KitGuru KitGuru.net – Tech News | Hardware News | Hardware Reviews | IOS | Mobile | Gaming | Graphics Cards
It looks like we can expect an AMD APU based on the Zen architecture in the future, though rather interestingly, it will include stacked High-Bandwidth Memory (HBM) with bandwidth speeds of 128GB/s along with a larger onboard GPU for better gaming performance.
This information comes to light from a paper(via: wccftech), which was co-authored by AMD graphics engineer, Mike Mantor. The paper shows that this Zen-based APU would make use of an upgraded version of the memory technology found in Carrizo, which started off as an improved design over the memory interconnects found on the current generation consoles.
On top of that, the Zen APU is shown to use stacked HBM with 128GB/s of bandwidth. Aside from that, we don't know much more at this time, though earlier reports have suggested that Zen-based APUs won't arrive until 2017, following the release of desktop Zen CPUs in the fourth quarter of 2016.
As far as the use of HBM goes, we do know that some of next year's GPUs from both Nvidia (Pascal) and AMD (Arctic Islands) will make use of second generation High-Bandwidth Memory though it looks like this Zen-based APU will be using first generation HBM technology instead.
KitGuru Says: 2016 is going to be a big year for AMD, especially with its new Zen architecture coming out towards the end of the year. Are any of you waiting for AMD to release Zen at the moment?
HURRYup
Q4 of 2016 = I’m switching to Intel. AMD sounds like they will release some incredible technology but the ball has been dropped. I was going to wait until the end of Q1, but Q4 isn’t happening. That sucks.
Intel for me. The CPUs are just better in performance per core and per watt than AMD’s.
Zen hasn’t even been released so you do not know of they will be more efficient or more powerful than their Intel counterparts. Wait for the benchmarks and reviews to make a logical decision.
AMD for me- I ain’t buying from a criminal outfit.
I see. You were waiting in a cave until now refusing to touch a computer, huh ? All of a sudden this article spurs you into action, chiefly consisting of disparaging AMD.
so zen will somehow use HBM? not looking good for nvidia since they’re going to get even less new RAM to play with on their new gpu.
Ill upgrade from my FX 6300 sometime after Zen is released. Hopefully be able to afford near higher end that time around 🙂
If this happens in the stated date and if zen is successful this thing will be a top notch product “if they don’t over blown the price!” can U just imagine? Next APU architecture is supporting DDR4 memory we should see a substantial performance gain on those and then this!
This hopefully will be for the Mobile/Laptop segment… APUs with HBM in the Desktop segment wont benifit as much from this as Laptop/Mobile will…
Imagine a Laptop with an APU consisting of 4 Zen cores (8 threads), a decent gaming GPU, all coupled via HSA 2.0 and HBM!!! Sounds pretty damn good to me!
AMD: Shut up and take my money!
We can get a rough idea. Zen is supposed to be 40% higher IPC than Excavator. Working on a core by core basis 4 core 8 thread Zen should be nearly as fast as the i5’s in single thread. I doubt it will be faster but it will be close, say 10-15%. Even if it is 20% behind it wont matter. Having their own form of HTing will close the gap in multithreaded apps but MT will likely not be any faster than Bulldozer derivatives as MT was their major strong point. Even now the i5’s just beat Piledriver. 8 core 16 thread should be twice as fast as the 4 core 8 thread but only if apps can use it. 8core 16 thread will be a beast of a chips as the i7’s are not twice as fast as the i5’s. Can’t say about power consumption though. Again this is all just speculation. They could have a single benchmark that Zen get 40% IPC boost over excavator, like they tried to pull off when bulldozer and piledriver was out. Piledriver traded blows with Sandybridge chips ranging from i3’s to i7’s depending on core utilisation and provided there were no FP calcs to be made. Who knows, I do like the fact AMD are being all hush hush about it though
I use an FX-6300 with 8GB of RAM and an GeForce GT 730 for my Media PC (connected to my TV)…
When Zen comes out I have a slick case and a 1000W PSU ready for it, hopefully it’s pretty schlick cause I’ll be building a gaming machine centered around it!
