AMD: Zen will offer 40% faster performance per clock than Carrizo

DogChainX said:
Which is why I have a 386, 486 and Pentium setups with original hardware (sound cards, etc) for my vintage gaming needs. ;-)

:D
Click to expand...

Pentium MMX and Athlon, both rocking 3dfx cards (V1 in the MMX, V5 in the Tbird) here. Glide emulators just don't cut it.
 
Hornet said:
That's true, but there are advantageous to emulators too.

I still have my PS2, but when I hooked it up to my LCD tv, it looked really bad. PS1 games are even worse. LCD screen just doesn't handle those low res well.

Even on the PS3, there's such an option to apply a smooth filter for PS1 games, and it does a great job IMO. Though I'm not going to repurchase my entire PS1 library from PSN, so I'll settle with a PC emulator.
Click to expand...

You know, you can just throw in your disc to any PS3 and it will play original PlayStation games. I've tried this on a 40GB fat and 120GB slim. Not sure about superslims though.

For me, emulators just aren't there yet for anything later than 16 bit machines. pSX or Xebra comes close, though.
 
Mr. Bluntman said:
Pentium MMX and Athlon, both rocking 3dfx cards (V1 in the MMX, V5 in the Tbird) here. Glide emulators just don't cut it.
Click to expand...

I recently discovered a couple of PCI Voodoo3 2000 boards and an AGP Voodoo 5 5500 board in my old boxes from college.

(This is really odd, as the last Voodoo board I remember owning was a 6mb Vodoo1, but I did go through a period where people left extra hardware outside my dorm room door for me to use in building game servers, so that might be where they were from.)

One of the Voodoo 3 boards was used to allow one of my headless servers to post for a while. Then when I rebuilt it, I swapped it out for a 2MB Matrox Millenium board I found, because it used less power. (those early 3d boards would use a good amount of power whether they were rendering or not...)

Maybe I could do something similar with them as well...
 
Zarathustra[H];1041886598 said:
I recently discovered a couple of PCI Voodoo3 2000 boards and an AGP Voodoo 5 5500 board in my old boxes from college.

(This is really odd, as the last Voodoo board I remember owning was a 6mb Vodoo1, but I did go through a period where people left extra hardware outside my dorm room door for me to use in building game servers, so that might be where they were from.)

One of the Voodoo 3 boards was used to allow one of my headless servers to post for a while. Then when I rebuilt it, I swapped it out for a 2MB Matrox Millenium board I found, because it used less power. (those early 3d boards would use a good amount of power whether they were rendering or not...)

Maybe I could do something similar with them as well...
Click to expand...

Hang on to that Voodoo5, they are getting extremely expensive on ebay these days. Last check I ran they're going for over $150, and about the same for 6MB Voodoo1 boards if they even show up.

If you can find one, something like my KT7A (KT133A) and a fast Athlon makes a very good pairing. EDIT: A friendly reminder just in case: whatever board you choose to pair with it, make sure it's a 3.3v capable or universal AGP. :)
 
Last edited:
Zarathustra[H];1041886508 said:
To those who are concerned that AMD has nothing new between now and 2016, I guess my take is as follows.
AMD is SIGNIFICANTLY behind Intel today.
They make their CPU sales from people who are not that interested in high end performance, or who shop out of brand loyalty.
They could spend money and engineering resources on yet another incremental revision on the flawed bulldozer architecture and get what, another 5% boost in performance? They'd still likely be stuck on the same process node.
In the end it comes down to this. Is being significantly behind intel really that much worse than being significantly behind intel +5%?
They will still likely make some dribs and drabs in sales from the same folks who are currently buying, and the truth is that 5% likely wouldn't make much of a difference.
As long as they have the cash on hand to be able to survive until a Zen launch date, it makes very much sense to cut their losses on the current lines, and go full steam ahead with all available resources on what promises to be a major leap, in large part due to architecture redesign, but also in large part because the smaller process nodes will be available then.
I think this is the right decision to make. Spending more money for yet another incremental development of a dead platform only to still be significantly behind their main competitor right now would be a very silly thing to do.
There is some risk here though. Zen ABSOLUTELY MUST live up to expectations, or they might as well liquidate and go home.
They do not have enough cash on hand (or enough assets to sell off) to survive yet another bulldozer...
Click to expand...

What a long list ;) If you look at single thread no contest there. However under Mantle/DX12/Vulkan as long as games are pushing 60K+ batches I would say AMD still got some very good performance.
They gave up on AM3+ ages ago so the people on Bulldozer is prolly not that much last revision of Excavator is on FM2+ sometime early next year (if that is still happening)?

