AMD sure got good mileage out of Trinity

[Charles Kozierok]

I'm just telling you what Intel said.. they said SDP would only be used in the case of low-power devices. I'm skeptical myself...

 

[Ketchup]

This is one of the reasons I just can't understand why AMD can't turn a profit, lol.

 

[Phynaz]

Oh, I know it's new and it's obviously not been calculated for older chips. I just making a silly example to demonstrate that Ark is not infallible, although it is clearly far better than the AMD offerings. I'm curious how it is not applicable to Xeon chips, though. Surely a metric intended to indicate power draw under a typical load would be even more useful for server markets which are constrained by performance/watt?

It's not even that. A Xeon wouldn't have an SDP. So Zero or NA, it means the same thing.

Scenario Design Power (SDP) is an additional thermal reference point meant to represent mainstream touch-first usages. It balances performance and mobility across PC and tablet workloads to extend capabilities into thin, thermally-constrained designs.

^ That's why it doesn't apply to Xeon.

 

[kornphlake]

From what I've been able to gather so far, there are no fewer than 1812 specific named CPU lines using the Trinity core. Many of them basically just rebrands of others.

It's bizarre. I get why they're doing it, but still.. wow.

If the wikipedia page is up to date and accurate Intel has 15 ivy bridge core i5 CPUs, confusion has nothing to do with AMD or their published specs, the confusion is the sheer number of options. Buying a processor has been a confusing decision ever since there has been more than one processor choice available. Even 20 years ago you had to choose between a 386 and a 386sx, Intel or AMD. There were at least 4 choices (and some minor options available as processors that were integrated into a motherboard,) the consumer has to make sense of it all. Last time I bought soup there were a dozen different choices of chicken noodle, it must be because Campbell's is in cahoots with the semi-conductor industry.

 

[Blandge]

From what I've been able to gather so far, there are no fewer than 1812 specific named CPU lines using the Trinity core. Many of them basically just rebrands of others.

It's bizarre. I get why they're doing it, but still.. wow.

I think every Intel Microarchitecture has about the same amount.

Sandy Bridge:
i7 extreme workstation
i7 desktop
i5 desktop
i3 desktop
Pentium desktop
Celeron desktop
i7 extreme mobile
i7 mobile
i5 mobile
i3 mobile
Pentium mobile
Celeron mobile
i7 ULV
i5 ULV
i3 ULV
Celeron ULV
Xeon SNB-EP
Xeon SNB-EN

 

[Charles Kozierok]

You're right, it's also true of Intel.

I guess it doesn't bother me as much with Intel because they do tend to avoid simply taking the same chip and rebranding it, which AMD does a lot. Plus, again, the "freedom of information" issue, which makes it easier to see what's what.

 

[Vesku]

You're right, it's also true of Intel.

I guess it doesn't bother me as much with Intel because they do tend to avoid simply taking the same chip and rebranding it, which AMD does a lot. Plus, again, the "freedom of information" issue, which makes it easier to see what's what.

I can see the information issue as the Intel Ark data contains more product specs than anything AMD maintains. But this same chip different chip argument doesn't really add up to me. SB/IB 2core, 4 core. Llano/Trinity 2 core, 4 core. Intel actually has more variation in terms of cache size and HT/no-HT.

 

[Shamrock]

That looks to be more like a templating auto-fill default value type situation. Intel must have updated their spec template but did not spend money on salaries for people to go back and vet the auto-generated template for zero-values for products which were alread EOL.

On the other hand it would be nice if AMD would actually fill in the spec values for the FX-8350 I bought, and that AMD still sells, considering it is an FX chip but I have no idea as to the max spec for operating temperature or voltage. Talk about selling something "as is"

Have you tried using the 8150 stats? I know it isn't exact. But it at least has the temps. Max temp is 61c. No voltages on it, though.

 

[Idontcare]

Have you tried using the 8150 stats? I know it isn't exact. But it at least has the temps. Max temp is 61c. No voltages on it, though.