My game pc runs on AMD hardware, 🙂 hehe
Only since Bulldozer. Phenom II chips were pretty good in terms of power consumption compared to Intel’s Core 2 chips and even 1st gen i series for comparable performance. They just came to the party a lil late with them. And again Performance per core is becoming less relevant, finally. We still see games where Piledriver is keeping up with i5’s. Which is fine. Zen is a different type of core design and is focusing on high throughput on each core with SMT making up the ground in multithreaded apps. MT wont be any faster than Bulldozer based chips as MT was their only strong point. They will be good. But you are entitled to your opinion and are right if based on the bulldozer series chips. I just do think Zen will be vastly different from Bulldozer. All I can do is speculate 🙂
Must be in a desperate need to upgrade eh? If you have anything older than say Sandybridge then I can understand. But if you have Sandybridge or newer there is no point upgrading anyway. I have a Sandybridge i7 2600 and even Skylake is not worth an $800aus+ upgrade to an i5 or nearly $1000aus for the new i7’s. They give like 20% IPC upgrade over Sandybridge after 4-5 years? Zen will be replacing my home server though, which is a FX 8350 with 16GB RAM unless it really is the ducks nuts then it may replace my Sandy i7 in the gaming rig. If you are really willing to spend all that dough on upgrades that give you f* *k all then go ahead. You’d be better off spending that money on a Fury or a Titan. You’d get much better improvements. Zen is a catch up chip. Skylake is to show they are doing something new. Both aren’t going to be light years ahead of what we already have.
I love AMD hardware just as much as the next guy, and I am stoked for Zen and APUs based on Zen with HBM… It’s about time a company created some solutions to rival Intel’s solutions!
The High End Desktop parts are tantalizing enough, the thought of 4 core/8 thread desktop APUs with HBM on Mobile Parts is enough to make your head explode!
2016/2017 will be an interesting year to say the least!
Zen CPUs don’t have to rely on HBM as much as you suggest… A Zen based APU using HBM can use this HBM for the GPU alone, and use DDR4 for the CPU…
Look at the latest offerings of GPUs from AMD, the HBM is used for the GPU alone, so an APU utilizing HBM doesn’t necessarily have to utilize the HBM for the CPU and GPU, it can utilize the HBM for the GPU exclusively and use the system DDR4 for the CPU.
This would be no different than an AM3+ system using an FX CPU with DDR3 memory and an R9 390X with HBM.
This is hbm1. gpus will use hbm2.
I will buy it. Apu with hbm can bring a revolution in apu performance.
i know, but aren’t they all made by the same producers? isn’t one of the big issues right now the fact that there’s a limit on production? i’d bet that increased HBM production would necessitate a decrease in HBM2 production, since there are a limited number of plants that make them. add to that AMD having first refusal on all of it.
maybe i’m wrong, but it just seems like it could further reduce amounts available for nvidia. either way it will be interesting to watch.
A few days ago new McLaren F1 subsequent after earning 18,512$,,,this was my previous month’s paycheck ,and-a little over, 17k$ Last month ..3-5 h/r of work a day with extra open doors & weekly paychecks. it’s realy the easiest work I have ever Do. I Joined This 7 months ago and now making over 87$, p/h.Learn More right Here
dm.
➤➤
➤➤➤ http://GlobalSuperEmploymentVacanciesReportsMoney/GetPaid/98$hourly…❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦
Hbm1 is being produced by hk hynix only. hbm2 by Samsung too though it’ll enter volume production in Q2 at both the places.
System memory wouldn’t be an issue as much as it was on old APU’s if it has HBM on the APU for the GPU to use……..and that was as confusing to type as it probably is to read with all the acronyms…..
Plz we need it to buy! We need laptops with it. Sell licences on third partys vliw 4 *5 Bulldozer x86 stars k10.
“Performance per core is becoming less relevant,” – wrong
It matter MOST. High IPC = slower clocks (less power), = less cores (smaller and cheaper CPU), = great performance in single threaded apps (still > 90% use time)
Nah more cores seem to be way to go. You can easily disable cores or put them in a low powerstate when not needed. Most companies dont get huge boosts in IPC’s but doubling the core count will nearly double the performance. Once placed on smaller nodes more cores can become very efficient. Look at ARM, no one is going higher IPCs they are going high core count to get the major improvements and best performance per watt. Also where did you pull that number from? ST still 90%used? Apps have been mt for years. Yeah single thread is still important. Like i meantioned in the text you so kindly noted “performance per core is becoming less relevant” not irrelevant. The reason AMD can get such a high ipc gain is focusing on dozers weaknesses which is of course better that making 16 core excavator chips. Generally IPC gains are expensive and hard to get massive improvements where as increasing core counts provides massive gains provided you have the die space to do it. It is easier to see that high core counts are becoming the norm and will become more important. You will eventually reach a wall in IPC gains once you find the best way of doing things. In short yes ipcs are important but they are becoming less and less relevant. Once AMD are competing on IPC you will core counts jump.
