• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

FEATURED AMD ZEN Discussion (Previous Rumor Thread)

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
Never gonna happen.
Intel has 100 times the R&D money that AMD does.
It will never get any better, because you can't divert income that you're not getting back into R&D. Unless AMD gets some backing cash from another party, they'll stay in a slow death spiral.
 
I'm referring to when AMD handled both design and manufacturing of their chips. AMD's 1st FX chips were the same or faster than the P4, while running 1 GHz slower. AMD came out with their x2 series and beat the crap out of Intels Dual Core Prescott's. At that time, the pricing of CPU's were reversed, with AMD on TOP and Intel slashing prices just to sell CPU's :)

Intel was never slashing prices as they never had to. AMD was always 30-60% cheaper for similar performance. So why it was always like that ? Because 70%+ computers are designed for office work and there Intel always had ~90% market.
Additionally Intel is the biggest GPU manufacturer with over 60% market shares and all of that are IGP ( there were discrete graphics too but maybe 10 years ago ). AMD has faster IGP ... really, who cares ? Barely anyone needs it.
Bigger companies don't trust AMD. In business generally barely anyone consider AMD as a good option. AMD was for years related with home PCs and home entertainment.

AMD servers were never popular but at least these 7-10 years ago they had some market %. Now they have maybe 5% and barely any company has AMD series servers in mass production. I mean if you ask about Intel you get offer of 10-15 models and most available in 1-7 days. When you ask about AMD then you get answer it's not available or you have to wait 30 days. In Polish distribution it's hard to find any AMD based server.

I was talking with IBM engineer couple of years ago ( well, about 7-8 ) and he said that they keep 2 AMD server lines ( vs 13 or something based on Intel ) for special customers, only because in 4 CPU configs AMD had much faster internal bus which let them to communicate with every CPU like CPU 1 could pass data to 2/3/4 and back while in Intel CPU 1 could pass data to 2/3 but not to 4 so 2/3 had to pass it to 4 and the same way back. It's hard to describe it and I don't really remember details. Anyway when you were working on many really small files in highly multithreaded environment then AMD was faster. These 2 server lines disappeared some time ago.
 
Last edited:
I'm referring to when AMD handled both design and manufacturing of their chips. AMD's 1st FX chips were the same or faster than the P4, while running 1 GHz slower. AMD came out with their x2 series and beat the crap out of Intels Dual Core Prescott's. At that time, the pricing of CPU's were reversed, with AMD on TOP and Intel slashing prices just to sell CPU's :)

I think you're forgetting that Intel also effectively paid resellers to use their chips to retain market share. Something which was proven and ruled as an illegal business practice. It had its desired effect of keeping Intel in systems sold to business and denying AMD revenue while they were ahead. Governments around the world fined Intel billions of dollars for this, funny though AMD never saw a penny. On the other hand, there's now an agreement between Intel and AMD that neither company will ever repeat that.....

Funny how people forget.....

IF AMD can get a better performing chip for a few years then hopefully this time they'll stand a chance, unless Intel resorts to illegal business practices once more. Have to wonder how many people in government were paid off to look the other way while that was happening really.
 
Intel was never slashing prices as they never had to. AMD was always 30-60% cheaper for similar performance. So why it was always like that ? Because 70%+ computers are designed for office work and there Intel always had ~90% market.
Additionally Intel is the biggest GPU manufacturer with over 60% market shares and all of that are IGP ( there were discrete graphics too but maybe 10 years ago ). AMD has faster IGP ... really, who cares ? Barely anyone needs it.
Bigger companies don't trust AMD. In business generally barely anyone consider AMD as a good option. AMD was for years related with home PCs and home entertainment.

AMD servers were never popular but at least these 7-10 years ago they had some market %. Now they have maybe 5% and barely any company has AMD series servers in mass production. I mean if you ask about Intel you get offer of 10-15 models and most available in 1-7 days. When you ask about AMD then you get answer it's not available or you have to wait 30 days. In Polish distribution it's hard to find any AMD based server.

