Jump to content

AMD Zen to Bring Greater than 40% IPC Increase, New Reports Suggests

Mr_Troll

If? Expect an 8-core Zen CPU to cost less than a 5820K.

 

AMD is not Intel.

 

Though I agree, it's not because they wouldn't if they could.  Even if ZEN were to outperform intels latest and greatest, they do not have the brand power to get away with intel pricing.   Regardless of the data.  AMD would need to play a consistent good game year after year for the better part of a decade to have the clout needed to try and pull off intels 60 point margins.

Relaying on the odd homerun every now and again just makes them a wildcard in peoples eyes and doesn't win you the series, just the odd game here and there.

 

No, if AMD deliver's on their promise, expect to pay $400+ for a Zen 8-core.

 

The only reason AMD's CPUs have been so cheap is because they can't compete well, not because AMD are just good guys. 

 

That's not really it either.  AMD has had better performance than intel before and it still didn't pan out that way.  Though I'm sure it would be more comparable, I don't see them getting away with tit for tat pricing without hurting themselves even more.  It's not like the mobos that support AMD are held in high regard either.

Link to comment
Share on other sites

Link to post
Share on other sites

If? Expect an 8-core Zen CPU to cost less than a 5820K.

 

AMD is not Intel.

 

 

No, if AMD deliver's on their promise, expect to pay $400+ for a Zen 8-core.

 

The only reason AMD's CPUs have been so cheap is because they can't compete well, not because AMD are just good guys.

 

This is true. It's easy to forget that AMD use to price their CPUs comparativly to Intel not that long ago.

Even then, a $600 Zen 8 core is going to be a lot cheaper than a 5960X or even a 5930k. Still a win in my books.

Would then, stands to reason a 6 core variant would be even cheaper again.

Thing to remember is the pricing we are seeing from Intel is because AMD doesn't have a competative high end CPU.

 

 

Zen won't be as cheap as the current range of AMD cpus, but they don't have to be as long as they perform. if AMD

can match intel on performance, hopefully they try to slightly under cut Intel to gain back market share.

Link to comment
Share on other sites

Link to post
Share on other sites

UGH!

 

Some reading comprehension would be really helpful!

 

This sums up all of AMD's claims regarding Zen's performance:

 

1. +40% IPC over Excavator

2. > 40% IPC over Piledriver

 

Confused??

 

Every time the "greater than 40%" comment is made by AMD, it is purely in reference to "datacenter."   Datacenter is Piledriver, simple as that.

 

AMD has never claimed >40% IPC over Excavator, and the performance target/expectation has NOT been changed by AMD.  Not once.

 

Knowing what I "know" about Zen's design*, I'm actually confused as to how it is only offering 40% higher IPC than Excavator.  But, then again, it was alluded to by AMD previously that that this figure was the pure architectural component of the added IPC and did not include process improvements... so that could be another 5% or more on top, enough to push it just above Haswell.

Link to comment
Share on other sites

Link to post
Share on other sites

AMD 8-core with Haswell-or-better IPC? Sign me up

Ya sign me up too! :) I love Intel 'cause they perform better, but I want AMD to be awesome in the CPU arena so Intel has a highly worthy competitor again.

RIGZ

Spoiler

Starlight (Current): AMD Ryzen 9 3900X 12-core CPU | EVGA GeForce RTX 2080 Ti Black Edition | Gigabyte X570 Aorus Ultra | Full Custom Loop | 32GB (4x8GB) Dominator Platinum SE Blackout #338/500 | 1TB + 2TB M.2 NVMe PCIe 4.0 SSDs, 480GB SATA 2.5" SSD, 8TB 7200 RPM NAS HDD | EVGA NU Audio | Corsair 900D | Corsair AX1200i | Corsair ML120 2-pack 5x + ML140 2-pack

 

The Storm (Retired): Intel Core i7-5930K | Asus ROG STRIX GeForce GTX 1080 Ti | Asus ROG RAMPAGE V EDITION 10 | EKWB EK-KIT P360 with Hardware Labs Black Ice SR2 Multiport 480 | 32GB (4x8GB) Dominator Platinum SE Blackout #338/500 | 480GB SATA 2.5" SSD + 3TB 5400 RPM NAS HDD + 8TB 7200 RPM NAS HDD | Corsair 900D | Corsair AX1200i + Black/Blue CableMod cables | Corsair ML120 2-pack 2x + NB-BlackSilentPro PL-2 x3

