windows 7 Pro Linux Mint 
MonitorKeyboardPowerCase
Seiki 50" 4K monitor (720P-2160i) 240-30Hz Das Keyboard Model S Pro Corsair AX1200 Case Labs SM8 
MouseAudio
Steelseries Rival Ultimate Ears TF10 
CPUMotherboardGraphicsRAM
8700K ASUS Maximus X Apex GTX 1080 2x8 GB G.Skill 4400c19 
Hard DriveCoolingOSCase
Samsung 960 Pro 360Rad + Chiller Win 10 HWBOT OBT 
CPUCPUMotherboardGraphics
7740X 7980XE ASUS Rampage Vi Apex 2x Nvidia Titan Xp SLi 
RAMHard DriveHard DriveHard Drive
G.Skill 3600c15 2x8GB kits Intel 900p WD Raptor 1T Plextor SSDs Raid 0 
Optical DriveCoolingOSOS
Plextor DVD/BR R/W (very) Custom Water Windows 10 Pro Windows 7 Pro 
MonitorKeyboardPowerCase
ASUS 144Hz Ducky Corsair 1500i Microcool Benchetto 101 
MouseAudio
Steelseries Rival Ultimate ears TF10s (IEMs) 
  hide details  
Reply
x99
(22 items)
 
Z370
(8 items)
 
x299
(18 items)
 
CPUCPUMotherboardGraphics
i7 5960X i7 6950X Asus Rampage V Extreme 10 2 GTX Titan X Pascal 
RAMHard DriveHard DriveHard Drive
GSkill 3200c14 TZs 8x8GB @ 3400c13 Intel 750 NVMe 400GB 2x Plextor SSD 256 Raid 0 (Win 7) Samsung NVMe 950 Pro M.2 
Optical DriveCoolingCoolingCooling
Plextor 810 2x XSPC RX360 Koolance 380i D5 
CoolingOSOSOS
Aquaero 6 Windows 10 x64  Windows 7 Pro Linux Mint 
MonitorKeyboardPowerCase
Seiki 50" 4K monitor (720P-2160i) 240-30Hz Das Keyboard Model S Pro Corsair AX1200 Case Labs SM8 
MouseAudio
Steelseries Rival Ultimate Ears TF10 
CPUMotherboardGraphicsRAM
8700K ASUS Maximus X Apex GTX 1080 2x8 GB G.Skill 4400c19 
Hard DriveCoolingOSCase
Samsung 960 Pro 360Rad + Chiller Win 10 HWBOT OBT 
CPUCPUMotherboardGraphics
7740X 7980XE ASUS Rampage Vi Apex 2x Nvidia Titan Xp SLi 
RAMHard DriveHard DriveHard Drive
G.Skill 3600c15 2x8GB kits Intel 900p WD Raptor 1T Plextor SSDs Raid 0 
Optical DriveCoolingOSOS
Plextor DVD/BR R/W (very) Custom Water Windows 10 Pro Windows 7 Pro 
MonitorKeyboardPowerCase
ASUS 144Hz Ducky Corsair 1500i Microcool Benchetto 101 
MouseAudio
Steelseries Rival Ultimate ears TF10s (IEMs) 
  hide details  
Reply
post #279 of 284
5960X results (definitely NOT using the 16 threads. Barely hit 45% on one core)

Brotli decompression speed: 364.29 MB/s
Brotli decompression speed: 336.342 MB/s
Brotli decompression speed: 285.19 MB/s
Brotli decompression speed: 336.624 MB/s
Brotli decompression speed: 339.146 MB/s
Brotli decompression speed: 266.176 MB/s

Results.txt 7k .txt file


there ya go @Sanmayce
thumb.gif
(I also have a 2700K, Q9650 and Q9300 running here)
x99
(22 items)
 
Z370
(8 items)
 
x299
(18 items)
 
CPUCPUMotherboardGraphics
i7 5960X i7 6950X Asus Rampage V Extreme 10 2 GTX Titan X Pascal 
RAMHard DriveHard DriveHard Drive
GSkill 3200c14 TZs 8x8GB @ 3400c13 Intel 750 NVMe 400GB 2x Plextor SSD 256 Raid 0 (Win 7) Samsung NVMe 950 Pro M.2 
Optical DriveCoolingCoolingCooling
Plextor 810 2x XSPC RX360 Koolance 380i D5 
CoolingOSOSOS
Aquaero 6 Windows 10 x64  Windows 7 Pro Linux Mint 
MonitorKeyboardPowerCase
Seiki 50" 4K monitor (720P-2160i) 240-30Hz Das Keyboard Model S Pro Corsair AX1200 Case Labs SM8 
MouseAudio
Steelseries Rival Ultimate Ears TF10 
CPUMotherboardGraphicsRAM
8700K ASUS Maximus X Apex GTX 1080 2x8 GB G.Skill 4400c19 
Hard DriveCoolingOSCase
Samsung 960 Pro 360Rad + Chiller Win 10 HWBOT OBT 
CPUCPUMotherboardGraphics
7740X 7980XE ASUS Rampage Vi Apex 2x Nvidia Titan Xp SLi 
RAMHard DriveHard DriveHard Drive
G.Skill 3600c15 2x8GB kits Intel 900p WD Raptor 1T Plextor SSDs Raid 0 
Optical DriveCoolingOSOS
Plextor DVD/BR R/W (very) Custom Water Windows 10 Pro Windows 7 Pro 
MonitorKeyboardPowerCase
ASUS 144Hz Ducky Corsair 1500i Microcool Benchetto 101 
MouseAudio
Steelseries Rival Ultimate ears TF10s (IEMs) 
  hide details  
Reply
x99
(22 items)
 