I was talking with IBM engineer couple of years ago ( well, about 7-8 ) and he said that they keep 2 AMD server lines ( vs 13 or something based on Intel ) for special customers, only because in 4 CPU configs AMD had much faster internal bus which let them to communicate with every CPU like CPU 1 could pass data to 2/3/4 and back while in Intel CPU 1 could pass data to 2/3 but not to 4 so 2/3 had to pass it to 4 and the same way back. It's hard to describe it and I don't really remember details. Anyway when you were working on many really small files in highly multithreaded environment then AMD was faster. These 2 server lines disappeared some time ago.
@Woomack
Like My Wife said, I'm a treasure trove of ancient hardware and also Computer Magazines. Here is a photo of the prices between AMD's and Intel CPU,s "The Picture was taken from CPU ( Computer Power User) magazine:

WHEN WAS THE LAST TIME YOU PAID 1K FOR A DESKTOP AMD CPU......

Just 1.jpg
 
AMD would be lucky to return to profit a Billy by just releasing something "As good as Intels". Even if cheaper it won't be enough. They need to out performance Intel at the high end to going through brand recognition they need to charge the prices they need to return to profitabillity.

Intel handle both design and manufacturers of their chips, while also being years ahead of everyone else in chips fabrication. AMD literally have to charge more to make the same profit per chips as Intel. As I previously mentioned, they have to be better, not just equally.

They don't need to be better than Intel. If that was the case they might as well close up shop now. As has been stated thousands of times Intel spends more money on research and development than AMD grosses annually. I am sure Intel is just coasting now. Working on decreasing TDP.
 
Intel is offering us almost only "fixes" for their last products. Even though they have a lot of money and a lot of time , they make faulty products without as good tests as they were performing in the past. Every new Intel platform has some issues. Even their last server motherboards had stability issues. On one of them after enabling CPU C stages OS was crashing ... I was waiting 3 months for fix.
On desktop boards were problems with chipset, USB, sleep/wake from S state, stability with SATA/SAS controllers ( what caused delays and finally they removed it from 2 chipsets ) etc.

Most users who bought Ivy Bridge see no point to change to Haswell, Broadwell or wait for Skylake. 10-20% CPU performance improvements mean nothing in home PCs. Who cares if game will run 2-3 FPS faster at the cost of new CPU/mobo/memory etc.
Laptops are not getting faster. What is funny, most of them are not even working longer on batteries. It's because hardware is using less power but manufacturers cut costs and sell laptops with worse/smaller capacity batteries. Also most laptop CPUs are now low power ( U series or something like that ) which never reach declared turbo clock so most of them run at about 1.8-2.1GHz. Many manufacturers are even disabling dual channel memory, especially in ultrabooks.
 
Last edited:
Just another example of why Intel needs competition. Why do they even need to do better right now. There's no-one can compete with them presently anyway so everyone will keep buying defective Intel and updated "fixed" products. Any other industry the manufacturer would have to recall and fix at their expense.

Intel is offering us almost only "fixes" for their last products. Even though they have a lot of money and a lot of time , they make faulty products without as good tests as they were performing in the past. Every new Intel platform has some issues. Even their last server motherboards had stability issues. On one of them after enabling CPU C stages OS was crashing ... I was waiting 3 months for fix.
On desktop boards were problems with chipset, USB, sleep/wake from S state, stability with SATA/SAS controllers ( what caused delays and finally they removed it from 2 chipsets ) etc.

Most users who bought Ivy Bridge see no point to change to Haswell, Broadwell or wait for Skylake. 10-20% CPU performance improvements mean nothing in home PCs. Who cares if game will run 2-3 FPS faster at the cost of new CPU/mobo/memory etc.
Laptops are not getting faster. What is funny, most of them are not even working longer on batteries. It's because hardware is using less power but manufacturers cut costs and sell laptops with worse/smaller capacity batteries. Also most laptop CPUs are now low power ( U series or something like that ) which never reach declared turbo clock so most of them run at about 1.8-2.1GHz. Many manufacturers are even disabling dual channel memory, especially in ultrabooks.
 
