TR Forums

OK I was thinking of AMD's upcoming 8 core 95w TDP Zen CPU and how cool it would be to stick it in a SFF case such as the  Silverstone ML08-H. I've been watching build videos using various MiniITX cases and I think if I was going to build a SFF system, then this is the case I would use. Looks easy enough to do without a bunch of hassle.

First I'll start out with a MiniITX board with built in WiFi
The CPU w/ a low profile HSF
32 GB of memory
2 large SSDs
GTX 1070 video card

Does this sound good?

Where are you going to stick the low-profile Milo ML08 mini-ITX case that a micro-ATX solution like the Grandia GD05, Sugo SG10 or Temjin TJ08-E wouldn't work?  I built my latest gaming PC in a mini-ITX Fortress FTZ01S because it seemed cool to make a tiny PC, but there are compromises in noise, expandability, ease of assembly and cost when you shrink down to mini-ITX from micro-ATX.

Additionally, why...

It's fun as an exercise, but invariably the tradeoff to putting high-TDP parts in a small space is more noise, and possibly a lower performance ceiling (limited overclocking, or even boost throttling...).

You just had to spoil my fun didn't you?

If you want tot do it for fun, then it's your money.  As long as you are aware of the compromises up front, go right ahead.  For me, the coolness of the small mini-ITX form factor doesn't make up for the worse cooling/noise options.  I will go back to micro-ATX for my next gaming PC.

JustAnEngineer wrote:

Where are you going to stick the low-profile Milo ML08 mini-ITX case that a micro-ATX solution like the Grandia GD05, Sugo SG10 or Temjin TJ08-E wouldn't work?  I built my latest gaming PC in a mini-ITX Fortress FTZ01S because it seemed cool to make a tiny PC, but there are compromises in noise, expandability, ease of assembly and cost when you shrink down to mini-ITX from micro-ATX.

Thank for mentioning those MicroATX cases, especially the Sugo SG10. I might look into that one instead.

Yeah, unless the size of the system is constrained by where you intend to put it, micro-ATX should give you more options for motherboards, and is better suited to a high(er) TDP part like the 8c/16t undoubtedly will be.

I had switched over to micro-ATX during the AM2/AM2+ generation, but then AM3+ hit, with its lack of decent micro-ATX motherboards. So I'm currently back on full ATX, but will probably downsize to micro-ATX again if/when I build a system around Zen.

JustAnEngineer wrote:

Where are you going to stick the low-profile Milo ML08 mini-ITX case that a micro-ATX solution like the Grandia GD05, Sugo SG10 or Temjin TJ08-E wouldn't work?  I built my latest gaming PC in a mini-ITX Fortress FTZ01S because it seemed cool to make a tiny PC, but there are compromises in noise, expandability, ease of assembly and cost when you shrink down to mini-ITX from micro-ATX.

The ML08 seems to have improved on the ease of assembly compared to the ML07/FTZ01 and Skylake Mini-ITX motherboard prices are lower than they were with Haswell.

Noise may still be an issue, although I'm not particularly convinced it'll be that much different between a small Mini-ITX case and a small Micro ATX case.

If you're going for a fairly expensive Mini-ITX system then I'd be inclined to go for one of the more premium cases. The Dan A4-SFX should come into stock in OcUK/Caseking in a couple of months, while the Sentry case starts their fundraiser soon:
https://www.youtube.com/watch?v=WDXKflm_RWo

In terms of Micro Micro-ATX cases there isn't actually a lot available. A decent selection of mid-sized options for higher end hardware like the SG10 but smaller ones for mid-range hardware are difficult to find. The Cooltek Coolcube Maxi aka Jonbo V4 looks reasonable, but there's a serious lack of options at the moment.

I'll almost certainly put 8C Zen in my NCase M1 if it has mITX options, ECC support is there, and the chipset isn't a terrible mess. Thermals and noise won't be issues, but I do have room for more CPU cooler than most cases this size.