I feel it is safe to assume that if AMD wanted us to assume 61C was the max temp for Piledriver as it is for Bulldozer then they would have prioritized someone to take the 500ms it takes to type "61C" in the max temp field for the spec sheet.

After all they felt it was worth the time and effort to have someone spend those 500ms when pulling together the public specs for bulldozer in the first place.

It stands to reason piledriver should have the same max temp as bulldozer, same 32nm process and all. But mine goes to 83C without throttling, and at 83C the computer just shuts off. It is very odd, not at all like Intel's TJmax.

 

[Shamrock]

Well, like you said, they just don't want any liabilities.

They will probably put all the info in, when they EOL the cpu.

I think it's safe to say you've voided your warranty, so 83c won't matter much. hehe

 

[NTMBK]

It's not even that. A Xeon wouldn't have an SDP. So Zero or NA, it means the same thing.



^ That's why it doesn't apply to Xeon.

Cheers, I hadn't seen that definition before. That makes more sense now.

 

[Atreidin]

If thermal solutions intended for 125W don't prevent the chip from throttling, isn't that a problem?

Where has that happened? The only instance I've heard of is the MSI motherboard with overheating VRMs because it was badly designed. The VRM overheating because it can't handle the current draw is not the same as the CPU overheating. If someone at MSI was stupid enough to use the TDP as a number for max power draw the VRMs should be able to handle, then some engineer in that company should get reprimanded or fired.

 

[mrmt]

Where has that happened? The only instance I've heard of is the MSI motherboard with overheating VRMs because it was badly designed. The VRM overheating because it can't handle the current draw is not the same as the CPU overheating. If someone at MSI was stupid enough to use the TDP as a number for max power draw the VRMs should be able to handle, then some engineer in that company should get reprimanded or fired.

As posted in another thread, ASRock boards also have the same problem. More important we have a MSI official statement saying that the 8350 breached the specifications, and no official rebuttal from AMD.

So between choosing an official statement from a MB company and mere faith in a processor which doesn't even have its official specs published by the manufacturer, the more logical choice is rather obvious.

I wonder if MSI, ASRock and AMD are liable to class action lawsuits. What are people waiting for? There won't be any money to receive after AMD goes bankrupt.

 

[Idontcare]

Where has that happened? The only instance I've heard of is the MSI motherboard with overheating VRMs because it was badly designed. The VRM overheating because it can't handle the current draw is not the same as the CPU overheating. If someone at MSI was stupid enough to use the TDP as a number for max power draw the VRMs should be able to handle, then some engineer in that company should get reprimanded or fired.

Impressive demonstration of "AMD can do no wrong".

Motherboards are spec'ed to handle specific TDP classes, this is not new information. There is a reason why a mobo spec'ed to support 125W CPUs cannot support 140W CPUs.

Remember all those boards that were burning up when AMD launched Phenom and people were popping their 140W TDP phenoms into their 125W mobos? There is a reason why that happened.

MSI is saying that their engineers have discovered some non-zero percentage of FX chips exceed the 125W rating, for all intents and purposes those chips have to be treated as if they are 140W TDP and not 125W TDP when it comes to mobo design.

Other mobo makers may be over-engineering their mobos, at added expense to the customer of course. Anyone can design a 95W mobo to handle 125W processors but not officially support them, but that "95W class" mobo is going to have a more expensive BOM than the competition who designed their 95W mobo's for 95W CPUs.

MSI designed their 125W mobos for 125W TDP CPUs, not 140W TDP CPUs, and then they found out AMD should be spec'ing their top-end FX chips as 140W TDP processors for all the intents and purposes that classifying TDP matters.

What is wrong with calling a spade a spade? AMD won't even spec the max voltage or temperature for their FX08350 processor, is it really all that suprising to consider that maybe AMD doesn't care if their 125W TDP processors should actually be spec'ed as 140W TDP processors? I don't think AMD gives two craps about it.

 

[Charles Kozierok]

What is wrong with calling a spade a spade? AMD won't even spec the max voltage or temperature for their FX08350 processor, is it really all that suprising to consider that maybe AMD doesn't care if their 125W TDP processors should actually be spec'ed as 140W TDP processors? I don't think AMD gives two craps about it.

