Phenom II 6 GHz+ OVERCLOCK

[justapost]

http://www.gamestar.de/hardware/test...nom_ii_x4.html

Unreal Tournament 3 shows the potential of the new Phenom II-series: Running at identical clock speeds of 2.7 GHz the simulated Phenom II X4 clearly beats the (overclocked) Phenom X4 9950 BE: 175.1 to 129.5 FPS in 1280x1024 and 167.6 to 128.2 FPS in 1680x1050 using high detail settings. But the Core i7 920 with its slightly lower clock speed of 2.66 GHz is still ahead of the Phenom II with 202.4 to 175.1 FPS (1280x1024) and 188,1 to 167.6 FPS (1680x1050).

In Call of Duty 4 the new Phenom II X4 is slower than its predecessor for no apparent reason. Even though we double-checked the BIOS and the game settings, the Phenom II X4 only managed 117.3 FPS (1280x1024) compared to the Phenom X4 9950 BE with 126.8 FPS, both using the same settings. This gap can also be seen in 1680x1050 with 106.8 (Phenom II X4) to 119.2 FPS (Phenom X4 9950 BE).
The internal benchmark of the compression tool WinRAR confirms the performance advantage of the Phenom X4 compared to its predecessor seen in Unreal Tournament 3 – with 1,528 KB/s it compresses data noticeably faster than the Phenom X4 9950 BE with 1,318 KB/s, both running with identical clock speeds. In optimized multimedia applications like WinRAR, Intel managed to almost double the performance going from Core 2 to Core i7 (see: Core i7 put to the test (in German)). AMD doesn't achieve anything similar with the Phenom II X4 – but that's no surprise given that there are hardly any technical differences between both generations of Phenom processors.

[Ghostbuster]

http://www.gamestar.de/hardware/test...nom_ii_x4.html

Looking at that chart according my calculations... Using Q6600 as the base, as Q9650 as the top.

That Opteron 2384 ("Phenom II X4") 2.7GHz is roughly equivalent to a 2.7GHz Core 2 Quad, clock-to-clock wise in UT3.

[demonkevy666]

http://www.gamestar.de/hardware/test...nom_ii_x4.html

why DX9 ? that is so lame to do.

they where also getting scores like that on i7 because of immature bios. where the older core 2 qaud where out doing them.

lets not for get that the operton is bottleneck with suck a low HT and L3 cache lock. I know most of you proably think thats not goign to matter but test show there can be 1-5% with change in both of those.

[informal]

What NB clocks did they run on this Opteron?Also the ram in question is ECC DDR2-800,compared to 1066 on Agena and C2Q.

I've just seen the update in the German version of the article:
http://www.gamestar.de/hardware/test...nom_ii_x4.html

Update: Response to some reader questions regarding the Opteron test system:
The Geforce GTX 280 was at full PCI Express 16x-speed connections (it is in the bios on the motherboard to configure server). As DDR2 memory came-800-RAM (CL6) with error correction (ECC, mode: Basic) to use. Since the Phenom X4 II (unlike the Opteron system) to show even faster RAM modules supports, should the gaming performance yet at the same clock by a few percentage points increase.

So the ram was CL6 ECC DDR2-800.One can assume that CL4/5 DDR2-1066 ram memory would give an extra boost over the one they used.Also we don't know what NB clocks will Deneb run at.It may be the same 2.2Ghz as Shanghai parts,or it may be higher than that.All of this coupled with a few BIOS settings we don't know about(CnQ active?<CPU TWEAK ACTIVE?-those are non existent in server bios i presume) can't give us a valid picture how exactly Deneb will perform.
But even in this short and not exactly accurate simulation,Deneb shows some massive gains in UT3 and Winrar(which is VERY sensitive to memory setting on K8/K10).

[demonkevy666]

What NB clocks did they run on this Opteron?Also the ram in question is ECC DDR2-800,compared to 1066 on Agena and C2Q.