STRONK COOLZ 9000

Spoiler

EK-Quantum Momentum X570 Aorus Master monoblock | EK-FC RTX 2080 + Ti Classic RGB Waterblock and Backplate | EK-XRES 140 D5 PWM Pump/Res Combo | 2x Hardware Labs Black Ice SR2 480 MP and 1x SR2 240 MP | 10X Corsair ML120 PWM fans | A mixture of EK-KIT fittings and EK-Torque STC fittings and adapters | Mayhems 10/13mm clear tubing | Mayhems X1 Eco UV Blue coolant | Bitspower G1/4 Temperature Probe Fitting

DESK TOIS

Spoiler

Glorious Modular Mechanical Keyboard | Glorious Model D Featherweight Mouse | 2x BenQ PD3200Q 32" 1440p IPS displays + BenQ BL3200PT 32" 1440p VA display | Mackie ProFX10v3 USB Mixer + Marantz MPM-1000 Mic | Sennheiser HD 598 SE Headphones | 2x ADAM Audio T5V 5" Powered Studio Monitors + ADAM Audio T10S Powered Studio Subwoofer | Logitech G920 Driving Force Steering Wheel and Pedal Kit + Driving Force Shifter | Logitech C922x 720p 60FPS Webcam | Xbox One Wireless Controller

QUOTES

Spoiler

"So because they didn't give you the results you want, they're biased? You realize that makes you biased, right?" - @App4that

"Brand loyalty/fanboyism is stupid." - Unknown person on these forums

"Assuming kills" - @Moondrelor

"That's not to say that Nvidia is always better, or that AMD isn't worth owning. But the fact remains that this forum is AMD biased." - @App4that

"I'd imagine there's exceptions to this trend - but just going on mine and my acquaintances' purchase history, we've found that budget cards often require you to turn off certain features to get slick performance, even though those technologies are previous gen and should be having a negligible impact" - ace42

"2K" is not 2560 x 1440 

Link to comment
Share on other sites

Link to post
Share on other sites

That's not really it either.  AMD has had better performance than intel before and it still didn't pan out that way.  Though I'm sure it would be more comparable, I don't see them getting away with tit for tat pricing without hurting themselves even more.  It's not like the mobos that support AMD are held in high regard either.

The last time AMD had better performance than Intel, AMD wasn't hurting for cash.

Zen's part of their payout if they actually deliver.

 

We'll have to wait and see then but i still have no doubt we won't be seeing a Zen 8-core under $300.

"If you ain't first, you're last"

Link to comment
Share on other sites

Link to post
Share on other sites

No, from excavator. Excavator is 10% ahead of bulldozer. 

 

Excavator is closer to 28% faster than Bulldozer, actually.  Zen should be about 78% faster than Bulldozer.  This would put it almost exactly dead-even with Haswell, though performance increases are never uniform, so AMD may be seeing +100% somewhere and -20% elsewhere and calling it +40%.  It's accurate, but nowhere near enough information.

Link to comment
Share on other sites

Link to post
Share on other sites

yes

 

it goes like this:

Bulldozer (First gen FX)

Piledriver (second gen FX, current GEN in sale, aka 43xx, 63xx, 83xx, 9370, 9590) +10% IPC over Bulldozer

Steamroller (APUs) +7% IPC over Piledriver

Excavator (APUs) +15% (AMD claims) IPC over Steamroller

 

If these numbers are real, and most are except we dont REALLY know how Carrizo adds up as there is ZERO CPU performance reviews from any reputable source, then 40% over Excavator will land ZEN at 3-5% above Haswell in IPC....

 

My numbers are:

Bulldozer:   100%Piledriver:  109%Steamroller: 116% (120% w/ L3*)Excavator:   128% (132% w/ L3*)Zen:         179%~185%* L3 simulation based upon performance impact via AMD's APUs head-to-head against the FX series.

Excavator is nearly spot-on with Penryn and Zen is right around Haswell.

Link to comment
Share on other sites

Link to post
Share on other sites

yeah but stable at 90C

And 2000 watts.

