New Zen information, AM3+ info, APU presentation, and video card information

Shintai said:
The only thing that matters for a product is how it sells. And that determines if its a fail or not. All construction cores so far have been a huge flop in the sales department. Why would Zen change that when the metrics compared haven't changed? Zen from the looks of it, is even worse than the first FX parts vs the competition.
Click to expand...

Rofl. From the looks off it you seem to hate anything amd. Yep Zen is failure and worse than fx series without seeing single official actual benchmark relating to the CPU. Looks like you got your daily dose of amd hate in!
 
NKD said:
Rofl. From the looks off it you seem to hate anything amd. Yep Zen is failure and worse than fx series without seeing single official actual benchmark relating to the CPU. Looks like you got your daily dose of amd hate in!
Click to expand...

So instead of relate to the substance, you draw the hate card when you disagree. Well done.
 
KazeoHin said:
The best damage control I saw from JF-AMD was the quote where he tried to say that the bulldozer IPC was higher than thuban because it was clocked higher.

People here were like "Thats BS" and JF-AMD was like "IPC means 'instructions per core' lol"
Click to expand...

He started out OK but I swear someone must have take over his account around the BD launch. Roy maybe? :p
 
AMD marketing exec specifically on the business side that went around to lots of enthusiast forums prior to BD launch and shilled for AMD, swearing all the leaks and ES benchmarks and all that weren't accurate and to wait for release, release came, all leaks and ES benchmarks were accurate, BD was garbage, he basically lied to multiple enthusiast communities, disappeared from all online forums shortly after launch, disappeared from AMD completely within a year. Since he was fairly high up the AMD chain AMD lost a lot of credibility due to him and hasn't had an official rep on most of the forums he was on since then.
 
However I think this is a different situation : somebody on the inside got too enthusiastic about the new hardware an wanted to prove a point, that the parent company never intended to. Which points to the fact that the product does not live to expectations and has missed some targets. If it would be otherwise, why hide it? IMHO I think AMD is betting on SMP right now, rather then improved IPC.
 
cageymaru said:
WCCFTECH wants your clicks. I didn't see the new info; maybe you will. 95w CPU is new maybe?
AMD Zen 8 Core 16 Thread & 4 Core 8 Thread CPUs Leaked - Impressive Performance At 95W & 65W
Click to expand...

Yeah it is rather everything they could find then you try and cram it on 1 page and there you have it

it’s important to remember that AMD’s latest Orochi dies feature Piledriver cores rather than Excavator. Excavator cores are roughly 15% faster per clock than Piledriver This in turn puts Zen at a lead in excess of 60% vs Piledriver in terms of performance per clock. Doubling the performance of the FX 8350 puts Zen in direct competition with Intel’s eight core Extreme Edition i7 5960X.
Click to expand...

There posting so much unsubstantiated shit it is not funny any more .....
 
Pieter3dnow said:
Yeah it is rather everything they could find then you try and cram it on 1 page and there you have it



There posting so much unsubstantiated shit it is not funny any more .....
Click to expand...
Yeah and that is how it starts. I said that 25% over piledriver @ same clocks would be good for a reasonable expectation and maybe considered a small win. Gotta keep in mind that this is a new design for AMD with nothing prior to build upon. If it can show some strength over previous showings with some room to grow then it is all good.
 
MrCaffeineX said:
I noticed this in my rounds this morning: AMD Zen Microarchitecture: Dual Schedulers, Micro-Op Cache and Memory Hierarchy Revealed

Does this architectural overview help set realistic expectations or is wild speculation still the norm?
Click to expand...

Not a ton of new detail; the cache information is interesting, but unless AMD has fixed it's very poor cache performance, slapping more of it on die won't accomplish much. Also worth noting is how the SMT implementation is pretty much standard, rather then the BD's implementation and 20% performance hit.
 
gamerk2 said:
Not a ton of new detail; the cache information is interesting, but unless AMD has fixed it's very poor cache performance, slapping more of it on die won't accomplish much. Also worth noting is how the SMT implementation is pretty much standard, rather then the BD's implementation and 20% performance hit.
Click to expand...

From what I've read, this is an entirely different architecture from the ground up. It's not related to Bulldozer at all.