I have been an AMD user since my first PC. I was a little disillusioned with the company lately but Zen would be something I’d look into.
Exactly. The architecture is based on a 14 nanometer process – which should bring down the temperatures, the wattage, and bring up the performance – as they can fit more transistors on the CPU itself.
THIS IS IT! I currently have an A10-7850k using only the onboard iGPU paired with 2400mhz Dual-channel ram, Overclocked the iGPU from 720Mhz to 1028mhz, play’s everything I throw at 🙂
How much more with ZEN? Soooo excited!
Not upgrading UNTIL Zen APU
4 Zen cores??? its not enough because intel will release core i7 with 10 core in 2017/2018
i’m still running Phenom IIx3 720BE with HD7750 GDDR5
Desktop APU’s with HBM would benefit as well if the APU is capable of using the HBM memory as an L3 or L4 cache type of storage. The APU could complete monstrous operations without ever talking to system RAM.
This is essentially what Intel was trying (and failing) to do with Iris Pro.
Boohoo, you do realize that MOAR CORES isn’t the answer to more performance anymore, right?
Anyway, AMD has a 16 core HPC version of Zen core slated for release as well. Hold on to your pants bro.
Hell yeah man! People don’t know ’bout my overclocked iGPU.
What the hell possessed you to pair a GT 730 with an FX-6300? You monster.
For now.
Ok, so what you’re saying is Zen is a pack of horses. 1 Horse is capable of running 40% faster than a Bulldozer horse. Somehow, you expect 4 Zen Horses together, to still run the same speed as a Bulldozer horse?
Alright buddy, let me know how that brain scaling works out for you.
Intel is releasing a 10 core LAPTOP CPU? REALLY?
Unless you’re space-constrained, if you REALLY want performance, ditch that compromised on-chip GPU to free up some TDP headroom and plug in a real, discrete GPU.
you know that 10 cores are absolutely useless for gaming and the iGPU will be much, much faster than those cores at productivity tasks such as rendering, editing, etc, etc.
it is, just not for gaming, for rendering and server workloads they are much more effective and efficient than stronger cores, but guess what the iGPUs are much faster than these cores.
….Not what I was saying at all. Single thread Zen will destroy Bulldozer like the i5’s do. Multithread Zen won’t be much faster than Piledriver again like the i5’s. Piledriver can keep up with i5’s but only in Multithread. Single thread there is not even any competition. SMT or Hyperthreading give about 30% improvements in most things. Now Excavator is not that much faster than Piledriver in raw IPC performance. What you do have it better communication between cores. Regardless you can see for the most part Zen will sit behind a i5 6600K or a i5 4790K in single thread maybe a bit in front and maybe with all 8 threads stressed.
As far as your horse theory goes??? Just to clarify 4 core chips will not beat an i5 on their own, they will need SMT to get there. Having 4 Core 8 Thread chips do not make it double the speed of a 4 Core 4 Thread chip. SMT or Hyper threading will only improve it by roughly 30% like it does in Intel Chips. Bulldozer/Piledriver 8 Core 8 Thread Chips have an advantage when all 8 cores are stressed as this architecture was made ground up to be great at Multi thread. Zen is being made ground up to be good at Single thread Hence them including SMT to increase MT performance. Just want to make sure you realise it is the 8 core Bulldozer/Piledriver chips I am talking about (not 4 core chips).
Anyhow longer than I wanted but it seems I have to explain twice. Zen 4c8t roughly going by AMD’s predictions = 6600K in ST and MT. FX8350 is already roughly the same speed as the 6600k in MT (ST is way way way slower.) So yeah In Multithread (which I mentioned earlier the trusty old 8Core Bulldozer/Piledriver will be as fast as a 4c8t Zen.
There was a time when dual and quads were useless for gaming too. Once ipc improvement have gainded all they can gain core count increases will be key.
Different markets don’t overlap or compete, they don’t even share potential customers. This quad-core w/HT Zen APU will only compete with whatever 4c8t CPU+IGP in LGA1151 exists by then, whether KabyLake or Cannonlake. In fact, nothing in mainstream ‘competes’ with anything from Intel’s HEDT platform.
