Overclock.net › Forums › Industry News › Hardware News › [AMD] Breakthrough Performance of “Zen” (Head to Head with Broadwell-E!)
New Posts  All Forums:Forum Nav:

[AMD] Breakthrough Performance of “Zen” (Head to Head with Broadwell-E!) - Page 25

post #241 of 580
Quote:
Originally Posted by Malinkadink View Post

Quote:
Originally Posted by Yttrium View Post

You're so naive, the 6900k is going for 1000$ and you're expecting amd to start a charity to
gain marketshare?! quite the opposite. Zen is going to be priced competitive; if performance is around intel's offering the the price is going to be around intel's offering. THATS what gains marketshare.

-an "AMD fanboy"

Have fun paying $900 for a competitive product

-an "Intel fanboy"
Quote:
Originally Posted by czin125 View Post


It's going to be 800-900 dollars for 8 cores vs the 1089 for 8 cores.

550 for 6 cores full pci-e lanes vs 6850K's 600+

As painfull as it is I must confess that 800-900 is too optimistic despite the slightly sarcastic comment of malink. The 6900k is priced around 1100-1200$~ in actual shops therefore I expect the consumer to get a performance equivalent ZEN for just 999.99 dollar$
post #242 of 580
Quote:
Originally Posted by formula m View Post

Quote:
Originally Posted by TheLAWNOOB View Post

My bad. Zen is going to match Broadwell-E in everything because it matches it in blender.


Single thread performance doesn't matter because it's 2016, and 8 core Zen is going to cost $250 because AMD cares about consumers' wallets.


Am I doing this right?


Edit: How is that review irrelevant? AMD claimed Zen to be 40% faster than Excavator, and that is an Excavator review.


Wouldn't you know, but Blender also uses single thread performance. And it is nearly equal to broadwell-e clock for clock. So, what do all intel chips normalized to the same clock, end up looking like? Which one has the most ipc?


Subsequently, you are reading the slides & presentation wrong. AMD has not released Zen's performance figures. Though, they did say that Zen's IPC is 40% more than Excavator, a was mentioned as one of the goals for AMD's team. Not the only goal, just one of them. 

ergo: Zen is not 40% faster than excavator, is has 40% more ipc than Excavator. Zen's actual architecture might yield other improvements and end up being 55% more efficient. We don't know, only that we know it can compute more per cycle.


Also, Excavator is not Zen.

I'm estimating the performance of Zen by looking at Excavator benchmarks.

I never claimed Zen was Excavator.

Did I?
Power Bac G5
(6 items)
 
  
CPUGraphicsRAMHard Drive
Intel i7 6950X EVGA GTX 1080 Ti GSkill 32GB 4133Mhz Samsung 960 Pro 1TB 
CoolingMonitor
Corsair H110i ASUS PG278Q 
  hide details  
Reply
Power Bac G5
(6 items)
 
  
CPUGraphicsRAMHard Drive
Intel i7 6950X EVGA GTX 1080 Ti GSkill 32GB 4133Mhz Samsung 960 Pro 1TB 
CoolingMonitor
Corsair H110i ASUS PG278Q 
  hide details  
Reply
post #243 of 580
Quote:
Originally Posted by Themisseble View Post

Quote:
Originally Posted by TheLAWNOOB View Post

My bad. 1 Zen core at 1 Ghz beats intel 22 core at 2.5Ghz.

GG intel.

Its very interesting. You are acting like you dont want competition. Well, of course we can expect huge amount of HYPE and yes ZEN beat Broadwell E in 1 benchmark. Good think is that the benchmark is FPU intensive.

Important thing is...? AMD is BACK! Well we need to wait to official benchmarks, but its looking very good so far. All I want is cheaper CPUs, I dont mind which one...

If you hate AMD ... no need to make troll post here.

ZEN is wider than HASWELL and I hope it will be faster than Haswell in 80% of benchmarks.

Yeah I hate competition.

Can't wait for $1000 dual core intel.

Zen is wider than Haswell? I hope it's dense too.
Power Bac G5
(6 items)
 
  
CPUGraphicsRAMHard Drive
Intel i7 6950X EVGA GTX 1080 Ti GSkill 32GB 4133Mhz Samsung 960 Pro 1TB 
CoolingMonitor
Corsair H110i ASUS PG278Q 
  hide details  
Reply
Power Bac G5
(6 items)
 
  
CPUGraphicsRAMHard Drive
Intel i7 6950X EVGA GTX 1080 Ti GSkill 32GB 4133Mhz Samsung 960 Pro 1TB 
CoolingMonitor
Corsair H110i ASUS PG278Q 
  hide details  
Reply
post #244 of 580
If we're speculating i think we'll have 2 or 3 octa-core SKU at launch
Anything at 10-15% near 6900k or closer on average and forget about aggressive pricing