Any other industry the manufacturer would have to recall and fix at their expense.

Right BUT this is not any other industry. Specs subject to change without notice and then as Woomack says, they drop the troublesome feature and on to the bank. But the overclocking community does not really care since all they want is that HI IPC and better benches. They buy the flaws no matter what as long as the benches hold up.

RGone...
 
Intel don't even care to fix their issues recently. They were replacing faulty B2 revision P67 chipsets but all knew about haswell sleep/USB 3.0 issues before the release and still they were manufacturing these chipsets for over 2 months and later we got them. The same with SATA issues in 9 series chipsets ( Z97 etc ) or constant USB 3.0 issues on all new Intel based boards.

What I mean here is that manufacturers are introducing useless new technologies like USB 3.1, Sata Express etc. while they can't even make old standards work as declared.

My favourite issues are recently related to laptops. I have often something new at work and I see that quality of mobile computers is dropping constantly. At least 50% new laptops are throttling because of overheating when you load CPU to the 100%. Simple XTU benchmark is showing it clearly. It's no big issue when you browse the web but it start to be a problem when you are using it for business.

Other things about what many users have no idea while buying laptops:

Examples:
- probably all card readers are connected to USB 2.0 ports and performance is below 2.0 standard
- video outputs have usually limited display resolution or support only integrated display and 1 monitor max ( even though are D-Sub+HDMI or DP+HDMI ports )
- many new laptops have 1 memory slot, even these 15" so you have dual channel memory controller but can't use it
- batteries have lowered capacity to lower laptop size and costs ( still low wattage CPUs are covering it )
- chipset supports 6+ USB 3.0 ports but manufacturer decide to install 1-2 USB 3.0 ( not even full speed ) and other USB 2.0
- non-business series have rarely available battery replacements other than manufactured for support needs ( in the past it was normal that you could buy it without issues )
- most laptop manufacurers are using low power CPUs - lower performance than older series just because turbo ratio is a myth on new OS with load balancing
- you simply can't find good laptop based on AMD. If it's on AMD then manufacturer made something wrong ... cheaper parts, low time on battery, lack of some features etc. I'm not saying that because I don't like AMD. It simply looks like that - Intel = games and business , AMD = you can browse web and spam something on facebook.
...
List can be much longer and I bet you have your own examples.

Anyway it looks like we have no big choice. Intel is releasing something constantly but in most cases with delays what doesn't mean they make perfect products. On the other hand AMD is barely releasing anything and when they show something then it's usually a fail. All their advantages are in GPUs/APUs which are not selling as good as all think.

I remember comments that AMD should wait some more and release good product but so far they are waiting too long. Already missed chance to beat nvidia when they had problems with GTX970 and soon they will lose next chance when Intel release skylake and new chipsets.
They just waste time on the mobile APU/GPU market about what barely anyone cares. Good in theory, waste of time in real world.
 
Last edited:
Some say that AMD Bought ATI because their offer to buy NVidia had too many stipulations. One of the rumored stipulations was if they bought NVidia, Jen-Hsun Huang had to be president after the merger. The current boss at that time was Hector Ruiz who had a BIG hand in designing AMD's K7 & K8 CPU's. The K10 was still in the design phase and they didn't know that it had bug's. AMD has gone on record to admit that they " Had thought to BIG on the K10 @ 65nm". It was not until AMD transitioned to a .45nm process, that the K10.5 was able to stretch it's legs.
 
I remember the hype around the K15's launch. I also remember "CHEW's OC record with the first K15", he was under a NDA at the time. He could not give details about the CPU but hint's. After the NDA was lifted, he went on record to state, "Think of the Bulldozer as a 4 Core 8 Thread CPU and not as a TRUE 8 CORE Processor".
 