But all the ZEN fear mongering aside did you think that management at AMD wants to fail or better put do you think that Jim Keller wants to fail badly at making the next generation cpu that would not only be bad for AMD .....

I think no one on this forum wants AMD to fail (cept for those few pesky forum members:) ) . The only thing that gave AMD the cold shoulder was the initial production of the 1st generation chip. Maybe AMD can use some luck on that (both Phenom and Bulldozer should not have had this faith).

But even if AMD is successful then it still depends on OEM contracts and last time AMD "won" that did not happen. And it seems that this is also something which AMD has no power forcing this issue, last time was a nightmare.
 
Let's be honest. Mobile spending is generating cash for the PC manufacturers right now compared to desktops. What AMD needs is for notebook manufacturers to make a true high end AMD based notebook. It's fine to have the low end stuff for various markets. But when your first experience with a company wares is a combination of the lowest end processor, HD, memory, screen, etc then you've lost a customer for life.

All they need is a couple of high end models from each manufacturer as a "Holy Grail" to get a good name again.
 
cageymaru said:
Let's be honest. Mobile spending is generating cash for the PC manufacturers right now compared to desktops. What AMD needs is for notebook manufacturers to make a true high end AMD based notebook. It's fine to have the low end stuff for various markets. But when your first experience with a company wares is a combination of the lowest end processor, HD, memory, screen, etc then you've lost a customer for life.

All they need is a couple of high end models from each manufacturer as a "Holy Grail" to get a good name again.
Click to expand...

Intel would shit the bed if an OEM did this - which is exactly why you don't see such laptops at all unless they have "intel inside" written on a badge stuck to the chassis. Release a quality laptop using a competitor's CPU? Whoops! There went your rebates, discount volume pricing, and guaranteed allotment of product.

Intel is called Chipzilla for a reason. They'll go kaiju on you if you do something they don't like.
 
There is this problem for netbooks/laptops is that AMD still using older manufacturing process. If AMD wants to bypass chipzilla it should approach OEM which do tablets and phones and want to branch out to laptops this way there would be no problems. Then again Intel just dumps chips , we saw this with baytrail where OEM could buy one get one free or some other form of compensation.

But the whole PC business is poisoned for AMD the amount of money they would have to spend not only to be competitive but also the deep pockets of Intel means they would have to generate cash flow outside of the x86 sandbox. Previous AMD management didn't see this at all..
 
Well all the people wanting for a high end AMD notebook, but that statement is contradictory in itself as it cannot be high end if it is using an AMD processor.
That is the sad reality. And i don't think there is any way even with zen they can compete on performance with Intels ULV chips, they will probably not have the resources to develop something competitive in this.
 
cageymaru said:
Let's be honest. Mobile spending is generating cash for the PC manufacturers right now compared to desktops. What AMD needs is for notebook manufacturers to make a true high end AMD based notebook. It's fine to have the low end stuff for various markets. But when your first experience with a company wares is a combination of the lowest end processor, HD, memory, screen, etc then you've lost a customer for life.

All they need is a couple of high end models from each manufacturer as a "Holy Grail" to get a good name again.
Click to expand...

I agree with you. I was trying to search for a good Carrizo-based notebook a while back. All I was able to get was a complete piece of garbage with the usual suspects: 15.6" TN panel, 1600MHz RAM, 5400 RPM HDD, cheap case, shitty speakers, etc.

Unbelievable. And it was a $700 dollar laptop.

Now that isn't even the worst thing about it. Carrizo comes in a configurable TDP - 15 and 35W IIRC... and pretty much ALL of them are setting it at 15W static and completely neutering the performance.

My laptop had the FX-8800P. The top of the line Carrizo. It completely sucked with that 15W TDP and I was powerless to change it.

If AMD would simply build an 11.6" laptop with an FX-8800P, 720P IPS panel, 2133 MHz ram, a scalable TDP, a standard SSD or at the bare minimum a hybrid HDD.... and price it in the mainstream range... it would actually be a very nice laptop.

There's no reason to have a 1080P panel in something less than 17", especially in a value segment. I know there are plenty who might disagree, but 720P would be cheaper and be easier on the APU when it comes to graphics processing.

Anyway... it seems AMD basically sells their chips for dirt and lets the OEMs scrap together whatever turd components they can get and barf out some kind of notebook. It's sad really as Carrizo is a good performer with an IGP niche that intel can only compete against with the Iris 6200 or above, in machines costing far more.