8C/16T unlocked, 140W TDP
3.2ghz base clock
3.4ghz full turbo
3.8ghz 1core turbo
Price: 800E++++

8C/16T unlocked, 95W TDP
2.8ghz base
3.2ghz full turbo
3.6ghz 1core turbo
Price: 500E+

8C/16T locked, 95-120W TDP
3Ghz base
3.2 full turbo
3.8 1core turbo
Price: between the 2 above

Expect 6cores at 300-500 and 4cores at 150-300.
If zen is a failure (according to the expectations that amd set, with the blender benchmark), like 20% slower than 6900k or more on average just divide all prices by 2
Edited by Marios145 - 8/19/16 at 2:47pm
post #245 of 580
I don't think AMD wants 140 W TDP, or locking theirs chips. That would hurt AMD against Intel quite a bit.
post #246 of 580
Quote:
Originally Posted by formula m View Post

Wouldn't you know, but Blender also uses single thread performance.

This was certainly a multi-threaded test. Different architectures will scale differently under different workloads.

Blender is as parallel and dependency free as it gets, which will cover up any performance issues from having Zen's eight cores and 16MiB L3 split into two quad-core modules.

Blender is also a low efficiency/low IPC app, meaning that it responds unusually well to SMT, which again, will cover up some potential Zen execution issues, as well as take good advantage of Zen's larger per-core cache.
Quote:
Originally Posted by formula m View Post

And it is nearly equal to broadwell-e clock for clock.

And there are many things about the Zen architecture and the Blender results that give the impression that Blender is likely an outlier, not something that is representative of average performance per clock. Everything released so far about Zen architecture makes it look like it would be ideal for Blender and the fact that Blender was the first test chosen to demonstrate how competitive Zen could be is testament to this.

Even if Zen is absolutely awesome (and I'm quite optimistic), people are likely setting themselves up for disappointment if they think Blender is representative of more than perfectly parallel, low IPC, low-to-moderate I/O, tasks.
Quote:
Originally Posted by formula m View Post

So, what do all intel chips normalized to the same clock, end up looking like? Which one has the most ipc?

Overall, newer architectures equal more IPC. Specific IPC is highly task dependent, of course.

In JIT compilation heavy tasks (emulation mainly), my i7 5820k (Haswell-E) is a solid 60% faster, clock for clock, than my i7 970 or Xeon X5670 (Westmere parts). This is because these tasks are extremely dependent on branch prediction, buffer/queue sizes, and L1/L2 cache performance, which were dramatically increased with Haswell.

In something like 7-zip, which is limited primarily by CPU integer performance and secondarily by memory latency (neither of which have significantly improved), IPC is almost the same between these architectures.
Quote:
Originally Posted by formula m View Post

ergo: Zen is not 40% faster than excavator, is has 40% more ipc than Excavator. Zen's actual architecture might yield other improvements and end up being 55% more efficient. We don't know, only that we know it can compute more per cycle.

Overall IPC is gauged by measuring performance and comparing it to clock speeds. If Zen averages 40% higher IPC than excavator it should perform roughly 40% better at the same clock speed than Excavator, on average.

Zen being 40% faster than Excavator will be contingent on Zen reaching Excavator clocks.
Quote:
Originally Posted by TheLAWNOOB View Post

Zen is wider than Haswell?

It is, though it's execution units also look less flexible.

This implies that while it's wideness may not always be easily leveraged in the form of extra ILP (vs. it's competition...it will certainly be far better in this regard than it's predecessors) that it could well see larger average performance gains from SMT than Intel (especially given that Zen has twice the L1-I and L2 cache of Intel, and cache contention is another limiting factor in SMP).

The info released is quite incomplete and there are a lot of unknowns, but these are the distinct impressions I get from looking at the info on hand, especially the architectural block diagram, and comparing them to Haswell/Broadwell.
Edited by Blameless - 8/19/16 at 3:05pm
Primary
(15 items)
 
Secondary
(13 items)
 
In progress
(10 items)
 