Doing more scrapping around...

...Well when you dig a little more things begin to hop out at you.

For a time I was really considering Hazewell-E setup. Much looking and just felt like a lot of money for performance I did not have to have at all. The lesser Has-E's had 26 PCIe lanes and the big price jump -E processor had 40 PCIe lanes if you ever truly intended to run a few video cards.

I have not seen this said many times in all the zEn hoopla but it is supposed to have only 22 PCIe lanes. Do what? M2 drives and Usb 3.0 and PCIe 3.0 support and multiple fast arse video cards and no freeken lanes? Faster but less is going to work out how? Hale if I know. Maybe things are not as they seem, but I was just bowled over at the number of lanes.

RGone...skisterzzzzzzzzzzz:chair:
 
...Well when you dig a little more things begin to hop out at you.

For a time I was really considering Hazewell-E setup. Much looking and just felt like a lot of money for performance I did not have to have at all. The lesser Has-E's had 26 PCIe lanes and the big price jump -E processor had 40 PCIe lanes if you ever truly intended to run a few video cards.

I have not seen this said many times in all the zEn hoopla but it is supposed to have only 22 PCIe lanes. Do what? M2 drives and Usb 3.0 and PCIe 3.0 support and multiple fast arse video cards and no freeken lanes? Faster but less is going to work out how? Hale if I know. Maybe things are not as they seem, but I was just bowled over at the number of lanes.

RGone...skisterzzzzzzzzzzz:chair:


SOUND OF CAR BRAKE'S

22 Version 3.0 PCIe lanes????
That would be (1) one x16 Video.... Figure in a NEW X-Link @ X6 = 22 ???
Or maybe (2) two x8 video and a NEW X-Link @ X6 = 22 ???

Either way with a NEW Uarch against Intel's Skylake..... = FAILURE:(
 
Last edited:
Okay.... I have slept some since my last post on this :)
If I remember correctly, The OPTERON has 3 HTT links per processor, 1 HTT link when moved to desktop. From Wikipedia "The theoretical transfer rate of HTT Ver 3.1 is 25.6 GB/s (3.2 GHz × 2 transfers per clock cycle × 32 bits per link) per direction, or 51.2 GB/s aggregated throughput, making it faster than most existing bus standard for PC workstations and servers as well as making it faster than most bus standards for high-performance computing and networking.
Okay.... Now we take the 51.2 GB/s and multiply by the current A-Link of (4) FOUR lanes between the CPU & NB = The theoretical transfer rate is 204.8 GB/s. We know this is BS because you can't send & receive on all 4 lanes at the same time.
If we go with a NEW " B-Link with 24 lanes / x 3 (HTT Link)" this would = 8 "B-Link" lanes between the CPU & NB. The 22 does not match a good DOUBLING of the current A-Link of 4 lanes.
This would give a NEW "theoretical transfer rate" of 409.6 GB/s between the CPU & NB. The PR section loves quoting numbers.

Here's a new link on the ZEN.......http://www.kitguru.net/components/c...e-amd-opteron-to-feature-quad-die-mcm-design/
And here is a PIC again of the ZEN with an I/O OF 64 LANES :)

f82923a2b3af2ff3d2aed183b6b61fec_XL.jpg
 
Does anyone have a better picture of the ZEN Core???????
I found one at http://wccftech.com/amd-zen-cpu-core-block/
From this block diagram, it clearly shows AMD using a SMT (Simultaneous multithreading) approach. This is (1) ONE CORE (2) TWO Threads, that have to share all resources of ONE CORE. Die area of second Core < 10%
The current K15 uses a CMT (Clustered Multithreading) approach. This is (2) Two (give or take) 80 % Cores (2) Two Threads, that have to share minimum resources between the TWO CORES. Die area of second Core < ->80%