I really don't know why AMD keeps bending over for the OEMs or at least start selling their own reference design through a different OEM. There is absolutely zero reason to buy a Carrizo laptop at this juncture.
 
Because AMD doesn't have any leverage? Remember, people like us know who AMD is. The general consumer? Not so much. AMD can't afford to push around the few OEMs willing to bother with AMD at this stage.

AMDs problem is a classic one: If they are "as good" as Intel, they lose. They need to be better.

Oh, and AMD just laid off another 5% of their workforce today.
 
grunwald said:
I agree with you. I was trying to search for a good Carrizo-based notebook a while back. All I was able to get was a complete piece of garbage with the usual suspects: 15.6" TN panel, 1600MHz RAM, 5400 RPM HDD, cheap case, shitty speakers, etc.

Unbelievable. And it was a $700 dollar laptop.

Now that isn't even the worst thing about it. Carrizo comes in a configurable TDP - 15 and 35W IIRC... and pretty much ALL of them are setting it at 15W static and completely neutering the performance.

My laptop had the FX-8800P. The top of the line Carrizo. It completely sucked with that 15W TDP and I was powerless to change it.

If AMD would simply build an 11.6" laptop with an FX-8800P, 720P IPS panel, 2133 MHz ram, a scalable TDP, a standard SSD or at the bare minimum a hybrid HDD.... and price it in the mainstream range... it would actually be a very nice laptop.

There's no reason to have a 1080P panel in something less than 17", especially in a value segment. I know there are plenty who might disagree, but 720P would be cheaper and be easier on the APU when it comes to graphics processing.

Anyway... it seems AMD basically sells their chips for dirt and lets the OEMs scrap together whatever turd components they can get and barf out some kind of notebook. It's sad really as Carrizo is a good performer with an IGP niche that intel can only compete against with the Iris 6200 or above, in machines costing far more.

I really don't know why AMD keeps bending over for the OEMs or at least start selling their own reference design through a different OEM. There is absolutely zero reason to buy a Carrizo laptop at this juncture.
Click to expand...
I'm sad after reading this handicap: https://community.amd.com/thread/185411
A fast ultra mini 720p laptop with AMD PRO A12-8800B could be a good ultrabook.
 
Skylinestar said:
I'm sad after reading this handicap: https://community.amd.com/thread/185411
A fast ultra mini 720p laptop with AMD PRO A12-8800B could be a good ultrabook.
Click to expand...

That's what I'm talking about. If that purchaser was your mom, dad, less tech savvy siblings, etc, then that's a lost customer for life. Why? Because the OEM cut savings by pairing the processor with a crappy heatsink. Thus to the person that purchased the notebook AMD APUs are as slow as molasses.

Hewlett Packard will NOT get the blame for the notebook being slow or the downclocking of the chip because of the poor design of the notebook. The next salesman will simply tell them to purchase Intel next time.
 
why do you want carrizo laptops, ppl in this thread? when it will not perform like an atom at best?(assumption)
you dont put atom in high end laptops!
 
a3venom said:
why do you want carrizo laptops, ppl in this thread? when it will not perform like an atom at best?(assumption)
you dont put atom in high end laptops!
Click to expand...

Carrizo is impressive the 35 Watt version is to be precise you get a lot of performance and very little power usage. It does hardware Hvec 265 and allows gaming with pretty good performance at Dota2 and LoL.

Btw this is the ZEN thread the Carrizo is another one ;).
 
Well the perception issues that AMD brought upon themselves over the years now apply to Zen, Carrizo, GPUs, etc. :)
 
Pieter3dnow said:
Carrizo is impressive the 35 Watt version is to be precise you get a lot of performance and very little power usage. It does hardware Hvec 265 and allows gaming with pretty good performance at Dota2 and LoL.

Btw this is the ZEN thread the Carrizo is another one ;).
Click to expand...

Is there any video on youtube showing this, because it is very interesting.
I saw a video of the 15w version (fx-8800p in a hp) and yes it failed to impress more than intel integrated graphics. (10 - 12 fps on farcry 4 at lowest possible settings)

I have a zenbook with i5-3317u, it plays wow (low ) and dota (med) 1080p at 45 fps on average.
Is that a 15w part in that or 35w? ( idont know that )
So the AMD counterpart at 35w should be way more impressive than that to gain attention shouldn't it?