I've been thinking some more about this some more and since I have hearing loss, I'm wondering just how much noise I will notice anyway? My desk has plenty of empty space beneath it so I have lots of room for my full size ATX Corsair 300R build for now. But in the future I could live somewhere with much less space available. So this one reason I'm looking a smaller form factors.
 

Depending on how Zen performs I will be using in in a smaller case but still matx form. I'm currently running AMD hardware in a small configuration (Raijentek Stryx) and have had zero issues with heat. FX8120 at 4.0Ghz, R9 380 overclocked are main sources of heat. The northbridge and VRMs but out quite a bit of heat. The 40mm dan on northbirdge keeps cool and VRM is exhuasted quickly. All the fans are controlled by NZXT grid on silent mode. On whatever next build in this case I plan to have no mechanical drives. I looked into various case flow and fan setups and profiles. Once you calibrate that I feel like you can do most anything.
 

I've been seeing this "Zen is going to be noisy in SFF cases" issue coming up in a couple other threads, and I don't completely get it. We're pretty sure by now that it has a 95W TDP comparable to that of typical Intel quads, and that TDP doesn't cause much trouble there. AVX w/ FMA is responsible for lots of the TDP on the Intel side, but the other four cores are responsible for lots of the TDP on the AMD side. Balancing those out, task energy probably isn't so different either. If some software like a game isn't making full use of eight cores and the performance isn't there, the power use is probably also quite a bit below max. Unless you want to leave a heavy full-time task like F@H running, I don't see the average noise levels being much different. Maybe noise would be a bit burstier when all eight cores are doing their thing, but if they're all in action and getting the job done that much faster, I don't see the problem.

synthtel2 wrote:

I've been seeing this "Zen is going to be noisy in SFF cases" issue coming up in a couple other threads, and I don't completely get it. We're pretty sure by now that it has a 95W TDP comparable to that of typical Intel quads, and that TDP doesn't cause much trouble there. AVX w/ FMA is responsible for lots of the TDP on the Intel side, but the other four cores are responsible for lots of the TDP on the AMD side. Balancing those out, task energy probably isn't so different either. If some software like a game isn't making full use of eight cores and the performance isn't there, the power use is probably also quite a bit below max. Unless you want to leave a heavy full-time task like F@H running, I don't see the average noise levels being much different. Maybe noise would be a bit burstier when all eight cores are doing their thing, but if they're all in action and getting the job done that much faster, I don't see the problem.

I guess we have to wait and see.

synthtel2 wrote:

I've been seeing this "Zen is going to be noisy in SFF cases" issue coming up in a couple other threads, and I don't completely get it. We're pretty sure by now that it has a 95W TDP.

But that's the point- putting a 95W TDP CPU in a small space and expecting it to perform quietly isn't a great idea in the first place, and the enclosure linked in the OP doesn't look terribly suited to quietly exhausting that amount of heat.

(though I do get your point that Zen may not be configured to put all of its performance and associated TDP to use for say gaming workloads, but I prefer to be optimistic)

Maybe we have different definitions of quiet? I'm thinking of a 4790(K)/6700(K) at stock clocks under one of the aftermarket coolers that take about as much space as the stock Intel one (the Noctua NH-L9s come to mind, but IIRC Noctua isn't actually the best at cooling in that form factor). The ML08 doesn't have forced case airflow, but if it's standing upright, there will be quite a bit of convection through the mobo area (source: I've run rigs like that before). That setup should be easily better than a stock Intel cooler in free air, and those are (IME) only mildly bad if you don't let them go to 100% and aren't OCing.

whm1974 wrote:

You just had to spoil my fun didn't you?

Ignore them, nothing wrong with mITX machines, you can do amazing ones.
Look up the SUGO 08 or Node 304 "owners club" threads at Overclock.net, it's chock full of good info.

Not *all* Zen, just the 8-core.

If you're building SFF in the first place, performance probably isn't your primary concern; size is. SFF typically limits airflow compared to a larger case, and places size constraints on the size of the CPU cooler. This is going to put a lower limit on how quiet you can make the system and still maintain reasonable CPU temperatures under full load. Yeah, you'll be able to get it reasonably quiet if you're smart about how you do things... but not as quiet as you could in a non-SFF form factor.