It could be even worse than that. It could be that they know that in order to remain even remotely competitive with Intel, they needed to up the clock speeds and not up the TDPs. So... they did -- on paper.

 

[Idontcare]

It could be even worse than that. It could be that they know that in order to remain even remotely competitive with Intel, they needed to up the clock speeds and not up the TDPs. So... they did -- on paper.

You have a point. Considering that at the time Intel had already reduced its TDP to 95W for its top-end mainstream class processors (i5-2500k, i7-2600k) and was in the process of reducing it further still to 77W. Not a good time to be fielding a 125W TDP product, let alone coming out and increasing it another 15W (which apparently is what MSI and Asrock engineers are saying is what should have happened with the release of the FX-8350 SKU).

So it may well be that this sudden lack of documentation from AMD is all part of an intentional smokescreen operation to keep themselves as technically in the clear as possible. Heh, like the Chinese "You don't technically have any proof that we are responsible for the widespread hackings, so technically we cannot be held accountable".

AMD has ensured themselves plausible deniability with the FX-8350 by refusing to publish technical specs of nearly every kind. Someone within AMD is clearly playing poker with the consumer and the mobo makers, the only questions are "why" and "what else are they keeping close to the vest"?

 

[Charles Kozierok]

I think both companies are playing games with these power figures. Last month it was Intel inventing "SDP". Both companies are putting out chips with lower base speeds and massive turbo boosts, lower TDPs while advertising "up to" the turbo speed.

I also noticed yesterday when comparing some Llano and Trinity chips that the TDP drops are rather surprising in some cases.

A4-3300M: 1.9/2.5 GHz, 2MiB L2, HD6480@444MHz.
A4-4355M: 1.9/2.4 GHz, 1MiB L2, HD7400@327/424 MHZ.

The Llano chip has a TDP of 35W, and the Trinity chip 17W.

Seems like a lot more of a power drop than I'd expect given the stats and the much smaller drops for other chips.

 

[mrmt]

I think both companies are playing games with these power figures. Last month it was Intel inventing "SDP". Both companies are putting out chips with lower base speeds and massive turbo boosts, lower TDPs while advertising "up to" the turbo speed.

I don't think Intel is messing with TDP yet. The whole SDP scheme is a a mean to better advertise their products for a specific scenario while at the same time keeping TDP as a credible metric.

It's not the same thing that AMD is doing, they are basically toasting TDP credibility (and whatever corporate credibility they have left) with this whole 8350 debacle. I wonder if some review site is interested in probing the TDP of the entire AMD line up. I bet they wouldn't find nice things there, but nobody will do that, as they would be forever blacklisted by AMD. It would yield a nice class action lawsuit also.

 

[Charles Kozierok]

I don't think Intel is messing with TDP yet. The whole SDP scheme is a a mean to better advertise their products for a specific scenario while at the same time keeping TDP as a credible metric.

Maybe. But I was very annoyed with the rollout of their "Y" chips, because they introduced SDP in order to come up with their 7W figure, and then spread that number around in such a way that people thought it was comparable to TDP figures from earlier processors.

Hopefully they will be more clear in the future.

 

[mrmt]

Maybe. But I was very annoyed with the rollout of their "Y" chips, because they introduced SDP in order to come up with their 7W figure, and then spread that number around in such a way that people thought it was comparable to TDP figures from earlier processors.

I was annoyed too, they should have been more transparent when rolling out the metric.

 

[Piroko]

MSI is saying that their engineers have discovered some non-zero percentage of FX chips exceed the 125W rating, for all intents and purposes those chips have to be treated as if they are 140W TDP and not 125W TDP when it comes to mobo design.

Other mobo makers may be over-engineering their mobos, at added expense to the customer of course.