CPU: Intel Core i7 7820X Cooling: Corsair Hydro Series H110i GTX Mobo: MSI X299 Gaming Pro Carbon AC RAM: Corsair Vengeance LPX DDR4 (3000MHz/16GB 2x8) SSD: 2x Samsung 850 Evo (250/250GB) + Samsung 850 Pro (512GB) GPU: NVidia GeForce GTX 1080 Ti FE (W/ EVGA Hybrid Kit) Case: Corsair Graphite Series 760T (Black) PSU: SeaSonic Platinum Series (860W) Monitor: Acer Predator XB241YU (165Hz / G-Sync) Fan Controller: NZXT Sentry Mix 2 Case Fans: Intake - 2x Noctua NF-A14 iPPC-3000 PWM / Radiator - 2x Noctua NF-A14 iPPC-3000 PWM / Rear Exhaust - 1x Noctua NF-F12 iPPC-3000 PWM

Link to comment
Share on other sites

Link to post
Share on other sites

Though I am hopefull for a comeback for the benefit of competition and better consumer pricing, I am as hesitant to believe what the CEO says. It's like the same 'overclockers dream' statement all over again.

ROG X570-F Strix AMD R9 5900X | EK Elite 360 | EVGA 3080 FTW3 Ultra | G.Skill Trident Z Neo 64gb | Samsung 980 PRO 
ROG Strix XG349C Corsair 4000 | Bose C5 | ROG Swift PG279Q

Logitech G810 Orion Sennheiser HD 518 |  Logitech 502 Hero

 

Link to comment
Share on other sites

Link to post
Share on other sites

Though I am hopefull for a comeback for the benefit of competition and better consumer pricing, I am as hesitant to believe what the CEO says. It's like the same 'overclockers dream' statement all over again.

"Overclocker's dream" should've really been made into a meme like "IT IS THE CURRENT YEAR" was.

 

I mean that was part of the sale's pitch, it's not like someone misquoted Lisa when she called Fiji that.

"If you ain't first, you're last"

Link to comment
Share on other sites

Link to post
Share on other sites

"Overclocker's dream" should've really been made into a meme like "IT IS THE CURRENT YEAR" was.

 

I mean that was part of the sale's pitch, it's not like someone misquoted Lisa when she called Fiji that.

Well, overclocking it -was- a dream :P  A near impossible dream that yielded very little if you could even do it.

Link to comment
Share on other sites

Link to post
Share on other sites

Yes, the 40% increase is independent of the new process node, so that also has to be taken into account for any performance improvements.

I still see two very large problems for AMD and potential customers.

The first one being that people expect dirt cheap enthusiast hexa and octa core processors. This simply won't happen. AMD needs to make lots of money. While I don't know how many octa core SKUs we'll see, I would be surprised if you saw one under $600. If you need the cores, you need to give the cash. Intel sure as hell won't undercut (perhaps only in OEM-like bulk orders, as they did in the past), so AMD will set the price has high as they reasonably can and if you're paying a thousand dollars for 8 cores from Intel, why would AMD price theirs at under half when you don't have any alternative? They'll most definitely undercut Intel somewhat, but a couple hundred dollars at the most for a similar top SKU.

The next problem is extremely scary for AMD. People want AMD to deliver a competitive product. Why? So that they can purchase cheaper and faster Intel and Nvidia products.

These two reasons are why AMD just can't win. The environment is so toxic for them. No matter what they do, people will stick to those who essentially fuck them over because that's what's familiar.

Link to comment
Share on other sites

Link to post
Share on other sites

My numbers are:

Bulldozer:   100%Piledriver:  109%Steamroller: 116% (120% w/ L3*)Excavator:   128% (132% w/ L3*)Zen:         179%~185%* L3 simulation based upon performance impact via AMD's APUs head-to-head against the FX series.

Excavator is nearly spot-on with Penryn and Zen is right around Haswell.

excavator is the wildcard.

there is so much "AMD said this" surrounding it. and not a single reliable benchmark.

The few benches we have have been between laptops using Kaveri and Carrizo. Where the Kaveri one has been TDP limited to 35w and the Carrizo part (which CAN be limited to 35w too at max) has been set to 20-25w...

Link to comment
Share on other sites

Link to post
Share on other sites

The next problem is extremely scary for AMD. People want AMD to deliver a competitive product. Why? So that they can purchase cheaper and faster Intel and Nvidia products.

These two reasons are why AMD just can't win. The environment is so toxic for them. No matter what they do, people will stick to those who essentially fuck them over because that's what's familiar.