Unless you really need (or want...) *both* small form factor *and* 8 cores, why wouldn't you go for a lower TDP quad core CPU, or step up to a micro-ATX form factor?

AbRASiON wrote:

whm1974 wrote:

You just had to spoil my fun didn't you?

Ignore them, nothing wrong with mITX machines, you can do amazing ones.
Look up the SUGO 08 or Node 304 "owners club" threads at Overclock.net, it's chock full of good info.

Thanks I'll do that.

Yeah, mATX will be much quieter but that's not always the point.

I would never use one myself (I'm a quiet > performance sort of person) but I had a lot of fun building into a FTZ01 with a GTX970 and Haswell i5 for someone's couch gaming build. Out of the box it was a bit noisy but some silicone sealant to mount different fans with and a bit of dremeling to remove airflow obstructions made a huge difference to the noise. The fun was in trying to make it quiet when 250W in a tiny space is not going to be by default. I was pleased enough with the results that I'd recommend that case for anyone willing to experiment (and presumably the Milo and Raven equivalents based on the same chassis and internal layout).

I don't think you can build an mITX case and just expect a quiet system without some careful component choice. You should also stick to mATX if you're not expecting to do some light modding.

Thanks Chrispy_ but I won't be building until 2018 or 2019. Hopefully AMD will have a much improved second generation Zen core by then.  

synthtel2 wrote:

I'll almost certainly put 8C Zen in my NCase M1 if it has mITX options, ECC support is there, and the chipset isn't a terrible mess. Thermals and noise won't be issues, but I do have room for more CPU cooler than most cases this size.

I sort of expected AMD to restrict ECC support to the Naples or however the "server" edition is called. Are you guessing that ECC support will be included or do you have information to that effect?

whm1974 wrote:

Thanks Chrispy_ but I won't be building until 2018 or 2019. Hopefully AMD will have a much improved second generation Zen core by then.  

By then, either Zen will already be a success or AMD will be planning their exit strategy from the CPU market. I don't think their CPU division can survive another Bulldozer.

ptsant wrote:

I sort of expected AMD to restrict ECC support to the Naples or however the "server" edition is called. Are you guessing that ECC support will be included or do you have information to that effect?

Restricting it to the server version would be somewhat of a break with tradition for AMD, though they have arguably been moving in that direction since they introduced the APU line (none of which support ECC). Aside from the APUs and a few random low-end Sempron SKUs, AFAIK all of their desktop chips since the original Athlon64 have had ECC support baked in.

Motherboard support, OTOH, has been a bit of a crapshoot. ASUS has generally been better than other vendors at including it on their consumer/enthusiast boards; I expect that if the desktop AM4 CPUs have ECC support, you'll want to put ASUS at the top of your short list if you want to use ECC DIMMs.

just brew it! wrote:

whm1974 wrote:

Thanks Chrispy_ but I won't be building until 2018 or 2019. Hopefully AMD will have a much improved second generation Zen core by then.  

By then, either Zen will already be a success or AMD will be planning their exit strategy from the CPU market. I don't think their CPU division can survive another Bulldozer.

Yeah that the thing if AMD will still be selling CPUs by then.

whm1974 wrote:

OK I was thinking of AMD's upcoming 8 core 95w TDP Zen CPU and how cool it would be to stick it in a SFF case such as the  Silverstone ML08-H. I've been watching build videos using various MiniITX cases and I think if I was going to build a SFF system, then this is the case I would use. Looks easy enough to do without a bunch of hassle.

First I'll start out with a MiniITX board with built in WiFi
The CPU w/ a low profile HSF
32 GB of memory
2 large SSDs
GTX 1070 video card

Does this sound good?

I'd GREATLY temper the idea with some history of AMD's cpus regarding temperatures.  Until we actually see 90% "load" on one and see the actual real world temps i'd hold off on dropping some coin.

If the 95TPD is 'optimistic' it may end up throttling a majority of the time when gaming or encoding. 

Well have to see in 90 days whats what with regard to actual operating temps at 3.5+ghz...

maxxcool wrote:

I'd GREATLY temper the idea with some history of AMD's cpus regarding temperatures. Until we actually see 90% "load" on one and see the actual real world temps i'd hold off on dropping some coin.

If the 95TPD is 'optimistic' it may end up throttling a majority of the time when gaming or encoding.

Well have to see in 90 days whats what with regard to actual operating temps at 3.5+ghz...

1. The temperatures are going to be more reflective of the cooling setup, and therefore "seeing" them isn't very helpful. You might actually be "seeing" how someone is running the processor in an enclosed shoebox with 100 gram aluminum heatsink and 1000rpm 60mm fan.

2. Is there any indication that any manufacturer is lying about their TDP numbers as a maximum? In fact, doesn't SDP at least partially exist because TDP is a larger value than what practical use usually generates (or how it can be technically constrained into lower tiers)? In other words, that TDP isn't "optimistic" at all, but rather "conservative"?

3. If it is throttling with their stock cooler in a reasonably-sized/cooled case at room temperatures, than it is not only falsely-advertised but plainly defective. It would also be immediately evident in the reviews, which all discerning people should be waiting for anyway.

Glorious wrote:

maxxcool wrote:

I'd GREATLY temper the idea with some history of AMD's cpus regarding temperatures.  Until we actually see 90% "load" on one and see the actual real world temps i'd hold off on dropping some coin.

If the 95TPD is 'optimistic' it may end up throttling a majority of the time when gaming or encoding.

Well have to see in 90 days whats what with regard to actual operating temps at 3.5+ghz...

1. The temperatures are going to be more reflective of the cooling setup, and therefore "seeing" them isn't very helpful. You might actually be "seeing" how someone is running the processor in an enclosed shoebox with 100 gram aluminum heatsink and 1000rpm 60mm fan.

2. Is there any indication that any manufacturer is lying about their TDP numbers as a maximum? In fact, doesn't SDP at least partially exist because TDP is a larger value than what practical use usually generates (or how it can be technically constrained into lower tiers)? In other words, that TDP isn't "optimistic" at all, but rather "conservative"?

3. If it is throttling with their stock cooler in a reasonably-sized/cooled case at room temperatures, than it is not only falsely-advertised but plainly defective. It would also be immediately evident in the reviews, which all discerning people should be waiting for anyway.

As far as #3 go I will probably have to buy an aftermarket HSF anyway since I want to install the CPU in a  mITX case in order to fit. And overclockers will have too as well.

Glorious wrote:

maxxcool wrote:

I'd GREATLY temper the idea with some history of AMD's cpus regarding temperatures.  Until we actually see 90% "load" on one and see the actual real world temps i'd hold off on dropping some coin.

If the 95TPD is 'optimistic' it may end up throttling a majority of the time when gaming or encoding.

Well have to see in 90 days whats what with regard to actual operating temps at 3.5+ghz...

1. The temperatures are going to be more reflective of the cooling setup, and therefore "seeing" them isn't very helpful. You might actually be "seeing" how someone is running the processor in an enclosed shoebox with 100 gram aluminum heatsink and 1000rpm 60mm fan.

2. Is there any indication that any manufacturer is lying about their TDP numbers as a maximum? In fact, doesn't SDP at least partially exist because TDP is a larger value than what practical use usually generates (or how it can be technically constrained into lower tiers)? In other words, that TDP isn't "optimistic" at all, but rather "conservative"?

3. If it is throttling with their stock cooler in a reasonably-sized/cooled case at room temperatures, than it is not only falsely-advertised but plainly defective. It would also be immediately evident in the reviews, which all discerning people should be waiting for anyway.

(1) Given my long history with AMD cpus in my lab,  'Seeing' with 80$+ coolers is indicative of the characteristics of a particular process and the ability to 'give up heat' .. and the past ratings were quite optimistic at best under sustained 100% load for weeks on end for my work.

