AMD CPUs, SoC Rumors and Speculations Temp. thread 2

Page 42 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Status
Not open for further replies.


To me i thought the CPU would actually be more competitive in gaming at the time not be below a 1100T. BD Was such a terrible "new" product i'll never forget that disappointment, and it was my fault for buying the board.
 


AM3+ has one additional pin and does have a different layout electrically and in what it supports. It still supports AM3 CPUs but was designed specifically for BD/PD chips.
 


IB's performance edge is somewhere around 30-35% or so. In addition, you make the assumption clocks won't be lower in Zen, which they almost certainly will. So if they get a 20% IPC gain, but also reduce clocks 10%, you still have a good 20% performance deficit versus IB, nevermind Haswell/Skylake.

That's AMDs problem: They pretty much HAVE to deliver 40% per-core IPC gains to be competitive with Intel at this point.
 


Maybe its just me.

EDIT: I did not own a 1100T I owned a phenom 2 x4 945 overclocked to 3.6.

I loved BD. I was amazed by PD. 8150 was faster than the 1100T when max overclocks for both were considered when running applications I run every day. on the side it also ran games within a margin of difference I could not discern from.

I had my 8150 pumping 4.5 on water for 2 years before I upgraded to PD and proceeded to clock that core to 4.8. both cpu's were faster in every encoding job I had to run for the money than their intel counterparts and STILL to this day my 8350 runs them faster than an identical priced i5.

Yes they are not as good for games, but do they still make my games hit 60 fps? Do I even have a monitor that can refresh faster than 60 hz?

you see why BD is a marketing failure and a money success? I pay next to nothing for every inch of performance I want and amazing fun overclocking to the max.

EVEN IF zen is a miserable 30% IPC improvement to PD (once again they claim 50% IPC improvement) AND the clocks are 2.8 stock 3.5 turbo I will be happy to buy it if it has 16 threads and is priced lower than intel's $1000 cpu.

NOBODY should expect this zen cpu to be as fast as intel at games. there isn't a game today that isn't totally restricted by the cpu single thread performance and we know zen will not match intel skylake or kaby lake (which will be out at that time). let the hype be hype, but make no mistake. zen, like PD, will be amazing at something and will be priced accordingly.

Last note please read the fine print people. im tired or repeating this. zen will have 40% IPC improvement over excavator cores. EXCAVATOR. excavator cores are faster than the 8350's piledriver cores. the zen claim is higher than 40% IPC gain over the 8350.
to assume amd will hit even a meager 60% of its goal IPC improvement, is still a 30% improvement over the 8350, performance similar to ivy bridge.
 
Well not saying you didn't love BD but for many it was a major disappointment actually even Amd themselves had to make an article addressing the performance issues. I for one ended up disliking BD over me having a 1100t OC 3.9ghz and heck even when I upgraded to a 8350 I still saw my GPU usage drop below 90% quite often meaning the CPU was bottlenecking my system.

People forget to actually monitor your usage and your frame rate now or even when I had my I3 my GPU usage was always better like 90%+ there was times when my usage dropped to like 70% with my 8350
 


its proven that an 8350 overclocked to 4.8 is faster(or equal to) in single core perf than an 1100T at 3.9.

we all were disappointed by the lack of improvement on this front and only matching performance by raising clocks.
Point is it DID match performance once clocks were considered!

then add to it the fact the 8350 had immense improvements in multithread rendering performance and we result in an overall improvement over the 1100T.
was it what the hype claimed? nope. should we have foreseen that a 8 core CMT design would not perform as well as intels 8 thread SMT design when amd is sharing a FPU for 2 cores? yes.

we now know what SMT is capable of thanks to intels years of use, the block diagram of zen, a claim of IPC and insider insight to clockspeeds that the 8 core 16 thread will be capable of reaching (@juanrga) For once I think we can make a reasonable assumption of performance and the hype is not like it has been in the past of the will kill intel. amd got realistic after the failure to meet the hype with BD specifically with games.

90% of pc builders are gamers. you look on this forum and all you see is I want to upgrade for gaming. BD was bad at gaming and so the PC building community dismissed it as a total failure, as it didn't do what they wanted it to.

if amd's Zen was a quad core quad thread that were 4 huge super fast 8 ghz cores and able to beat intel in all single core perf and cream them at gaming, the community would rave about how great it was, despite its lack of ability to do the encoding jobs I need it to do. I would go to intel or simply keep my 8350 for a separate encoding machine and get zen for gaming. and I don't really want 2 computers :/

BD was a disappointment to you because you expected it to be good at gaming.

BD was the best cpu ive ever owned because I expected it to be good at encoding.
 


Working in IT I can say that I rarely see these anymore. With laptops being thinner and lighter and tablets becoming more powerful they are picked more than thin clients.

I would prefer to see AMD pull near 50% of the server market share. At least then they would have the money to properly do R&D.
 
Thin clients are kind of a dead market but whats replacing them are pretty much the same thing but more local processing and storage. I think what ever chip AMD is using in that space could get design wins for enterprise nucs too, maybe whoever makes those clients will use AMD for different products. Its good that they have some niches they can sell in.
 

Sadly i got a bad one to so i could only get mine to 4.3ghz(8350 never owned a 8150) with all the cores and it resulted in a lost plus it ran hotter even, overall i saw a slight increase in FPS and in benchmarks also it did beat my friends lower clocked haswell in chess was easily slower in dolphin something i use often. Not to sure about Zen just based on Amd's last 10 years but i kind of expect ivy based IPC something tells me Amd didn't have everything Keller wanted to really push things foward. I'm with juan i think people will be disappointed on the price.