The cache setup looks more flexible based on the increased communication links and increased size vs the BD architecture.

They did do a side-by-side comparison with a Broadwell-E (both clockspeeds locked at 3ghz) running Blender with the Zen chip finishing first by a few seconds. If that's an indication, the worry is still clockspeeds.
 
Magic Hate Ball said:
From what I've read, this is an entirely different architecture from the ground up. It's not related to Bulldozer at all.

The cache setup looks more flexible based on the increased communication links and increased size vs the BD architecture.

They did do a side-by-side comparison with a Broadwell-E (both clockspeeds locked at 3ghz) running Blender with the Zen chip finishing first by a few seconds. If that's an indication, the worry is still clockspeeds.
Click to expand...


Agreed. But they supposedly are holding clock speeds close to the vest. Given what I heard early about this exceeding expectations and outperforming them, My guess is that this sucker is going to be set aggressively.
 
Zion Halcyon said:
Agreed. But they supposedly are holding clock speeds close to the vest. Given what I heard early about this exceeding expectations and outperforming them, My guess is that this sucker is going to be set aggressively.
Click to expand...

This is the company that released a 220 watt CPU in the form of the FX-9590 factory clocked to 4.7ghz. I have a feeling they're not going to pull any punches for the enthusiast grade CPUs.
 
Magic Hate Ball said:
They did do a side-by-side comparison with a Broadwell-E (both clockspeeds locked at 3ghz) running Blender with the Zen chip finishing first by a few seconds. If that's an indication, the worry is still clockspeeds.
Click to expand...

They didn't tell HW configurations, they didn't say if they used custom compilers etc for the Blender. Or if it was actually a GPU load. The runtime, lack of standard bench and the almost perfect time match raises a lot of questions.

A trip back in history shows this:
AMD FX 8150 Looks Core i7-980X and Core i7 2600K in the Eye: AMD Benchmarks
 
Shintai said:
They didn't tell HW configurations, they didn't say if they used custom compilers etc for the Blender. Or if it was actually a GPU load. The runtime, lack of standard bench and the almost perfect time match raises a lot of questions.

A trip back in history shows this:
AMD FX 8150 Looks Core i7-980X and Core i7 2600K in the Eye: AMD Benchmarks
Click to expand...

True, but I guess we have new leadership now. I think Lisa Su has been pretty careful not to exaggerate. We'll know when the reviewers have samples in hand in a few month's time I guess.
 
Magic Hate Ball said:
This is the company that released a 220 watt CPU in the form of the FX-9590 factory clocked to 4.7ghz. I have a feeling they're not going to pull any punches for the enthusiast grade CPUs.
Click to expand...

Isn't there still a concern that Zen simply won't be able to handle the kind of voltage/current that made those FX CPUs generate so much heat due to the new manufacturing process and the physics related to FinFET??
 
Magic Hate Ball said:
True, but I guess we have new leadership now. I think Lisa Su has been pretty careful not to exaggerate. We'll know when the reviewers have samples in hand in a few month's time I guess.
Click to expand...

Have she? Kaveri, Carrizo, Fiji, Polaris etc?

I think we first see reviews at release when sales have begun. And we may have to wait until early next year for that.
 
Shintai said:
Have she? Kaveri, Carrizo, Fiji, Polaris etc?

I think we first see reviews at release when sales have begun. And we may have to wait until early next year for that.
Click to expand...

Yeah, maybe her more recent work overshadowed what I had forgotten from the past. But if you ignore the community over-hype, Polaris is a good product, and a solid improvement over the previous generation.
 
The problem is that we still don't know how far Zen will clock and if the 4 core version will differ in respect to the 8 core. There have been suggestions by certain people that the 4 core 8 threads would clock a good deal higher.

Where PCPer lists clocks above 3 ghz let it not be 3.1ghz because that would certainly be a silly thing to mention ....
 