If anything, only the Zen 6- and 8-core w/HT is going after the higher end i7 processors. Will it out perform, no, but that isn’t the point of competition in the business sense. The idea is to steal away potential customers by filling gaps, so I wish AMD the best.
I have to agree, I have the same cpu that you’ve got, think I got it in 2011 and wouldn’t upgrade the cpu at the moment, but I’m planning doing a big upgrade late 2016 when the new microns on the cpu’s and gpu’s come out and I have to admit, with DirectX 12/Vulcan likely to be used a lot more going forward, it makes me wonder if I need to upgrade my cpu at all with the new api’s lowering cpu usage a lot, but I think I will just so I have a new motherboard with all the new stuff, ddr4 and what not, thankfully ddr4 has droped a lot when I checked them recently so looking good for late this year.
Yeah I know what you mean. You will get some performance improvements with the new ones especially I think running big cards or Dual cards. DDR4 I think with games running on Frostbite3 which seems to love high speed RAM. I think I am just going to upgrade my GPU or try find a good R9 290 to run along side my one if I can find one cheap enough. I know this is pretty much betting on the impossible but I really hope AMD have something amazing up their sleeve. I mean not just 8 core Zen chips but 8 core Zen chips that compete with Intel extreme chips. But I think I would have more chance of surviving a trip to the sun with nothing but a scuba tank and thermal undies than AMD bringing out a killer like that.
i3 sales mean developers have zero incentive to cut out owners of dual core 4 thread chips by making engines that truly take advantage of so many threads/cores.
Hawaii is pretty impressive considering how well it has held up, performance-wise.
Bulldozer is not a laptop CPU. It’s a 32nm high power node design.
Nvidia’s big numbers. Surely 730 is better than 580, right? (Yes, both companies are guilty of this nonsense.)
A few days ago new McLaren F1 subsequent after earning 18,512$,,,this was my previous month’s paycheck ,and-a little over, 17k$ Last month ..3-5 h/r of work a day ..with extra open doors & weekly paychecks.. it’s realy the easiest work I have ever Do.. I Joined This 7 months ago and now making over 87$, p/h.Learn More right Here
4kkw………
➤➤
➤➤➤ http://GlobalSuperEmploymentVacanciesReportsCloud/GetPaid/98$hourly…❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦
HBM iGPU should be pretty nice and the HBM, if it can be used as an L4 cache when working with a discreet GPU, should be better.
The time to buy a new expensive GPU is when both companies are on 14/16nm.
I hope they don’t use the previous gen GPU in this one.
That, it is. They are apparently going to be pushing for some ridiculous amount of transistors out of these next gen cards. I am looking forward to it. I wish CPU’s move along at the same pace as GPU’s
You are definitely right there. That is why we still have the option to lower our detail levels in games so they can still cope. I think a great engine is an engine that can adapt to the hardware from tiny little rigs to the big beasts that only a handful can actually afford to buy. It is not in any companies interests to cut anyone out. But no one will develop for what isn’t there. IE 10 core chips.
16 cores, pfft I will get excited when I see how 1 core performs.
Prob why Bulldozer derivatives overclocked like a mofo too. Don’t see that happening with Zen as much.
It’s a MEDIA CENTER PC… I don’t need to game on it, just get decent performance.
I was looking for best bang for buck, Tomshardware suggested it, and it works fine…
Upgrading my gpu is the only thing I’ve done since my last big upgrade, got a 280x about a year ago, this is one of my slowest upgrade cycles I’ve ever had, on the one hand I shouldn’t complain as I’m saving money but it would be nice to see some progress, the real problems been the last gen console cycle lasted too long, 8 years which in the tech industry is too long, and AMD not competing with Intel on the cpu front, with any luck, the new micron and Zen chips from AMD can change that.
All in all, this year is looking like a interesting one on the PC with the new microns to DirectX 12/Vulcan, to VR, 4k, Windows 10 and the PC growing at a healthy rate, at least in gaming anyway, which in all honesty is the main reason for me to upgrade the PC.
I would imagine that Zen APU will feature an upgraded architecture – probably based on what is rumored to be ‘Polaris’/Arctic Islands – on the integrated gpu, more gpu cores (that much has been confirmed), and far more bandwidth than current offerings.
Add HBM to the mix, and it is very possible that this integrated gpu will surpass most of the other offerings (from Intel and possibly Nvidia too).
Keep in mind that AMD’s igp’s performance seems to be bandwidth limited for the most part, so an HBM addition would provide a much needed boost in bandwidth, and subsequently performance (in addition to architectural improvements, addition of more cores, etc.).