 

Thing is though, AMD was never and will never get those peoples money anyway, so it isn't like they lost it. Whats been happening the last couple of years is that the people that do go by price and performance (for CPUs at least) are going to Intel.

Link to comment
Share on other sites

Link to post
Share on other sites

Even if this is true, I it wouldn't cost what an FX8350 costs now. Why? Because it would be competitive to Intel's 8-core flagship and guess what... It will cost around the same price. The new Zen chips will also have new sockets and boards and if it was anything like in the age of the phenom, a decent board would be quite expensive too. 

CPU: AMD Ryzen 9 - 3900x @ 4.4GHz with a Custom Loop | MBO: ASUS Crosshair VI Extreme | RAM: 4x4GB Apacer 2666MHz overclocked to 3933MHz with OCZ Reaper HPC Heatsinks | GPU: PowerColor Red Devil 6900XT | SSDs: Intel 660P 512GB SSD and Intel 660P 1TB SSD | HDD: 2x WD Black 6TB and Seagate Backup Plus 8TB External Drive | PSU: Corsair RM1000i | Case: Cooler Master C700P Black Edition | Build Log: here

Link to comment
Share on other sites

Link to post
Share on other sites

Still can't believe there going to use the FX branding for Zen... FFS its a new thing, use a new name, not the one that is associated with the shitty bulldozer CPU's, i swear if Zen delivers on performance, the FX name is going to be the reason they flock because people is going to associate them with the old FX CPU's thinking there going to be shit. Nintendo made that mistake with the Wii U and we all know the absurds amounts of confusion upon the consumer which hurt sales.

this is one of the greatest thing that has happened to me recently, and it happened on this forum, those involved have my eternal gratitude http://linustechtips.com/main/topic/198850-update-alex-got-his-moto-g2-lets-get-a-moto-g-for-alexgoeshigh-unofficial/ :')

i use to have the second best link in the world here, but it died ;_; its a 404 now but it will always be here

 

Link to comment
Share on other sites

Link to post
Share on other sites

Still can't believe there going to use the FX branding for Zen... FFS its a new thing, use a new name, not the one that is associated with the shitty bulldozer CPU's, i swear if Zen delivers on performance, the FX name is going to be the reason they flock because people is going to associate them with the old FX CPU's thinking there going to be shit.

wrong

FX rocks

FX kicked pentuim 4 butt in the mid 2000s leading to the most successful period in AMD's history. FX was the branding used for the Athlon 64 flagship

Link to comment
Share on other sites

Link to post
Share on other sites

Should name it Photon.

CONSOLE KILLER: Pentium III 700mhz . 512MB RAM . 3DFX VOODOO 3 SLi

 

 

Link to comment
Share on other sites

Link to post
Share on other sites

This made me curious how much ahead Intel is in terms of IPC.

So I went to Anandtech's bench and compared the i5 6600K vs the FX-8320. Then I cut the AMD's score in half and then added a bit (on the multithreaded tasks). That's to make up for the 4 Intel cores vs 8 AMD cores in Bulldozer.

 

So here are some scores (same clock speed, and rough compensation for the extra cores on the AMD chip):

Intel (Cinebench R10): 8372

AMD (Cinebench R10) : 3987

Intel lead: 109%

 

Intel (Cinebench 11.5): 1.92

AMD (Cinebench 11.5) : 1.05

Intel lead: 83%

 

Intel (x264 1st pass): 133.48

AMD (x264 1st pass) : 43

Intel lead: 210%

 

Intel (x264 2nd pass): 42.8

AMD (x264 2nd pass) : 20

Intel lead: 113%

 

Intel (7-zip): 18090

AMD (7-zip) : 11000

Intel lead: 64%

 

 

So AMD is waaaay behind in IPC (in these benchmarks). A 40% IPC over Bulldozer will still put AMD a lot behind Intel in terms of IPC, but if AMD releases true 8 cores then they will actually be very competitive with Intel's quad cores.

Sadly I think the chances of AMD not only exceeding their 40% IPC goal, matching Intel's features, everything working out as planned (so no major issues being revealed) are fairly small.

Also, the "greater than 40% IPC" claim is just what they might achieve if they keep improving the architecture at the rate they are currently doing.

 

It would be fantastic if AMD could trade blows with Intel in the CPU market again, but it's been so long since they did I have kind of lost hope. Also, AMD have been caught talking bullshit so many times in the past I wouldn't be surprised if not even half of the stuff they claim turns out to be true.