CPUMotherboardGraphicsRAM
5820K @ 4.2/3.5GHz core/uncore, 1.175/1.15v Gigabyte X99 SOC Champion (F22n) Gigabyte AORUS GTX 1080 Ti (F3P) @ 2025/1485, 1... 4x4GiB Crucial @ 2667, 12-12-12-28-T1, 1.34v 
Hard DriveHard DriveHard DriveCooling
Plextor M6e 128GB (fw 1.06) M.2 (PCI-E 2.0 2x) 2x Crucial M4 256GB 4x WD Scorpio Black 500GB Noctua NH-D15 
OSMonitorKeyboardPower
Windows 7 Professional x64 SP1 BenQ BL3200PT Filco Majestouch Tenkeyless (MX Brown) Corsair RM1000x 
CaseMouseAudio
Fractal Design Define R4 Logitech G402 Realtek ALC1150 + M-Audio AV40 
CPUMotherboardGraphicsRAM
X5670 @ 4.4/3.2GHz core/uncore, 1.36 vcore, 1.2... Gigabyte X58A-UD5 r2.0 w/FF3mod10 BIOS Sapphire Fury Nitro OC+ @ 1053/500, 1.225vGPU/1... 2x Samsung MV-3V4G3D/US @ 2000, 10-11-11-30-T1,... 
RAMHard DriveHard DriveHard Drive
1x Crucial BLT4G3D1608ET3LX0 @ 2000, 10-11-11-3... OCZ (Toshiba) Trion 150 120GB Hyundai Sapphire 120GB 3x Hitachi Deskstar 7k1000.C 1TB 
CoolingOSPowerCase
Noctua NH-D14 Windows 7 Pro x64 SP1 Antec TP-750 Fractal Design R5 
Audio
ASUS Xonar DS 
CPUMotherboardGraphicsRAM
i7-6800K @ 4.3/3.5GHz core/uncore, 1.36/1.2v ASRock X99 OC Formula (P3.10) GTX 780 (temporary) 4x4GiB Crucial DDR4-2400 @ 11-13-12-28-T2, 1.33v 
Hard DriveHard DriveCoolingOS
Intel 600p 256GB NVMe 2x HGST Travelstar 7k1000 1TB Corsair H55 (temporary) Windows Server 2016 Datacenter 
PowerCase
Seasonic SS-860XP2 Corsair Carbide Air 540 
  hide details  
Reply
Primary
(15 items)
 
Secondary
(13 items)
 
In progress
(10 items)
 
CPUMotherboardGraphicsRAM
5820K @ 4.2/3.5GHz core/uncore, 1.175/1.15v Gigabyte X99 SOC Champion (F22n) Gigabyte AORUS GTX 1080 Ti (F3P) @ 2025/1485, 1... 4x4GiB Crucial @ 2667, 12-12-12-28-T1, 1.34v 
Hard DriveHard DriveHard DriveCooling
Plextor M6e 128GB (fw 1.06) M.2 (PCI-E 2.0 2x) 2x Crucial M4 256GB 4x WD Scorpio Black 500GB Noctua NH-D15 
OSMonitorKeyboardPower
Windows 7 Professional x64 SP1 BenQ BL3200PT Filco Majestouch Tenkeyless (MX Brown) Corsair RM1000x 
CaseMouseAudio
Fractal Design Define R4 Logitech G402 Realtek ALC1150 + M-Audio AV40 
CPUMotherboardGraphicsRAM
X5670 @ 4.4/3.2GHz core/uncore, 1.36 vcore, 1.2... Gigabyte X58A-UD5 r2.0 w/FF3mod10 BIOS Sapphire Fury Nitro OC+ @ 1053/500, 1.225vGPU/1... 2x Samsung MV-3V4G3D/US @ 2000, 10-11-11-30-T1,... 
RAMHard DriveHard DriveHard Drive
1x Crucial BLT4G3D1608ET3LX0 @ 2000, 10-11-11-3... OCZ (Toshiba) Trion 150 120GB Hyundai Sapphire 120GB 3x Hitachi Deskstar 7k1000.C 1TB 
CoolingOSPowerCase
Noctua NH-D14 Windows 7 Pro x64 SP1 Antec TP-750 Fractal Design R5 
Audio
ASUS Xonar DS 
CPUMotherboardGraphicsRAM
i7-6800K @ 4.3/3.5GHz core/uncore, 1.36/1.2v ASRock X99 OC Formula (P3.10) GTX 780 (temporary) 4x4GiB Crucial DDR4-2400 @ 11-13-12-28-T2, 1.33v 
Hard DriveHard DriveCoolingOS
Intel 600p 256GB NVMe 2x HGST Travelstar 7k1000 1TB Corsair H55 (temporary) Windows Server 2016 Datacenter 
PowerCase
Seasonic SS-860XP2 Corsair Carbide Air 540 
  hide details  
Reply
post #247 of 580
Quote:
Originally Posted by Raghar View Post

I don't think AMD wants 140 W TDP, or locking theirs chips. That would hurt AMD against Intel quite a bit.
I don't think Zen will clock that well since they are using the same aggressive clock gating that was present on the Fury. Don't know exactly how it translates to overclockability but Fury wasn't a great overclocker under any circumstances.
Gaming Rig
(15 items)
 