As for how much more powerful AMD ZEN APU with HBM will be…
Won’t know conclusively until we see it, but we might have some indications based on this:
http://www.geek.com/games/amd-plans-to-offer-5x-performance-for-playstation-5-and-next-xbox-1639548/
Consoles are of course a closed system for which games were specifically optimized, but the hardware would still need to be powerful enough to drive 4K and Virtual Reality (Which I think is something AMD is focusing on with Zen).
I suspect that AMD might be focusing on providing maxed out gaming at 1080p with Zen APU’s that have HBM… 4K might be an option, but I have no clue if the hardware will be capable of pulling it off (perhaps the Polaris architecture will be specifically optimized for DX12, 4K and VR).
Software optimization will likely play a huge role in this, but if current news about Zen are accurate, then the hardware will likely be VERY powerful.
Why not i got an trinity bulldozer tier 2
It is crazy how little has changed over this many years especially in the mid to high cpu market.
You should already be excited. There is also a 32 core chip rumored and we already now that it’s going to perform at least 40% better than Excavator cores.
The APU’s shown on all of the websites that cover AMD leaked information show 16 and 32 core APU’s. Imgine THAT!
First of all, why would you even include a 4c/8t version of Zen. Zen is going to be a minimum of 8c/16t, but there will also be 16c/32t and 32c/64t versions. With 40% better IPC and much better power efficiency, I predict that Intel will have a lot of competition to deal with, especially at the pricepoints that AMD will be at. Also, Zen will be much faster than Piledriver in Multithreading. Thats the whole reason why a Fx8350 is the same speed as an i5 even in multithreading, it’s because the i5 does’t have multithreading at all. With the updated cores AND a much better method of hyperthreading, 4c/8t version of Zen would be faster than an i5, give it an 8c/16t version and forget about it, I don’t even think current i7’s could keep up. Intel will need to release those 10c/20t versions and at a much lower pricepoint if they want to keep a hold on market share.
This makes me moist.
They aren’t. They’ll be using Greenland (Polaris), at the least, in their upcoming Zen APUs.
I know, consoles are partly to blame for giving us little incentive to upgrade, Microsoft being slow on the windows front but that seems to be changing and AMD for not competing enough but can’t blame them, they only have so many resources and with them battling both Nvidia and Intel and both of them having more resources then AMD, it’s amazing they compete at all. Good news is the new micron should help them out a lot.
1. i5 is a multithread chip it has four cores and is able to handle 4 threads. FX8350 has 8 threads available but is only as fast as the i5’s because of how well Piledriver works in Multithreaded apps. So no a 4 core 8 thread will not be that much faster than Piledriver when all 8 threads are stressed. The 8c16 will destroy it though.
2. 8c16t chips will not likely be in the same price range of the i3’s and i5’s losing out on much of the market. 4c8t chips will be available to compete with i3 and the i5’s. Quad cores are still the most supported chips on the market. AMD will not release a chip that is supposed to compete with the i7’s at the same price point as the i5/i3s. If they have a product they can charge for they do.
3. Do the math, with 40% Instructions Per Clock increase, core for core or clock for clock, Zen will still leave Intel with a lead in performance.
4. 8 core 16 thread chips will not be amazing for gaming. The massive IPC improvements will help greatly but .New API’s are lowering CPU usage not only that, 8 core chips don’t get a lot of support in this area.
5. Intel will not have to release 10 core chips at better price points. 16c32t, 32c64t will likely be only for the server market not retail. So 16c/32c Zen chips will not be competing with 4c or 8c i7’s.
6. Zen is a catch up chip. It is a node jump and a redesign so that AMD can be competitive with Intel core for core. While 40% IPC is significant and will make AMD relevant again it is only closing the already massive gap in performance between the 2 brands. Something I do think is that Zen will be great and I will be building a couple of systems on it.
1) Quad core i5’s don’t have hyperthreading, that’s a fact, they are 4c/4t, straitforward. Also, Piledriver’s method of hyperthreading (CMT) was poor, which is why an i5 can outperform an FX8350 even in multithreaded applications, which is why a 4c/8t Zen will outperform the 8c/8t Piledriver. SMT>CMT.
2) No one knows what the pricing will be until AMD officially releases is, but based on History, it will depend on the performance of Zen. If Zen is as good as we’re beaing led to believe, then perhaps the first FX Zen chip will be around $300, but if we look right now, AMD FX 8c CPU’s are selling in the i3 price range, not even hitting i5 prices, but when they were released they were selling for much more, but still less than the i7 desktop CPU’s.