Last edited:
Guru3D
AMD Shares ZEN info and talks about based 32 core 64 thread product

  • AM4 8 cores with 95W TDP (Summit Ridge)
  • AM4 4 cores with 65W TDP (Bristol Ridge)
  • SP3 24 cores with 150W TDP
  • SP3 32 cores with 180W TDP (Naples)
The "Summit Ridge" Zen family will feature a unified AM4 socket with its GPU-equipped "Bristol Ridge" APU counterparts, and feature DDR4 support and a an expected 95W TDP. Newer roadmaps don't confirm the TDP for desktop products, they suggest a range for low-power mobile products with up to two Zen cores from 5 to 15W and 15 to 35W for performance-oriented mobile products with up to four Zen cores. Though unconfirmed we expect each Zen core will have four integer units, two address generation units and four floating point units, and the decoder can decode four instructions per clock cycle. L1 data cache size is 32 KiB and L2 cache size 512 KiB per core.
  • DDR4 Memory
  • PCIe Gen 3
  • USB 3.1 Gen2 10Gbps
  • NVMe
  • SATA Express

Additional “Zen” architectural features will be detailed next week in a presentation at Hot Chips 28.
 
Shintai said:
Have she? Kaveri, Carrizo, Fiji, Polaris etc?

I think we first see reviews at release when sales have begun. And we may have to wait until early next year for that.
Click to expand...

Polaris is handled by the RTG.

I can't say I heard a lot of hype over Carrizo or Kaveri.
 
I just read the AT article about Zen this morning and a few things caught my eye:

  • Zen now includes a micro-op cache (which BD did not have). This should help instruction throughput for instructions that have previously been decoded (as in a loop) and should help reduce power consumption a little bit by not having to re-decode the same instructions again.
  • The L1 data cache is write-back instead of write-through as on BD. The problem with a write-through cache is that for any data that changed in the L1 cache, it has to update this new value in the L2 cache/main memory immediately. This adds latency when data is written to cache/memory. A write-back L1 cache only has to write the data to the cache and once that is done can say the instruction is complete. This data is then written back to memory when data is evicted from the cache or when another core needs this data. This should be faster for programs where only one core/thread is performing operations on a set of data and this data can fit within the L1 cache.
  • Caches are larger than on BD and have a higher associativity and are similar to Intel processors. The higher the associativity the more likely that data will be able to be placed in the cache. The L2 cache is actually twice as large per core as the Intel L2 caches on Broadwell and Skylake.

Hopefully this means that we will see some good performance from the processor. Of course we will have to see what the final clock rates are, but I'm being skeptically optimistic. I'm sure that Blender workload was chosen as one of the better performing benchmarks but clock for clock Zen looks like it will be competitive with Intel.
 
Schmave said:
I just read the AT article about Zen this morning and a few things caught my eye:
  • The L1 data cache is write-back instead of write-through as on BD. The problem with a write-through cache is that for any data that changed in the L1 cache, it has to update this new value in the L2 cache/main memory immediately. This adds latency when data is written to cache/memory.
Click to expand...
No, it does not necessarily add latency to use an L1 WT cache instead of a L1 WB cache. The write-through operation to L2/mem can proceed asynchronously to the CPU and L1 cache. What using WT does do, however, is increase the number of writes to the L2/mem, which is a power and bandwidth issue, not a latency issue. It also forces your L2 cache to be inclusive, which can make for slightly slower hit rates for the combined L1/L2 subsystem.
 
pcgeekesq said:
No, it does not necessarily add latency to use an L1 WT cache instead of a L1 WB cache. The write-through operation to L2/mem can proceed asynchronously to the CPU and L1 cache. What using WT does do, however, is increase the number of writes to the L2/mem, which is a power and bandwidth issue, not a latency issue. It also forces your L2 cache to be inclusive, which can make for slightly slower hit rates for the combined L1/L2 subsystem.
Click to expand...

What you wrote is correct that the write through operation would not necessarily add latency, but depending on the implementation it could. I don't honestly know how Bulldozer does this so I can't say one way or the other. Your other points are good, especially in that write-back reduces the number of writes to L2 which, as far as performance goes, saves memory bandwidth to the L2 cache.
 
Schmave said:
What you wrote is correct that the write through operation would not necessarily add latency, but depending on the implementation it could. I don't honestly know how Bulldozer does this so I can't say one way or the other. Your other points are good, especially in that write-back reduces the number of writes to L2 which, as far as performance goes, saves memory bandwidth to the L2 cache.
Click to expand...
Well, I did used to architect 32-bit RISC microprocessors for a living, after all. :)
 
Deathroned said:
man you surely do hate AMD they used the same system, reviewers were there (PC perspective, Hot Hardware etc),
Click to expand...

that doesn't mean anything. AMD is well known as cheaters and liars in each announcement and presentation they make.. so anything out of the direct hand of AMD can be considered fake or very close to it.. =) no one have to hate AMD to know that...
 