  
CPUMotherboardGraphicsRAM
Intel i5-4670K Asrock Z87 Extreme6 Powercolor RX 480 G-skill Ripjaws x-series  
Hard DriveOptical DriveCoolingOS
Adata SPG 900X adata 120GB Antec H2O 620 Windows 10 
MonitorKeyboardPowerCase
MG279Q 27inch free sync CM Storm quickfire Seasonic 650W Fractal Design R5 
MouseMouse PadAudio
Logitech G502 ZERO Sennheiser PC-363D 
  hide details  
Reply
Gaming Rig
(15 items)
 
  
CPUMotherboardGraphicsRAM
Intel i5-4670K Asrock Z87 Extreme6 Powercolor RX 480 G-skill Ripjaws x-series  
Hard DriveOptical DriveCoolingOS
Adata SPG 900X adata 120GB Antec H2O 620 Windows 10 
MonitorKeyboardPowerCase
MG279Q 27inch free sync CM Storm quickfire Seasonic 650W Fractal Design R5 
MouseMouse PadAudio
Logitech G502 ZERO Sennheiser PC-363D 
  hide details  
Reply
post #248 of 580
Quote:
Originally Posted by Tobiman View Post

I don't think Zen will clock that well since they are using the same aggressive clock gating that was present on the Fury. Don't know exactly how it translates to overclockability but Fury wasn't a great overclocker under any circumstances.
That explains that 3.05 GHz clock in engineering samples.
post #249 of 580
Quote:
Originally Posted by TheLAWNOOB View Post

Well, AMD claimed 40% IPC increase. You can't possibally expect AMD to deliver more than what they claim.
Depends on where that 40% IPC increase is. If it's on Integer performance, where Bulldozer is actually fairly strong, then that would put it up with Intel's offerings. And considering the FP performance should no longer be gimped simply due to switching from Bulldozer to a more Intel like uArch, then the relative IPC increase in that area should be astronomical.

If it's 40% on FP performance, then RIP.

Edit: I found where you got your numbers from, Cinebench. A floating point heavy application. If the integer performance is increased by 40% from Excavator to Zen, you're looking at 80% increase in floating point easily.
Edited by lolerk52 - 8/19/16 at 4:25pm
post #250 of 580
Quote:
Originally Posted by Tobiman View Post

I don't think Zen will clock that well since they are using the same aggressive clock gating that was present on the Fury. Don't know exactly how it translates to overclockability but Fury wasn't a great overclocker under any circumstances.

It's a wait and see, CPU cores are more flexible than GPU. It really depends on the base voltage and the multipliers and co-dependencies for how frequencies are handled.
Power Tower
(22 items)
 
SteamBox
(9 items)
 
Doge Miner
(7 items)
 
CPUMotherboardGraphicsRAM
Ryzen 1700X AX370-Gaming 5 AMD Radeon R9 200 Series G.Skill DDR4-2400 
RAMRAMRAMHard Drive
G.Skill DDR4-2400 G.Skill DDR4-2400 G.Skill DDR4-2400 Samsung 840 Pro 
Hard DriveHard DriveHard DriveHard Drive
CX300 Crucial 480GB Toshiba 4TB Toshbia 4TB Western Digital Black 1TB 
CoolingOSMonitorMonitor
h110i Windows 10 42" LG TV 20" Digitizer ASUS 
KeyboardPowerCaseMouse
Corsair Vengeance Mechanical Keyboard  850watt Vampire Gold Rated NZXT S340 Elite Corsair RGB FPS Mouse 
Mouse PadAudio
Borderlands Mousepad Realtek HD 
  hide details  
Reply
Power Tower
(22 items)
 
SteamBox
(9 items)
 
Doge Miner
(7 items)
 
CPUMotherboardGraphicsRAM
Ryzen 1700X AX370-Gaming 5 AMD Radeon R9 200 Series G.Skill DDR4-2400 
RAMRAMRAMHard Drive
G.Skill DDR4-2400 G.Skill DDR4-2400 G.Skill DDR4-2400 Samsung 840 Pro 
Hard DriveHard DriveHard DriveHard Drive
CX300 Crucial 480GB Toshiba 4TB Toshbia 4TB Western Digital Black 1TB 
CoolingOSMonitorMonitor
h110i Windows 10 42" LG TV 20" Digitizer ASUS 
KeyboardPowerCaseMouse
Corsair Vengeance Mechanical Keyboard  850watt Vampire Gold Rated NZXT S340 Elite Corsair RGB FPS Mouse 
Mouse PadAudio
Borderlands Mousepad Realtek HD 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Hardware News
Overclock.net › Forums › Industry News › Hardware News › [AMD] Breakthrough Performance of “Zen” (Head to Head with Broadwell-E!)