3) The 40% IPC improvement is just one of the known performance improvements and that is over Excavator cores, which are supposed to be 15% better than Steamroller, but since there was never an Excavator desktop CPU made, we really can’t know how well it will perform, but even with just the 40% in IPC improvement, I see Zen trading blows with an i5 depending on the benchmark (and that’s in benchmarks that only utilize up to 4 threads. Anything beyond that and Zen will destroy an i5, like future games for instance).
4) You seem to have this part completely backwards. The more cores you have, the better it will be for gaming (it wasn’t like this so much in the past, but today and into the future it will be). The new API’s are increasing CPU usage across all cores, not decreasing it. In the past only 1 CPU core was ever able to communicate with the GPU at any given time, but now with DX12, all cores will be able to communicate with the GPU, hence all companies, especially AMD going for more cores and a lot of them at that. 8-core CPU’s are what all console games are based on, and future DX12 games will be based on at least a minmum of 4 with 8+ cores being taken into consideration as DX12 games are created and with every console game that’s ported over to PC.
5) Intel doesn’t “have” to do anything, but what they ARE doing is releasing greater core count desktop enthusiast chips. This most recent generation saw the first 8c/16t Extreme CPU. Next gen will feature the first 10c/20t, but those cost $1000 or more. Once Zen is released, they are going to have a tough time selling those unless they change their pricing. Based on available information, there will be a 16c/32t that will be a desktop enthusiast part as well. The 32c/64t, from what information is out there, will be the next Opteron successor.
6) Glad you like it. I do too. Will build my next system based on AM4 in preparation for an eventual Zen CPU. Will most likely need to go with an Excavator APU in the meantime, which will be interesting to bench at the least.
A few days ago new McLaren F1 subsequent after earning 18,512$,,,this was my previous month’s paycheck ,and-a little over, 17k$ Last month ..3-5 h/r of work a day ..with extra open doors & weekly paychecks.. it’s realy the easiest work I have ever Do.. I Joined This 7 months ago and now making over 87$, p/h.Learn More right Here
4gxp……..
➤➤
➤➤➤ http://GlobalSuperEmploymentVacanciesReportsFirst/GetPaid/98$hourly…❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.
1. Just to clarify multithread does not mean hyper threading. I did state that i5’s are 4c4t chips meaning I am aware they don’t have hyperthreading. Hope that didn’t sound bitchy. Don’t mean too. I actually thought Piledriver was great at multithread. Especially when it came out it traded blows with the Sandy i7’s but also with i3’s. I figured it’s incredibly poor single thread performance was its demise. Because the chips is still relatively powerful when compared to i5’s provided all chips are used. Maybe it is the way I see it could be wrong.
2. You are right that we wont know. But going on AMD’s pricing (at least in my area) they tend to go high then lower it down. I remember their last competitive chip was the 6 core Phenoms which were a beast of a chip but these came out at nearly the i7 prices at the time which was a bit overpriced. This price settled down a bit months after it came out.
3. I really hope they are being sneaky turds on this and just giving us snippets of what to expect from Zen. Comparing Carrizo against Kaveri but doesn’t show much more than 5-10% on most benchmarks other than GPU based ones. There is truth in what you say about not knowing all the performance details, but still Looking at 3DMark CPU score (I know, I know, Synthetic benches) of the Steamroller based Kaveri A10 7850K scored 4,748 VS the i5 6600K’s 7,413. Add 10% onto Kaveri’s score for the Excavator score and you get 5222. Add 40% on top of that and you get 7311. I mentioned before I it will fall just short of the i5’s core for core but add AMD’s SMT on top and the 4c8t will be faster in multithreaded apps I’d imagine anywhere between 10-30% depending on how well AMD SMT actually is. Now doubling AMD’s core generally give you double the performance. The difference between Intels 6600 and 6700 isn’t double making things very interesting 😀
4. Numbering points makes things easy 🙂
I may not have explained myself very well here. What I really meant was the new API’s are lowering CPU usage like you said. I didn’t mean they will use less cores. DX12 is definitely going to be interesting but I fear for many Dev’s it may be too hard to develop and optimise properly. Yes I understand that DX12 gives them more control over the available resources but I think we wont get the most out of it for a while. But I am not a game designer so I don’t really know. I feel many will be looking towards engines like UE4, Cryengine and Uniengine (I think its called) as they have a lot of the work done for them. I could be wrong though.