Link to comment
Share on other sites

Link to post
Share on other sites

Yes, the 40% increase is independent of the new process node, so that also has to be taken into account for any performance improvements.

 

IPC has nothing to do with nodes what so ever. It's only about architecture and numbers of cores/transistors.

 

This made me curious how much ahead Intel is in terms of IPC.

So I went to Anandtech's bench and compared the i5 6600K vs the FX-8320. Then I cut the AMD's score in half and then added a bit (on the multithreaded tasks). That's to make up for the 4 Intel cores vs 8 AMD cores in Bulldozer.

 

So AMD is waaaay behind in IPC (in these benchmarks). A 40% IPC over Bulldozer will still put AMD a lot behind Intel in terms of IPC, but if AMD releases true 8 cores then they will actually be very competitive with Intel's quad cores.

Sadly I think the chances of AMD not only exceeding their 40% IPC goal, matching Intel's features, everything working out as planned (so no major issues being revealed) are fairly small.

Also, the "greater than 40% IPC" claim is just what they might achieve if they keep improving the architecture at the rate they are currently doing.

 

It would be fantastic if AMD could trade blows with Intel in the CPU market again, but it's been so long since they did I have kind of lost hope. Also, AMD have been caught talking bullshit so many times in the past I wouldn't be surprised if not even half of the stuff they claim turns out to be true.

 

You have no clue what IPC is do you? That #CEO math has nothing to do with reality. The fact that you claim ZEN needs to be octo core to compete with Skylake quad core, should tell you that you are way off.

 

Bolldozer is still a package of 2 cores (ALU) and one FPU, which is similar to a normal core. You can't just halve it and call it a day.

Watching Intel have competition is like watching a headless chicken trying to get out of a mine field

CPU: Intel I7 4790K@4.6 with NZXT X31 AIO; MOTHERBOARD: ASUS Z97 Maximus VII Ranger; RAM: 8 GB Kingston HyperX 1600 DDR3; GFX: ASUS R9 290 4GB; CASE: Lian Li v700wx; STORAGE: Corsair Force 3 120GB SSD; Samsung 850 500GB SSD; Various old Seagates; PSU: Corsair RM650; MONITOR: 2x 20" Dell IPS; KEYBOARD/MOUSE: Logitech K810/ MX Master; OS: Windows 10 Pro

Link to comment
Share on other sites

Link to post
Share on other sites

This made me curious how much ahead Intel is in terms of IPC.

So I went to Anandtech's bench and compared the i5 6600K vs the FX-8320. Then I cut the AMD's score in half and then added a bit (on the multithreaded tasks). That's to make up for the 4 Intel cores vs 8 AMD cores in Bulldozer.

 

So here are some scores (same clock speed, and rough compensation for the extra cores on the AMD chip):

Intel (Cinebench R10): 8372

AMD (Cinebench R10) : 3987

Intel lead: 109%

 

Intel (Cinebench 11.5): 1.92

AMD (Cinebench 11.5) : 1.05

Intel lead: 83%

 

Intel (x264 1st pass): 133.48

AMD (x264 1st pass) : 43

Intel lead: 210%

 

Intel (x264 2nd pass): 42.8

AMD (x264 2nd pass) : 20

Intel lead: 113%

 

Intel (7-zip): 18090

AMD (7-zip) : 11000

Intel lead: 64%

 

 

So AMD is waaaay behind in IPC (in these benchmarks). A 40% IPC over Bulldozer will still put AMD a lot behind Intel in terms of IPC, but if AMD releases true 8 cores then they will actually be very competitive with Intel's quad cores.

Sadly I think the chances of AMD not only exceeding their 40% IPC goal, matching Intel's features, everything working out as planned (so no major issues being revealed) are fairly small.

Also, the "greater than 40% IPC" claim is just what they might achieve if they keep improving the architecture at the rate they are currently doing.

 

It would be fantastic if AMD could trade blows with Intel in the CPU market again, but it's been so long since they did I have kind of lost hope. Also, AMD have been caught talking bullshit so many times in the past I wouldn't be surprised if not even half of the stuff they claim turns out to be true.

 

The FX-8320 is two generations behind for AMD's core tech.  About 15~18% slower than Excavator - the baseline for Zen's performance.  Zen will, of course, not catch up to Skylake - it should be close to Haswell - or about 64% faster than the FX-8320 in single threaded workloads (with that gain very unevenly distributed).  It's hard to guess how Zen will perform in any particular benchmark from the average performance number.