 


Zen's price will be terrible because they will have many many people who are die hard AMD fans buying up the first batch if the performance is even remotely ok. I can still remember bulldozer launched at $250 and was sold out for the first month or so even with the terrible reviews.
 


tbh $250 wasn't horrible for the 8150. should have been around $230 starting price imo and then it droped to $200 really quick, and that was perfect.

@esrever your right. amd fans have been waiting for the return to high end computing and they will pay a $50 premium to even a $100 premium just to have zen first.

me personally? hard to say. I am a huge amd fan, but I didn't buy BD or PD until they droped to under $200 each. I also make more money now sooo might snag one early for the premium anyway. :wahoo:

sorry to hear you got a bad one jdwii! ive had two and they both hit 4.8 easy :/
 
It's true for everyone. For the first one to two months on any new CPU/GPU or chipset it's very expensive. Companies launch well before the production is scaled up and retailers don't downprice them. Look at Skylake, say- weeks before they got close to Haswell in cost, and all the CPUs could be released. Fury X too- problems with availability. So expect to have to wait one or two months after launch for a decent system at a reasonable price.
 
New user been reading this forum for quite some time and the old forum.
Anyways, I was wondering with the new AM4 socket or AM4+ either or, what is the deal with chipsets, northbridges, southbridges do we have any info on that? Are they planning to move away from that?
 


the apus, most of them will be SoCs. they will have NB and SB (aka UNB and FCH respectively) on die. amd may choose to use external FCH for desktop usage (disable the FCH) and fully enabled SoC for mobile pcs.
not yet sure about the cpus, although both the cpus and the apus will have plenty of room for FCH. depends on whether amd can save money from making a bigger chip with GloFail or aim for higher yields with two chips.
 
Rough estimation of Zen costs

Die size

1) About 50% of Kaveri die is 4 SR-core CPU.

2) Samsung/Glofo 14nm is about 2x more dense than 28nm.

3) Zen core ~ 2 SR cores

4) Thus 8 Zen-core CPU (without L3) would occupy about 245mm² on 14nm.

5) Add L3 and the die size will be close to 300mm².

Transistor cost

6) Not all transistors are the same. 14FF is more expensive than 28PL; first because of the jump to 3D and, second, because the smaller size requires extra techniques as multi-patterning.

7) Thus ~300mm² die on 14FF will be about so expensive as ~600mm² on 28PL [I am using cost data from another guy].

Design/Validation

8) All of above simply considers number of transistors and transistor costs. Add the R&D cost from designing Zen micro-architecture from scratch. Add the cost from validating such design (validation costs increase nonlinearly with design complexity).

Extras

9) AMD has a huge debt and other economic problems and needs huge margins. This is why AMD will not focus on cheap products anymore. In Lisa Su terms: AMD is no longer the cheap company.

Conclusion: As I stated many time ago, Zen will be an expensive product. Anyone expecting octo-core Haswell-like performance at 8350 prices or i5 prices will be disappointed.
 


Thank you for the reply appreciate it.
I am curios to see what they do I don't really see them moving away from their system now, Is there any performance boost of having the chips integrated on the chip or?

 

it depends on whether the integration affects performance. for example, amd adopted integrated pcie controller in their apus but never showed competitive gaming performance as intel's cpus as it's cpu was average at best.
otoh, cost is a big advantage of integration. 2 chips strategy will require separate packaging, fabbing, validating, designing and other stuff. integration saves cost, reduces area required on the mobo, requires only one thermal dissipation system (even if the fch is usually left bare), advantageous for hsa (broadly speaking). soc will help offset some (or a lot) of the expense of moving to 14nm FF node.

and if amd has worked out the cpu, memory issues, we might see the upcoming chips really taking advantage of integration (the igpu already does that, btw).
 


Honestly, it depends on the leakage, power consumption, and other factors for the die overall.

Essentially, when Intel first moved everything on die (haswell, iirc), they actually showed performance loss in terms of overclocking. Such to the point that a well overclocked Ivy Bridge tended to be about 10% faster than a well overclocked haswell due to the difference in achievable clock speeds.

Additionally, adding additional complexity on die increases the power draw to the die for something that was previously not drawing power on die, but on the board. So, depending on how AMD adapts to that new feature of the uarch, cooling may be extensively more demanding to keep the chips cool, especially if the thermal windows for the uarch to perform well are low like BD/PD/SR/EX...if the thermal envelopes are higher like Intel, then you may well see a scenario where they have more overclocking headroom.

It all just depends, really...however...we can get a lot better idea once the early silicon starts coming around.
 


I would imagine you will see it for roughly the same as a higher end i7 quad though...which is very fair, especially considering.
 


I believe octo-core Zen CPU will be priced between six-core and octo-core Broadwell CPU, with AM4 boards being significantly cheaper than Broadwell-E boards.
 


Sounds fair enough, I saw an article last week stating AMD have confirmed 4, 6 and 8core Zen sku's (sorry haven't got link to hand). I don't know if the lower core count parts are smaller dies or harvested from the 8 core?

Either way, I'd expect quad core zen (8 threads) to be priced between i5 and i7, with 6 and 8 core parts being higher. That makes a lot of sense provided the performance is there.
 


Only if the performance is the same. If it is lower it wont be. If it is higher it will be priced higher.

AMD needs to make money and playing the cheap company gets them no where.
 


+1 Juanrga.

octo zen will be $600+ launch imo
 


You make the assumption price is based off performance. The top chips will be selling north of $500 at launch, regardless of performance. They'll correct over time as supply/demand, which is performance driven, takes hold. But initial prices will be jacked up to catch the early adopter.
 
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom