AMD RyZen CPU Architecture for 2017

Well I havn't seen board with backside M.2.

In most board the only available space is between PCE ports and when ur gaming the VGA will literally be spitting hot air to the drive and we know how susceptible these drives are about the heat. Also you can have only 1 or 2(MSI now show a 4 stack drives array but it only make the thermal trothing problem bigger). and im sure the drives makes the boards more expensive than just normal ports since they require more engineering time to not just make the traces but move components around them.
 
Well I havn't seen board with backside M.2.

In most board the only available space is between PCE ports and when ur gaming the VGA will literally be spitting hot air to the drive and we know how susceptible these drives are about the heat. Also you can have only 1 or 2(MSI now show a 4 stack drives array but it only make the thermal trothing problem bigger). and im sure the drives makes the boards more expensive than just normal ports since they require more engineering time to not just make the traces but move components around them.

I built an mITX system for my fiance in late 2015 that had the M2 slot on the back of the board. It was a Gigabyte Z170 board I believe.
 
Last edited:
I built an mITX system for my fiance in late 2015 that had the M2 slot on the back of the board. It was a Gigabyte Z170 board I believe.

I haven't seem one myself although I dont look for small boards which has even less space so I think it was a good solution. But then again it even may require to get out the board of the case to get access to the drive which is basically my point(its harder for the user and its extremely limited in the number of drives and usually good implementations are just for high end boards with high prices. We need a standard with the ease of use of "classical" drive where you can swap or "scale" drive easily. Also with a good placement for cooling because right now M2 are in the worst possible parts of the board with most of them getting hit by (at least 60c hot air?) from the VGA.

Btw I was thinking. could it be possible to enable SMT in just some cores? like a 4c cpu with 2 cores with SMT to get 4c6t?
 
The cooling of M2 drives AFAIK is not an issue outside stress tests & benchmarks.

On the flip-side of placing these drives close to possibly hot gpus, the gpus themselves will get better cooling, IMO. Consider that the fans at the front of the case can now get their air to (or out of) the motherboard directly, without cables , drives and cages in the way. This is very important
 
The cooling of M2 drives AFAIK is not an issue outside stress tests & benchmarks.

On the flip-side of placing these drives close to possibly hot gpus, the gpus themselves will get better cooling, IMO. Consider that the fans at the front of the case can now get their air to (or out of) the motherboard directly, without cables , drives and cages in the way. This is very important

without cables , drives and cages? you mean replacing ur drive set up with a single(or dual) M2? can be possible although for around 300 for 1TB M.2 to 50 for an hdd(or 6 times the price(although It makes no sense to me since M.2 have less materials and can be fully robotic manufacture...)) its still a large gap. and yes ideally you would want to have the faster storage possible but in most cases an HDD is just fine for storage things like music, pics and videos.

But still we have the scalability problem. With HDDs if you need more space you just bough another one with higher capacity and add that to ur system, but with M.2 you have very limited number of bays and even worse you get a very limited number of lines to connect them. So if you want one single drive it would be Ok but if you want to get more you can start getting into problems.

The M.2 drives remember be that Asus board with the RAM solded to the PCB. A good idea with great performance but no future proof and limiting. Maybe we can make cables? for M.2 drives and get many more PCIe lanes and then overcome all the cons of the technology but I can't predict the future.
 
I've personally experienced the sudden failure of a 3TB Seagate drive after just 18-24 months of usage. I don't want a repeat of that. If one needs a lot of storage certainly one also cares about data safety (they do, just didn't realize it yet). So IMO storage should be handled by an external drive, the cloud or a dedicated NAS.

The price argument is of course still valid. But at least now you can throw money at the problem, which I intend to .. This was not possible before. And prices will go down, too
 
I have bad experience with seagate myself, never going to buy another drive from them again. But have good experience with WD and hitachi. And my argument was not that M2 are not useful but that it is not the long term solution to the problem.
 
Would a quad core zen with 8 logical cores be able to replace 8 physical Jaguar cores in every situation that could arise in the Xbox library? Its tiny too, Zens gonna be like 3x it's size at least.
 
im no expert but zen cores and jaguar core are even in the same league. you can answer the question simply by seeing the difference in single threat between jaguar and excavator and add 40% to get to zen that will tell how much faster the zen core is compare to jaguar(even if it will be just a raw estimate) then take core numbers into account with SMT core = 25%(since delv will use it to its full potential).
 
I have bad experience with seagate myself, never going to buy another drive from them again. But have good experience with WD and hitachi. And my argument was not that M2 are not useful but that it is not the long term solution to the problem.

I have rather good experience with Seagate, specifically OLD drives that are still good as cheap OS drives (linux) or supplementary storage. Current 7200.14 ones seem to work. I still had a 160GB fail hard but that's what a hard drive does. They're rather quiet for the purpose. e.g. I would have liked to use a Quantum drive as they look good, but the rotation noise is insane.
Of course, the 7200.11 Seagate drives are the really bad ones, I guess the good thing is I doubt there are survivors left (the worst is to use one as a backup drive : your backup drive will fail)

WD did a stupid thing : "Green" 3.5" drive buried in the smallest USB enclosure they could manufacture for it, with no air vents whatsoever. It's a drive that came with its own coffin. So if you try to do the good thing by using external drives, check that they didn't do something really stupid like that.
 