 

Sadly, Anandtech's numbers don't have turbo disabled, which gives some vagueness in the value of the numbers for IPC comparison, but there are sources which disclosed the clockrates achieved and those who overclocked or disabled turbo, preventing the shifting.  Those results are quite predictable.

 

Excavator is about as fast as the Core 2 (Penryn), so we really only need to see what Intel has done, on average, since then to figure out where Zen will land (Penryn + 40%):

 

intel_ipc_claimsvbench.jpg

Link to comment
Share on other sites

Link to post
Share on other sites

excavator is the wildcard.

there is so much "AMD said this" surrounding it. and not a single reliable benchmark.

The few benches we have have been between laptops using Kaveri and Carrizo. Where the Kaveri one has been TDP limited to 35w and the Carrizo part (which CAN be limited to 35w too at max) has been set to 20-25w...

 

There are many sources for Excavator numbers.  Most, rightly, compare to previous mobile FX numbers.  Some show more of an advantage than others because their prior numbers are wrong.  I base my numbers on the test platform benchmark results.  So no throttling, no artificial limitations imposed by laptops.

 

Awesomely, we will be getting better numbers on how Excavator will work on a more appropriate fabrication process relatively soon (months).  Bristol Ridge AM4 APUs will be giving us a good insight.

Link to comment
Share on other sites

Link to post
Share on other sites

This made me curious how much ahead Intel is in terms of IPC.

So I went to Anandtech's bench and compared the i5 6600K vs the FX-8320. Then I cut the AMD's score in half and then added a bit (on the multithreaded tasks). That's to make up for the 4 Intel cores vs 8 AMD cores in Bulldozer.

 

So here are some scores (same clock speed, and rough compensation for the extra cores on the AMD chip):

Intel (Cinebench R10): 8372

AMD (Cinebench R10) : 3987

Intel lead: 109%

 

Intel (Cinebench 11.5): 1.92

AMD (Cinebench 11.5) : 1.05

Intel lead: 83%

 

Intel (x264 1st pass): 133.48

AMD (x264 1st pass) : 43

Intel lead: 210%

 

Intel (x264 2nd pass): 42.8

AMD (x264 2nd pass) : 20

Intel lead: 113%

 

Intel (7-zip): 18090

AMD (7-zip) : 11000

Intel lead: 64%

 

 

So AMD is waaaay behind in IPC (in these benchmarks). A 40% IPC over Bulldozer will still put AMD a lot behind Intel in terms of IPC, but if AMD releases true 8 cores then they will actually be very competitive with Intel's quad cores.

Sadly I think the chances of AMD not only exceeding their 40% IPC goal, matching Intel's features, everything working out as planned (so no major issues being revealed) are fairly small.

Also, the "greater than 40% IPC" claim is just what they might achieve if they keep improving the architecture at the rate they are currently doing.

 

It would be fantastic if AMD could trade blows with Intel in the CPU market again, but it's been so long since they did I have kind of lost hope. Also, AMD have been caught talking bullshit so many times in the past I wouldn't be surprised if not even half of the stuff they claim turns out to be true.

 

Your math is a bit wrong. Because you use the difference between the intel value. Meaning you "choose" intels score as the "correct" answer.

The real answer is the average between the two. Because you cannot say processor X is the "correct product". Simply because that would insinuate that anything but YOUR chosen value is wrong, including other intel products scoring higher or lower.

 

Instead, i urge you to look at single core CB R11.5

because if you use R11.5, you can compare Original Bulldozer, aka 81xx, Piledriver aka 83xx, Steamroller aka A10 7850k/860k, Sandy, Ivy, Haswell, broadwell, Skylake.

 

In short, to save you the maths

If we use i5s, which is generally same clock speeds as FX 8 series (3.2-3.5Ghz ish)... then Haswell is roughly 70% ahead of Piledriver (8350) in single core performance.

But why look at single core?

Because of CMT.

 

During multi.treading, CMT doesnt always function as 8 "cores". It only does so with the correct workload. If there is a need for integer operations or pure 32bit float, one of the ALUs can be put on hold to complete the task faster. This is done automatically inside the CMT based architecture.