Hardware pron. Extremely closeup shot of hot naked Motherboard with GPU and CPU straddling it at the 2:55 mark of this video. May get heart palpitations after viewing this! You have been warned!

2 way Server CPU gang bang on a AM4 motherboard at the 3:30 mark.

 
Araxie said:
that doesn't mean anything. AMD is well known as cheaters and liars in each announcement and presentation they make.. so anything out of the direct hand of AMD can be considered fake or very close to it.. =) no one have to hate AMD to know that...
Click to expand...
A bit much... I don't think they cheat, maybe lie but that is a bit of a stretch as it is more akin to painting themselves in the best light with known strong points. They generally post marketing material which given how marketing works you never post negatives about your product but show it in the best light possible. Their GPU benches pre-release always are shady as they tend to use only certain settings to make them look favorable. It sucks but that is how it is done. Also some of what they state is without clarification like they state a 4K card because they believe as some sites around the world that 30fps is the goal, although enthusiast require 60fps avg or higher for the extreme nuts.
 
Araxie said:
that doesn't mean anything. AMD is well known as cheaters and liars in each announcement and presentation they make.. so anything out of the direct hand of AMD can be considered fake or very close to it.. =) no one have to hate AMD to know that...
Click to expand...

Riddle me this why would they gain from lying to PC perspective, Hot Hardware etc they were there they saw it with their own eyes, are you afraid of, considering Intel and nvidia shenanigans you should not be throwing accusations. of I was AMD I would just leave the PC market let Intel and Nvidia jack prices we will then see how much you enjoy shitting and crapping on and to make you feel better about your crappy life.
 
Shintai said:
Let me hear the details then. Please list it.
Click to expand...
Do I look like google, the same threads there are articles posted if you bothered to read to grasp information instead of looking for something to bash you would have known, its the same system difference is the cpu .
 
Deathroned said:
Do I look like google, the same threads there are articles posted if you bothered to read to grasp information instead of looking for something to bash you would have known, its the same system difference is the cpu .
Click to expand...
I have to applaud, AMD have outdone themselves here, getting Zen ES working on LGA2011-3 with X99 mobos.
Deathroned said:
Riddle me this why would they gain from lying to PC perspective, Hot Hardware etc they were there they saw it with their own eyes, are you afraid of, considering Intel and nvidia shenanigans you should not be throwing accusations. of I was AMD I would just leave the PC market let Intel and Nvidia jack prices we will then see how much you enjoy shitting and crapping on and to make you feel better about your crappy life.
Click to expand...
Well, what did they gain from lying about Polaris power efficiency? Yet they did.
 
lolfail9001 said:
I have to applaud, AMD have outdone themselves here, getting Zen ES working on LGA2011-3 with X99 mobos.

Well, what did they gain from lying about Polaris power efficiency? Yet they did.
Click to expand...
Actually no, they didn't outright lie about efficiency. One they didn't giver the direct comparison, with which we can compare. Nor did they give the scenario. I think and others are just now coming to the same conclusion that the situation is under Frame Limiting allowing the power gating to work. That is the one thing it seems that no one has actually tried.

As I stated above, AMD doesn't outright lie, they tend to limit the situation to a point where what they state is correct. Every company does this one way or another when it comes to marketing their product.
 
JustReason said:
Actually no, they didn't outright lie about efficiency. One they didn't giver the direct comparison, with which we can compare. Nor did they give the scenario. I think and others are just now coming to the same conclusion that the situation is under Frame Limiting allowing the power gating to work. That is the one thing it seems that no one has actually tried.

As I stated above, AMD doesn't outright lie, they tend to limit the situation to a point where what they state is correct. Every company does this one way or another when it comes to marketing their product.
Click to expand...
Actually they gave both for their 2.8x perf/watt claim. They claimed stock 270x was eating 180W in their tests and that is well known to be not the case.
 
You must log in or register to reply here.
Back
Top