Which is why i thought there is no point for any OEM to make ultrabooks with an AMD chip, because ultrabooks have to meet a specification of dimensions, screen etc... ( i think ultrabooks need to have mandatory intel CPUs too, but lets leave that part out for this topic )
 
There is a video several months old now showing gaming and streaming on the AMD youtube channel. Sadly no one got some independent stuff for now (Carrizo 35 Watt).

Think someone was mentioning the HP elite books having a TDP limited version of Carrizo as well...
 
Pieter3dnow said:
There is a video several months old now showing gaming and streaming on the AMD youtube channel. Sadly no one got some independent stuff for now (Carrizo 35 Watt).

Think someone was mentioning the HP elite books having a TDP limited version of Carrizo as well...
Click to expand...

But that's AMDs own video, they couldve just made a reference laptop with their chip and give to reviewers.
 
rvborgh said:
some new interesting reading on Zen:

10 pipelines

http://dresdenboy.blogspot.com/2015/10/amds-zen-core-family-17h-to-have-ten.html

i should add from a few days ago:

http://seekingalpha.com/user/19326301/comments

"What I heard from some AMD guys is that they already start testing "their new cpu" in the lab and seeing haswell level performance."
Click to expand...

Fresh rumors! Thanks for the share! Looking at the diagram it seems like zen is a smaller core than haswell (which has 12 or 13 piplelines iirc) so the high core count rumors seem to mesh with this one.
 
Darakian said:
Fresh rumors! Thanks for the share! Looking at the diagram it seems like zen is a smaller core than haswell (which has 12 or 13 piplelines iirc) so the high core count rumors seem to mesh with this one.
Click to expand...

8 cores 16 threads ?
 
I'm a believer. Now that Apple has a Haswell IPC ARM core, I don't think it's all that hard to believe.
 
defaultluser said:
I'm a believer.
Click to expand...

That's what many of us said about Phenom or FX. We all know how those flops turned out.

Still, I hope that AMD does deliver with something as fast as Haswell, as it would put them back in the game and make Intel get off it's ass.
 
Basically its looking like Zen (for desktop) is going to be like an i7-5960x (8 core/16 threads) with a bias towards all out integer performance, and reasonable good enough fp performance (to save power), running at a higher frequency, for a whole lot less $$$.

A winning combo as far as i can see. i want one.
 
Last edited:
Darakian said:
I thought I heard 16 cores somewhere. Either way if the core is smaller than haswell and about as fast as haswell then we're looking at something pretty good.
Click to expand...

The question is how they will do this when they move to SMT.

Will there be dedicated decode units sharting the rest of the core, like with intel's HT?

How much of this is covered by Intel patents, I wonder. I would suspect that nothing HT would be included in the IA32 license...
 
Darakian said:
I thought I heard 16 cores somewhere. Either way if the core is smaller than haswell and about as fast as haswell then we're looking at something pretty good.
Click to expand...

Well 16 cores is nice but will that still be within "the range" of 140 Watt. I mean as soon as we hear specs of AM4 we can at least stop guessing.
 
Darakian said:
I thought I heard 16 cores somewhere. Either way if the core is smaller than haswell and about as fast as haswell then we're looking at something pretty good.
Click to expand...

What would you prefer? 16 cores at 2Ghz or 8 cores at 4ghz? :p

I'd take one for my desktop, and the other for my server :p
 
Zarathustra[H];1041891761 said:
The question is how they will do this when they move to SMT.
Will there be dedicated decode units sharting the rest of the core, like with intel's HT?
How much of this is covered by Intel patents, I wonder. I would suspect that nothing HT would be included in the IA32 license...
Click to expand...

I thought Hyper Threading was an excellent example of how not to do it...
 
Pieter3dnow said:
I thought Hyper Threading was an excellent example of how not to do it...
Click to expand...

Well, it worked better than the bulldozer module :p But that's not saying much.

Is it possible Zen will just be straight up cores, without any kind of logical core setup?
 
Zarathustra[H];1041891761 said:
The question is how they will do this when they move to SMT.

Will there be dedicated decode units sharting the rest of the core, like with intel's HT?
Click to expand...
That would be my guess since that's the whole idea behind SMT

Zarathustra[H];1041891761 said:
How much of this is covered by Intel patents, I wonder. I would suspect that nothing HT would be included in the IA32 license...
Click to expand...

Little to none. Both IBM and Sun/Oracle have implemented SMT without talking to intel.