Why do we know this to tbe true?

there has been some tests done, i dont remember where i read this. They tested workload latencies within the CPU. During different sequential workloads, FX performs normally, however during mixed workloads, certain tasks have notably increased latency. Suggesting that something was put on hold to complete the other task.

This is why it is 4 modules and 8 threads. Because under given conditions, the whole module is needed to do a task rather then the ALU which in and of itself is only capable of lighter workloads.

 

Due to cache split, latencies and whatnot, simply cutting score in half wont do. There is too many odd things at work inside the CMT core to make such a method remotely accurate.

For instance, During benchmarks, if you disable certain ALUs (for my FX8320 it was 0,2,4,6) in task manager, forcing CB R15 to use the other threads, i scored higher then when forcing CB R15 to use thread 1,3,5,7.... This suggests that resources inside the FX cores are indeed NOT equally split, as AMD claims. And that the pipeline is marginally shorter using one ALU rather then the other.

Link to comment
Share on other sites

Link to post
Share on other sites

You have no clue what IPC is do you?

Instructions per clock. How much work can be done in one clock cycle.

 

That #CEO math has nothing to do with reality. The fact that you claim ZEN needs to be octo core to compete with Skylake quad core, should tell you that you are way off.

 

Bolldozer is still a package of 2 cores (ALU) and one FPU, which is similar to a normal core. You can't just halve it and call it a day.

That is not #CEO math.

It is a (very rough) estimate. You can't count the octa core AMD chip as a single core, especially not in benchmarks such as the 7-zip one which is very ALU dependent. In those scenarios the AMD chip might as well be an octa core. I did however bump the AMD number up a bit after cutting it in half. That was to make up for imperfect scaling and other minor differences.

 

Why would saying Zen needs to be an 8 core to compete with Skylake be way off? We don't have Zen so anyone claiming that it won't be is just making stuff up.

By the way, I did not "claim" it needed up to be an octa core to keep up. That's just you projecting very hard.

 

 

 

 

 

The FX-8320 is two generations behind for AMD's core tech.  About 15~18% slower than Excavator - the baseline for Zen's performance.  Zen will, of course, not catch up to Skylake - it should be close to Haswell - or about 64% faster than the FX-8320 in single threaded workloads (with that gain very unevenly distributed).  It's hard to guess how Zen will perform in any particular benchmark from the average performance number.

 

Sadly, Anandtech's numbers don't have turbo disabled, which gives some vagueness in the value of the numbers for IPC comparison, but there are sources which disclosed the clockrates achieved and those who overclocked or disabled turbo, preventing the shifting.  Those results are quite predictable.

 

Excavator is about as fast as the Core 2 (Penryn), so we really only need to see what Intel has done, on average, since then to figure out where Zen will land (Penryn + 40%):

 

<image>

Good point, although I am pretty skeptical of that benchmark. Nehalem to Sandy Bridge was a huge leap and in that graph it only shows it being like a 10% increase in IPC.

Also, do we have good numbers for how Excavator performs, and are we sure AMD means 40% increase over Excavator and not over Bulldozer?

 

 

 

 

Your math is a bit wrong. Because you use the difference between the intel value. Meaning you "choose" intels score as the "correct" answer.

The real answer is the average between the two. Because you cannot say processor X is the "correct product". Simply because that would insinuate that anything but YOUR chosen value is wrong, including other intel products scoring higher or lower.

That makes no sense...

 

Intel was leading in the benchmarks.

We want to know how much AMD has to improve to match them.

We calculate how much that is in %.

 

For example:

If we have 80 and want to get to 100, we need to add 25%.

25% of 80 is 20.

80 + 20 is 100. Therefore, we need 25% more.

That's how I counted. If Intel scores 8372 and AMD scores 3987 then we need to add 109% to AMD's score to match Intel's. That type of math is 100% correct.

Link to comment
Share on other sites

Link to post
Share on other sites

Is IPC actually relevant to direct real world performance. Or is it just another big number we can all gawp at.

 (\__/)

 (='.'=)

(")_(")  GTX 1070 5820K 500GB Samsung EVO SSD 1TB WD Green 16GB of RAM Corsair 540 Air Black EVGA Supernova 750W Gold  Logitech G502 Fiio E10 Wharfedale Diamond 220 Yamaha A-S501 Lian Li Fan Controller NHD-15 KBTalking Keyboard

Link to comment
Share on other sites

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now


×