Z370
(8 items)
 
x299
(18 items)
 
CPUCPUMotherboardGraphics
i7 5960X i7 6950X Asus Rampage V Extreme 10 2 GTX Titan X Pascal 
RAMHard DriveHard DriveHard Drive
GSkill 3200c14 TZs 8x8GB @ 3400c13 Intel 750 NVMe 400GB 2x Plextor SSD 256 Raid 0 (Win 7) Samsung NVMe 950 Pro M.2 
Optical DriveCoolingCoolingCooling
Plextor 810 2x XSPC RX360 Koolance 380i D5 
CoolingOSOSOS
Aquaero 6 Windows 10 x64  Windows 7 Pro Linux Mint 
MonitorKeyboardPowerCase
Seiki 50" 4K monitor (720P-2160i) 240-30Hz Das Keyboard Model S Pro Corsair AX1200 Case Labs SM8 
MouseAudio
Steelseries Rival Ultimate Ears TF10 
CPUMotherboardGraphicsRAM
8700K ASUS Maximus X Apex GTX 1080 2x8 GB G.Skill 4400c19 
Hard DriveCoolingOSCase
Samsung 960 Pro 360Rad + Chiller Win 10 HWBOT OBT 
CPUCPUMotherboardGraphics
7740X 7980XE ASUS Rampage Vi Apex 2x Nvidia Titan Xp SLi 
RAMHard DriveHard DriveHard Drive
G.Skill 3600c15 2x8GB kits Intel 900p WD Raptor 1T Plextor SSDs Raid 0 
Optical DriveCoolingOSOS
Plextor DVD/BR R/W (very) Custom Water Windows 10 Pro Windows 7 Pro 
MonitorKeyboardPowerCase
ASUS 144Hz Ducky Corsair 1500i Microcool Benchetto 101 
MouseAudio
Steelseries Rival Ultimate ears TF10s (IEMs) 
  hide details  
Reply
post #280 of 284
Quote:
Originally Posted by kpzero View Post

"Clean Sheet" has been said several times including by Keller at the Core Innovations event on May 5, 2014. Keller quotes from that event: "New from scratch x86 core", "A lot of clean pieces of paper in that plan", "High level design as a clean piece paper", "We know how to build lots of components and we brought that in and filled in that plan fast"

It isnt really a complete clean sheet because it would be monumentally stupid and time consuming to ignore the bits and pieces that they know work well. That is probably what he means when he says that they used components or parts of what they had.

Yea, those are the comments we are wanting to see. Glad you brought them out. I have enough faith left in me that they can pull something great off. Although the waiting is going to be kind of crazy, to bad it isn't now!
    
CPUMotherboardGraphicsRAM
Intel i7 5820K AsRock Extreme6 X99 Gigabyte GTX 980 Ti Windforce OC 16 GB Corsair Vengeance LPX 
Hard DriveHard DriveCoolingOS
Samsung 840 EVO 250GB - HDD Speed Edtition Samsung SM951 512 GB - I still hate Samsung!  Noctua NHD14 Windows 10 
MonitorMonitorMonitorKeyboard
Achieva Shimian QH270-Lite Overlord Computer Tempest X27OC  Acer Predator XB270HU Filco Majestouch 2 Ninja 
PowerCaseMouseMouse Pad
Seasonic X-1250 Fractal Design R5 Razer Naga Razer Goliathus Alpha 
AudioAudio
AKG K702 65th Anniversary Edition Creative Sound Blaster Zx 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
Intel i7 5820K AsRock Extreme6 X99 Gigabyte GTX 980 Ti Windforce OC 16 GB Corsair Vengeance LPX 
Hard DriveHard DriveCoolingOS
Samsung 840 EVO 250GB - HDD Speed Edtition Samsung SM951 512 GB - I still hate Samsung!  Noctua NHD14 Windows 10 
MonitorMonitorMonitorKeyboard
Achieva Shimian QH270-Lite Overlord Computer Tempest X27OC  Acer Predator XB270HU Filco Majestouch 2 Ninja 
PowerCaseMouseMouse Pad
Seasonic X-1250 Fractal Design R5 Razer Naga Razer Goliathus Alpha 
AudioAudio
AKG K702 65th Anniversary Edition Creative Sound Blaster Zx 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Rumors and Unconfirmed Articles
Overclock.net › Forums › Industry News › Rumors and Unconfirmed Articles › [HeXus]Legendary CPU architect Jim Keller leaves AMD