5. Agreed 8c16t for $1500 (in Aus at least) is a little bit BS considering the 4790k beats it in single core performance in some areas by 35%
http://cpuboss.com/cpus/Intel-Core-i7-5960X-vs-Intel-Core-i7-4790K
I Really hope that 16c chip ain’t too expensive because that might just be the sucker I put in my home server. Not that I use my 8350 enough.
6.Yeah don’t take my pessimistic outlook on Zen to seriously. Maybe I am not getting my hopes up. But I am Excited. I am hoping for it to replace my i7 2600 eventually and I know it will replace my FX 8350 in my home server even just on power draw. Carrizo chips are amazing on power compared to the old Piledriver chips so I do have High hopes for Zen in this regard. Zen will be good. Just how good we will have to wait. Like I said they are being pretty hush hush about it. Anyhow, good chat 🙂
Provided it is reasonably priced. Over here is Australia the A10 7870k and decent speed RAM to go along with it is almost as expensive as an i3 with a basic but faster GFX card. Making the APU a hard sell. But still they look amazing.
And yet Ashes of the Singularity scales very well across cores especially in DX12 mode. There’s a big difference between “working” with 2C/4T and getting the optimal experience with four physical cores or more.
DX12 will make better use of existing CPU’s but also provide the scalability for better experiences with more cores. Basically a “win-win” if you will.
Plus, there are LOTS of games that launched with very high system requirements so historically your argument doesn’t hold water.
Yeah, trust me, as one of the biggest AMD fans that exist, I had really high hopes for their CMT based architecture, but they just marketed it wrong and rushed it out to production before the architecture was completely mature. Now, with Excavator, we see what Bulldozer should have been from the get-go, but it’s a bit too late at this point. If we had Excavator desktop cores during the time of Sandy Bridge, we’d have a really good thing, even Steamroller cores would have been great, but that’s where the architecture needed to start from, not end with. I still love the idea of CMT, which is why it’ll be fun to actually run an Excavator APU once it’s out for a short time before I swap it out for the Zen based APU. What I always wished for was SMT on top of CMT, which would have made the most out of the FX chips we were offered. Imagine, 4m/8c/16t FX chips ftw! Oh well. I’ve been waiting years and years to finally upgrade my Phenom II [email protected], and unfortunatley, Vischera wasn’t enough to pursuade me, but I’m super happy that AMD has finally made some excellent decisions and will bring out a worthy successor to the Phenom II. All those Constructacon CPU designs were simply meant to pave the way for Zendastator.
I think i know what you mean. I went the jump to Piledriver in the hope that there would be better support for it later on. Which there was. If every game supported the fx 8350s like say bf4 amd would have been in a much better position. That said it is amds fault as well for bring out a different core design. I think i will do the same and get an excavator chip and run it along the 8350 to see what the difference is. Are they bringing out an 8 core excavator?
Unfortunately they are not making an 8c version. It’s still gonna be made on the 28nm node and with that, there simply isn’t enough room for more than 2 modules. And rather than design something new with Excavator cores for 14nm, they’re just gonna go strait with Zen at that point, so unfortunately we gotta deal with 2 module BS, but at least we can purchase the board, ram, and have a working system to hold us over. I’m building a mini-ITX system, or preferably a pico-ITX, so APU’s are it for me anyway, but man, how much more glorious a 16c/32t Zen APU will be, paired with an iGPU that will rival the performance of the Fury, its going to be one helluvan upgrade over the Excavaotor APU. No more discrete GPU’s or cases that need to be big enough to fit one anymore for me. Tiny desk ornament with Hurculean processing strength will be powering all my future gaming needs.
Agreed, that’s what APU’s have needed all this time. Now we’ll get to see just how Viable APU’s can really be. I think it will make for a huge performance gain, at least bigger than most will suspect.
Present generation apus are not that interesting because of bulldozer architecture. Next generation shouldn’t have that problem theoretically. We’ll know how zen performs by the end of this year or early next year. Ram bandwidth limitation should be solved by HBM. It’ll be priced according to its performance or slightly higher. The best thing is there will be more competition and we’ll be benefit from it.
and Benefit we shall! 😀
Unfortunately there will not be an 8c version of Excavator. It is being made on the 28nm node still, and since it’s only going to come in an APU design, that only leaves room for 2 modules, so we’re going to be stuck with that until Zen comes out, but at least we’ll be able to purchase the AM4 boards, RAM, and have an APU to provide a working system all while we wait to swap it out for the ZEN APU/CPU once it’s released.