I've just seen the update in the German version of the article:
http://www.gamestar.de/hardware/test...nom_ii_x4.html



So the ram was CL6 ECC DDR2-800.One can assume that CL4/5 DDR2-1066 ram memory would give an extra boost over the one they used.Also we don't know what NB clocks will Deneb run at.It may be the same 2.2Ghz as Shanghai parts,or it may be higher than that.All of this coupled with a few BIOS settings we don't know about(CnQ active?<CPU TWEAK ACTIVE?-those are non existent in server bios i presume) can't give us a valid picture how exactly Deneb will perform.
But even in this short and not exactly accurate simulation,Deneb shows some massive gains in UT3 and Winrar(which is VERY sensitive to memory setting on K8/K10).

also that sever board is running on Ht 1.0.

No sever until 09 will have Ht3.0/or Hht3.1

[gallag]

What NB clocks did they run on this Opteron?Also the ram in question is ECC DDR2-800,compared to 1066 on Agena and C2Q.

I've just seen the update in the German version of the article:
http://www.gamestar.de/hardware/test...nom_ii_x4.html



So the ram was CL6 ECC DDR2-800.One can assume that CL4/5 DDR2-1066 ram memory would give an extra boost over the one they used.Also we don't know what NB clocks will Deneb run at.It may be the same 2.2Ghz as Shanghai parts,or it may be higher than that.All of this coupled with a few BIOS settings we don't know about(CnQ active?<CPU TWEAK ACTIVE?-those are non existent in server bios i presume) can't give us a valid picture how exactly Deneb will perform.
But even in this short and not exactly accurate simulation,Deneb shows some massive gains in UT3 and Winrar(which is VERY sensitive to memory setting on K8/K10).

Not really a massive gain in winrar, Is it? I would expect deneb to be better than this if not hindered by server parts.

[informal]

Not really a massive gain in winrar, Is it? I would expect deneb to be better than this if not hindered by server parts.

Well i told you what matters for Winrar:it's extremely sensitive to memory subsystem(if you don't believe me,check with someone else).
I can jump ten percent with my X2 only by going from cas 5 to cas4.Obviously larger L3 pays a role here,but even on a "crippled" (server) platform,that opteron shows 15% gain clock for clock over Agena which ran with uber fast ddr2 ram(non ECC too).
I can bet when AM3 version of Deneb comes out it will eat the AM2+ version of Deneb in Winrar.That's why you see Core i7 perform stellar in this test:3 channel DDR3 IMC + Turbo + HT +large L3 = massive gain.Deneb will lack turbo and HT but will have fast DDR2/DDR3 memory on it's disposal and a large chunk of L3(with lower latency than the one in Barcelona\Agena).

15% on this "desktop unfriendly" platform for a "die shrink" is a great score IMO.

[PetNorth]

Well i told you what matters for Winrar:it's extremely sensitive to memory subsystem(if you don't believe me,check with someone else).
I can jump ten percent with my X2 only by going from cas 5 to cas4.Obviously larger L3 pays a role here,but even on a "crippled" (server) platform,that opteron shows 15% gain clock for clock over Agena which ran with uber fast ddr2 ram(non ECC too).
I can bet when AM3 version of Deneb comes out it will eat the AM2+ version of Deneb in Winrar.That's why you see Core i7 perform stellar in this test:3 channel DDR3 IMC + Turbo + HT +large L3 = massive gain.Deneb will lack turbo and HT but will have fast DDR2/DDR3 memory on it's disposal and a large chunk of L3(with lower latency than the one in Barcelona\Agena).

15% on this "desktop unfriendly" platform for a "die shrink" is a great score IMO.

Expect at least another 20% gain (to add to that 15%), going from ECC 800 CL6 to NO ECC 1066 CL5. I've done a quick run with winrar 3.80 (Phenom 9850). Five runs (average): 1226 (667 CL5); 1410 (800 CL4). Thats 15% gain.

[informal]

Expect at least another 20% gain (to add to that 15%), going from ECC 800 CL6 to NO ECC 1066 CL5. I've done a quick run with winrar 3.80 (Phenom 9850). Five runs (average): 1226 (667 CL5); 1410 (800 CL4). Thats 15% gain.

Thanks for the numbers .
As i said,Winrar is very sensitive to memory subsystem on K10.
Deneb + fast(CL4) 1066 DDR2 ram+high(er) NB clock+much larger L3 = big gain in Winrar.