I didn't mean to take a stab at Seagate. Rather, at hard disks in general. Now that that SSDs have matured mostly, I 'feel' (it is just a feeling, real data which I'm unaware of may contradict it) hard disks are less reliable than their NAND alternatives. Further, I just hate the noise they make. Even for thouse that have rather silent operation. I'm still annoyed by their startup sound after being idle. And they take up space..

I've made myself a NAS with HGST drives now which at least is in another room. Still not happy yet, but my problems are OS-related now. Couldn't get freeNAS to power down the drives when not used, as it probably optimizes them for response time.

That said, my main Vishera based desktop still has a WD Blue drive I use for storing games. It's 7 years old and I'm fine with its noise too.
 
Last edited:
Would a quad core zen with 8 logical cores be able to replace 8 physical Jaguar cores in every situation that could arise in the Xbox library? Its tiny too, Zens gonna be like 3x it's size at least.

If they ran at 3.2GHz, 4 Ryzen cores even without hyperthreading could probably do anything that 8 Jaguar cores at 1.6GHz could, and I'm talking worst-case situation where a certain instruction can't be run any faster-per-cycle in Zen than in Jaguar (if such a thing even exists).

However, I'd say if AMD ever disables Hyperthreading, then it'll be for product differentiation in the modular PC market. They most probably won't produce or design Zen cores without Hyperthreading (just like Intel never did since SandyBridge on non-Atom cores).
That said, if Scorpio ever comes with Zen cores then it'll most definitely have hyperthreading.
 
im no expert but zen cores and jaguar core are even in the same league. you can answer the question simply by seeing the difference in single threat between jaguar and excavator and add 40% to get to zen that will tell how much faster the zen core is compare to jaguar(even if it will be just a raw estimate) then take core numbers into account with SMT core = 25%(since delv will use it to its full potential).
Assuming Excavator has 30% faster IPC than Jaguar and Zen has 40% over Excavator, we get 82% faster IPC total. Let's be conservative and estimate SMT adds 20%. Total performance per core would thus be 2.18x, assuming identical clock rate. Now add a modest clock rate increase for Zen: 1.6 GHz -> 2.5 GHz (+56%). We are now at 3.4x per core. 4 core Zen should thus beat 8 core Jaguar handily. Even at very conservative clock rate.

4 core Zen would also have equal amount of hardware threads (virtual cores) as Jaguar, meaning no extra OS context switches. Core locked threads also map perfectly. Zen core also has 2x higher SIMD math (SSE/AVX) throughput compared to Jaguar, so no problems there either. Of course CPU timing wouldn't be exactly the same, so there's always a risk that some poorly programmed game with a bug (race condition) misbehaves. But most games nowadays are cross platform and/or devs have PC debug builds -> this kind of bugs get likely caught during development.
 
Last edited:
I didn't mean to take a stab at Seagate. Rather, at hard disks in general. Now that that SSDs have matured mostly, I 'feel' (it is just a feeling, real data which I'm unaware of may contradict it) hard disks are less reliable than their NAND alternatives.

I have good experience with my 100MB Micropolis SCSI drive. It has (indeed I still have it, working!, 22 years old) capacitators the size of peas (5 1/2 size, half height, server disk with 1KG steel frame). One day I wanted to water my plant which stood by my SCSI-"array" on the window-sill, and I accidently watered my drive. I didn't realized before I heard the water drops boiling between the contacts. I immediately turned off the power and was scared to death. I have a positive attitude in general, so I went to the disco, letting it dry up. And so it was, it worked perfectly fine after I came home in the morning.
I loved the drive, because literally everything on the board was gigantuan, I always thought I can go to the electronic shop to buy replacements when I need to, because the resistors fe. had standard color-coding. :)
 

Some information on the Neural Net Prediction:
Neural Net Prediction: This one is super nerd-level exciting for me. Oh man. Inside the Zen architecture is an actual artificial neural network that is learning and predicting what instructions inside the CPU will be required. This learning AI is also predictive just like Smart Prefetch, but instead of predicting data, NNP is concerned with knowing what instructions and pathways inside the chip will be needed for executing an application. Staying on top of how an application can best flow through the chip also neutralizes a potential source of latency penalties.
...

RH: The “buffer” for pattern learning is not megabytes big or anything like that, so there is not room to remember hours or days of history. The buffer is typically flushed when you move from one app to the next, or when the system is restarted.
...

This does seem to point to some kind of contextual information in addition to branch history being stored, some of it would appear to be data that was derived from information the branch instruction wouldn't touch. It could be something basic like whether an instruction cache line or page hits the FPU, but perhaps it can record information from execution trends in the back end?

The life span indicates it is temporary and will eventually be overwritten. Saying that it usually loses history after switching applications, if defined as a user switching tabs or something like that, may mean that this history is tied to something that can survive a thread context switch--which would happen more frequently than changing applications. I'm not sure if that means there's a dedicated store for it, or if it's extra context in something that might live for a while, like the instruction cache or uop cache. The predictor/BTB/TLBs might be a place for that, although their space and timing is usually at a premium.
 
Back
Top