Yeah that makes sense. No point putting Excavator on smaller nodes to fit 4 modules on it. I would be happy with an APU with PS4 power. But I guess we will be exceeding that in no time. I am wondering what the new Nintendo will be like once it is released and if it using AMD’s hardware too 😀
That would be sweet just having to upgrade a Nuc like PC every few years. One site said AMD were looking at putting high speed non-volatile memory on chip. Imagine 256GB of High speed Non-volatile HBM on chip that can be used as storage/System/GPU RAM/cache. Would be a true System on Chip. I’d imagine that is along way off though. If they could pull something like that off. Might be a while off yet.
Actually, they are putting HBM on their Zen APU, but just how much I’m not sure. I’ll have to look at the details again to see if they make any mention of the exact amount, but the fact that they are actually putting HBM on the chip itself will lead to new levels of APU performance, considering that APU performance was limited by having to access slow (in comparison) system memory. Also, they could stay at 28nm and fit 4 modules into a single die, but it would be a CPU only, just like the FX8350, which I always wanted AMD to do with each iteration of the architecture, but they aren’t going to do it. I wanted an FX-Steamroller and an FX-Excavator in 8c versions, because they would be better than having to stick with the old FX-Bulldozer and Piledriver, but who knows, I wasn’t in charge of AMD, so unfortunately us enthusiasts had no great upgrade options without going to the dark side.
If that DX12 feature can take advantage of multiple GPU’s then I can see the advantage of having big iGPU’s but if it takes more than it gives then I think AMD should just bring out a small iGPU version with bigger/more CPU cores. Maybe it will be too many SKU’s but have a High performance CPU with light weight iGPU and have a high powered iGPU with a smaller CPU. Might get confusing though. I guess it’s not important to have huge amount of CPU cores ATM (not that 16 cores with 32 threads are a small amount). I like your Idea of 4m8c16t chips and it would be a great way to use those extra integer cores that don’t get used but would that even be possible? I’m sure it was just wishful thinking but, I was under the impression that AMD had to get Microsoft to hand tasks to bulldozer chips like it would in an SMT setup to maximise the cores properly?
Apparently they are putting 16GB of HBM memory on the APU. It wont have to pull from system memory, that’s gonna be sweet. And as for DX12 using multiple GPUs, yes, in fact, as strange as it is, for the first time, it allows for mixing any GPUs you want one AMD and one NVidia, no problem. And what is really weird is that, so far, the absolute best performing setup is with an AMD FuryX in slot 1 with a 980Ti in slot 2, specifically in that order. It performs worse the other way around, perhaps having just a teeny bit closer proximity to the HBM is the reason, not sure though.
And in terms of 4m/8c/16t, yeah, it’s possible, but there’s no point anymore. Why split the cores when you can just add a crap ton more whole cores with SMT instead. It was an idea (and a cool and unique one at that) when four cores were the max they could fit on a die and they wanted to improve multi-threaded performance, but with 14nm, we’re gonna have up to 32 cores, so no need to split them anymore. Now, since more than enough cores can fit on the die, they just want to maximize the potential of each core by utilizing as much of the unused cycles as possible. That is why an i7 isn’t twice as fast as an i5, if your core is at 66% duty cycle, then the virtual core will only be a maximum of 34% of a whole core, thus 50% better performance rather than 100%, but at least we can rest easy knowing that all of the CPU we paid for will be utilized and that 34% wont just go to waste. So, even with CMT, you could have one module with 2 cores, one at 66% the other at 80% and then apply SMT to create two virtual cores using up the remaining 34% and 20% of the modules cores, completely maximizing the usage efficiency of the chip. Oh well, wishful thinking of the past, lawlz.
What are you even talking about? I own an AMD 6-core FX that is still on the 32nm fabrication process. AMD should have reached 22nm 2 years ago. Fine, I waited and held my breath. Now instead of Q1 2016, it’s suddenly Q4. Not Q2 or Q3, but 4. Come on. Too late.
and yet a core i7 is showing no more than 2-3% FPS gains in games than a core i5. Video games are GPU intensive and are getting more and more CPU agnostic.
This is the apu that’s gonna power the NINTENDO NX CONSOLE THAT’S GONNA BE ANNOUNCED AT THE E3 CONVENTION LATER THIS SUMMER