(2) Not yet, that is why I said we have to 'wait and see'

(3) And we will have to wait and see.. 

synthtel2 wrote:

We're pretty sure by now that it has a 95W TDP comparable to that of typical Intel quads, and that TDP doesn't cause much trouble there.

Let's not forget that historically AMD has had a tendency to underestimate their TDP.  Whereas Intel's TDP numbers typically track fairly close to power draw, it's difficult to explain why (for example) the FX8370 draws 61W more than the i7-6700K despite a TDP difference of 34W.  I have a hard time believing that the AM3+ platform power usage is responsible for ALL of that additional 27W power draw.

DPete27 wrote:

synthtel2 wrote:

We're pretty sure by now that it has a 95W TDP comparable to that of typical Intel quads, and that TDP doesn't cause much trouble there.

Let's not forget that historically AMD has had a tendency to underestimate their TDP.  Whereas Intel's TDP numbers typically track fairly close to power draw, it's difficult to explain why (for example) the FX8370 draws 61W more than the i7-6700K despite a TDP difference of 34W.  I have a hard time believing that the AM3+ platform power usage is responsible for ALL of that additional 27W power draw.

Keep in mind the 9* W TDP of the 6700k includes the iGPU.

maxxcool wrote:

(1) Given my long history with AMD cpus in my lab, 'Seeing' with 80$+ coolers is indicative of the characteristics of a particular process and the ability to 'give up heat' .. and the past ratings were quite optimistic at best under sustained 100% load for weeks on end for my work.

The ratings are for wattage, not temperature: TDP specifies the maximal power the chip will consume, not the temperature it is while doing it. The manufacturers do not "optimistically" rate the temperature these things will run at, no, they specify an operating temperature range the device must stay within in order to properly function. As it is range, it has a lower and higher bound, so if your ambient is Antarctica or Death Valley you're probably violating it either way before you even switch the power on.

Thus whatever it is you are saying is, at best, not even wrong.

maxxcool wrote:

(2) Not yet, that is why I said we have to 'wait and see'

If you were saying anything intelligible at all, you just said in 1) that the TDP is "optimistic". So that's contradictory.

And, if we ignore that, so what? They *might* eventually just make it up? Well, dude, they *might* sell me a two-decade old Cyrix chip in that Zen box, they *might* sell me cow chips (kinda like what you're trying to sell me on now!), etc...

maxxcool wrote:

(3) And we will have to wait and see..

I don't have to "wait and see" in order to understand that you are completely talking out of your posterior, and neither does anyone else. You're just running your mouth and spewing nonsense.

DPete27 wrote:

Let's not forget that historically AMD has had a tendency to underestimate their TDP. Whereas Intel's TDP numbers typically track fairly close to power draw, it's difficult to explain why (for example) the FX8370 draws 61W more than the i7-6700K despite a TDP difference of 34W. I have a hard time believing that the AM3+ platform power usage is responsible for ALL of that additional 27W power draw.

Not really difficult at all, actually: different chipsets (990FX is like 20W TDP, Z170 is 6w), different boards (most power is consumed by the VRM setup, so even minor differences in component choice there can add up, and that additional space between feeding 91w and 125w adds up, especially since it's always an efficency curve) plus speedy DDR3 at 1.5v versus normal-speed DDR4 at 1.2v (that can potentially be 5-10 watts at load, just right there). The differences add up.

EDIT: Also what DrDominodog51 said which I didn't think of: because Intel must budget the iGPU in, there is reason to think a i7-6700K at full CPU load isn't as close to its TDP as the FX8370 would be, because the FX's TDP doesn't have to account for that.

AMD's TDP definition isn't exactly the same as Intel's. Yes, it's looser. But it's not THAT loose, and it really can't be: OEMs with real engineers who design stock systems rely on the same number to come up with adequate but cost-conscious cooling systems. The margins are razor thin, particularly on what are effectively "budgetish" parts, so they'd be screaming bloody murder if AMD was reliably under-estimating by a 25%.

And, on the other side, why would AMD publish those jokes at 220w a couple of years ago if the numbers were so flexible? Why not 200? or 175?