Pieter3dnow said:
Well 16 cores is nice but will that still be within "the range" of 140 Watt. I mean as soon as we hear specs of AM4 we can at least stop guessing.
Click to expand...
Probably. It's hard to say X cores means Y watts when we know only rumors. It's possible (maybe not probable) that AMD will reverse course and release a 16 core arm chip with a 50 watt TDP :p Also I may have been reading something regarding a server chip with respect to the 16 core thing.

Zarathustra[H said:
]
What would you prefer? 16 cores at 2Ghz or 8 cores at 4ghz?
Click to expand...
The 8 core at 4Ghz. But if there were a 16 core at 3.5Ghz I'd be more tempted by the extra cores.

Speculation is fun :)
 
Zarathustra[H];1041891799 said:
Well, it worked better than the bulldozer module :p But that's not saying much.

Is it possible Zen will just be straight up cores, without any kind of logical core setup?
Click to expand...

The way it comes across is where AMD still gets performance under Mantle/DX12 from those CMT "cores" where SMT just does not.

What you are saying makes sense if they were short on time. from
http://techreport.com/review/28228/amd-zen-chips-headed-to-desktops-servers-in-2016
The Zen core will feature simultaneous multithreading (SMT), or the ability to track and execute multiple threads per core. Although SMT can extend beyond this limit, Zen's version of SMT will stop at two threads per core, like today's big Intel cores. The inclusion of dual threads per core follows a proven template for success in big x86 CPUs, and it also should put AMD on more equal footing with Intel from a marketing standpoint.
Click to expand...

It seems they have it working but the details are still unknown. The way AMD back then described the Bulldozer CMT was that it would function at 80+% of a normal core (from what i recall). How this version of SMT functions will prolly be announced at some point in time where they need some media attention before the release :) .

http://dresdenboy.blogspot.nl/2015/10/amds-zen-core-family-17h-to-have-ten.html
As heard earlier this year, Zen will use SMT and an improved cache subsystem while being designed from scratch with new ideas combined with reusing existing components (to reduce the effort). This might even include already existing and somewhat developed ideas not realized in previous designs. A lot of the new functionality has been filed for patenting. For example there was a mention of checkpointing, which is good for quick reversion of mispredicted branches and other reasons for restarting the pipelines. Some patents suggest, that Zen might use some slightly modified Excavator branch prediction.
Click to expand...

It seems that there is still somethings dresdenboy wants to blog about regarding Zen :).
 
Last edited:
i run a 48 core Opty as my home PC and i can tell you that just about nothing is even remotely capable of utilizing it. 16 threads is plenty for desktop :)

i wish AMD came out with an 8 core/16 thread APU with shared HBM2 and a Fiji on it :) i'd pay $1200 for it.

Pieter3dnow said:
32 cores or bust :)
Click to expand...
 
rvborgh said:
i run a 48 core Opty as my home PC and i can tell you that just about nothing is even remotely capable of utilizing it. 16 threads is plenty for desktop :)
Click to expand...

Please tell me you are at least virtualizing or something on that bad boy. Or maybe encoding/rendering?
 
Darakian said:
So I've been doing multi-threaded coding recently and it's a lot of fun to have these huge core counts.... so... 128 cores or bust! :D
Click to expand...

Once you go tons of cores you never want to go back - so glad to have convinced TPTB years back to get 24 core Opterons for our SQL Server and .NET Application Servers.
 
My server has 12 cores (24 logical) and my desktop has 6 cores (12 logical) and that is sufficient for me :p

(In fact, it is probably way more than I need. I don't think I've ever seen the server with more than 35% CPU load, except for during stability testing :p )
 
Pieter3dnow said:
I thought Hyper Threading was an excellent example of how not to do it...
Click to expand...

And intel Know it... the return and re-implementation of OoOE (Out-of-order Execution Technology) to being a successor of Hyper threading I think its a good example and a good route for intel in cannonlake and I think that's the main reason why they are planing to increase the core count of mainstream chips by not using HT but relegating the work to OoOE which is supposed to offer +30% multi-thread performance at the same energy cost than HT but also allowing higher FP performance.. so 6 logic cores with 12 OoOE threads will be great for mainstream market.

Intel had a great time of refine OoOE with those tiny Silvermont and goldmont Atom chips with haswell IPC in 4 cores/ 4 threads and just 5W TDP mobile chips or 8c/8t server chips. which work great so I guess all that experience and refining added to cannonlake will be just at least Interesting..
 
You must log in or register to reply here.
Back
Top