zen.jpg


Right off the bat just by looking at the block diagram we can tell that Zen will have a substantially higher single threaded performance compared to Excavator and the Bulldozer family. Both in integer and floating point workloads. The bulldozer family will likely maintain a higher total throughput on integer if you compare a single Excavator module with two cores vs a single Zen core. But this is a sacrifice that has to be made for Zen to achieve better per thread performance. And once more, Bulldozer’s Integer throughput was already quite phenomenal as it rivaled Intel’s extreme i7 parts. So the overall throughput once you add all the threads was never a problem, the per thread / single threaded performance was the issue

Comparing both floating point units of Excavator and Zen we can see that AMD has introduced a floating point that’s twice as wide as that of Excavator. Featuring two FMAC 256-bit units that in all probability will be able to fuse together to process 512-bit AVX flaoting point instructions. This is compared to the two 128-bit FMAC units found in AMD’s Bulldozer family, which can either process one 128-bit SIMD instruction each per clock or fuse together to process a single 256-bit AVX instruction per cycle. Hence the assumption above that we could see a similar behavior with Zen’s FPU which would allow both FMACs to cooperate and process 512bit AVX instructions.

Read more: http://wccftech.com/amd-zen-cpu-core-block/#ixzz3ZyenMTep
 
Last edited:
If you look at the block diagram of the ZEN Core, it only has 6 pipelines??????
I know the bulldozer has 4 but increased the stages from K10's 12-15 to around 20.
I could not find info on how many pipelines the K10 - K10.5 had :(

AMD is making a BIG JUMP. There current CPU is done on a .32nm process and their GPU is on a .28nm process.
From what I have read, AMD is going STRAIGHT to a .16nm process:(
AMD has historically had serious problems when they release a New Uarch + a NEW Process Node.
 
Correct me if I wrong on this guy's.......
You do not release a block diagram picture of your CPU, unless you already have ES chips out being tested.
You have to remember, that even though they have ES chips out, it will still take 6 months or more before a RETAIL chip hits the market. That would put the first release in Nov-Dec 2015 or Jan-Feb 2016. Unfortunately AMD does not have the $$$$ like Intel, when last year Intel in July or August announced that their New CPU would support DDR4. ......BAM..... Intel has a New platform ( X99) for sale at Christmas, that alone helped generate several Million Dollars. It would be very nice if AMD can hit the DECEMBER time line.
What I find funny about Intel's CPU's (Having to buy a CPU+Mobo at every release) is that they went from a socket 1156 > 1155 > 1150 and now their going to a socket 1151 :)
 
Correct me if I wrong on this guy's.......
You do not release a block diagram picture of your CPU, unless you already have ES chips out being tested.
You have to remember, that even though they have ES chips out, it will still take 6 months or more before a RETAIL chip hits the market. That would put the first release in Nov-Dec 2015 or Jan-Feb 2016. Unfortunately AMD does not have the $$$$ like Intel, when last year Intel in July or August announced that their New CPU would support DDR4. ......BAM..... Intel has a New platform ( X99) for sale at Christmas, that alone helped generate several Million Dollars. It would be very nice if AMD can hit the DECEMBER time line.
What I find funny about Intel's CPU's (Having to buy a CPU+Mobo at every release) is that they went from a socket 1156 > 1155 > 1150 and now their going to a socket 1151 :)

Hopefully they will be released as close as Skylake as possible so people don't just jump to it instead of waiting, it definitely looks promising so far.

As for intel socket changes, since 2009 they change socket every tick-tock cycle, and AMD's FM socket has done the same (FM1, FM2, FM2+ and soon the fourth iteration for Carrizo) in a shorter space of time. I think it's a consequence of the CPU having more integrated components than they used to, requiring more frequent socket changes.

Hopefully by then DDR4 is quite a bit more affordable than it is right now too, otherwise this could eat into the performance/price ratio than AMD usually wins on.
 
Back