From a consumer standpoint I don't support MSI here. All electronics age at least a bit with usage, even solid caps and VRMs. Thus, designing a board which can just barely run the processors it was designed to run while being factory new just reeks of trouble to me.
A new (premium) forklift truck can easily stem 1.3 times of its maximum load, because it will still be able to handle its rated load 10 years from now. I can plug two Headsets into the same jack of my X-Fi even though the impedance will be out of spec, I can plug a 2.5" usb drive into one USB port only and be fine, put 1.5V cells into my 'NIMH only' mouse, oc my Graphics card, I could even oc my Monitor because its chip is designed with a robust safety. Consumers do not lose anything with sensible over-engineering because that is - and should be - the norm, not the exception.

Of course it's bad if AMD actually exceeded their own specs (and it's good if there are well designed safeties in place), but I would expect a new 125W board to handle a 137W load (+10%) for some months at least.

 

[Idontcare]

From a consumer standpoint I don't support MSI here. All electronics age at least a bit with usage, even solid caps and VRMs. Thus, designing a board which can just barely run the processors it was designed to run while being factory new just reeks of trouble to me.
A new (premium) forklift truck can easily stem 1.3 times of its maximum load, because it will still be able to handle its rated load 10 years from now. I can plug two Headsets into the same jack of my X-Fi even though the impedance will be out of spec, I can plug a 2.5" usb drive into one USB port only and be fine, put 1.5V cells into my 'NIMH only' mouse, oc my Graphics card, I could even oc my Monitor because its chip is designed with a robust safety. Consumers do not lose anything with sensible over-engineering because that is - and should be - the norm, not the exception.

Of course it's bad if AMD actually exceeded their own specs (and it's good if there are well designed safeties in place), but I would expect a new 125W board to handle a 137W load (+10%) for some months at least.

As I understand it, the throttling threshold was set to not kick-in unless the processor was actually exceeding 140W, well above the spec'ed TDP.

I don't think MSI did anything that violates the spirit of your post. They are professionals, after all, this isn't like some no-name generic brand manufacturer is doing this with gray-market goods.

But MSI is not a charity, and they are competing with mobo suppliers who are not charities. Everything that is over-engineered is done so at an increase in the BoM and ultimately the customer picks up that tab.

For every MSI customer who buys their AM3+ mobo but doesn't buy an FX-8350 to go into it, the mobo is over-kill because nothing short of an 8150 or 8350 is going to touch the 125W TDP spec.

Is it really fair to ask every FX-4100 to FX-4170, or FX-6100 to FX-6200, or FX-8100 to FX-8140 to pay extra for their MSI mobo just because MSI had no other choice than to make it bullet proof over-engineered for that handful of 8150 and 8350 owners that might come along with an over-spec processor?

That is a rather unfair policy IMO, expecting the majority to subsidize the few just because the few want to buy a limited selection of processors that are not binned correctly with respect to the manufacturers spec.

The irony here is that AMD is the culprit, they consciously decided to force the mobo makers and cpu buyers into making these decisions (over-design a mobo to benefit a few people at the expense of elevating cost for everyone else) because AMD doesn't want to rate the FX-8350 as the 140W TDP product that it is...and yet the consumer outrage is directed towards the mobo maker who has been left holding the bag.

 

[Piroko]

I think we have to be a bit more precise here, the original thread was about Prime95 and other burn-in tools which may cause throttle and which at the same time are excluded from "typical load at which TDP is defined". The processor is actually allowed to exceed its own spec in this case.

I hate car analogies, but kinda like a car exceeding its rated top speed because you're driving down a hill. Most cars will just let you because they're over-engineered, but some cars will slow down by cutting the fuel injection for a second. I had this exact feature on a Toyota Corolla some 15 years ago, the first time the car slowed me down I thought something was broken...

Anyways, the majority will also subsidize recalls from people who think their boards are defective because of the odd behaviour. MSI could just as easily exclude 8350s from the compatibility list and make a public statement as to why. That would have put pressure on AMD to keep the safety margins higher on their end. Instead we now have two companies who pushed their safety margins too low and the consumers are loosing...

 

[Vesku]

AFAIK (quick peak online) the other more robust 970 motherboards compete well on price with the troubled MSI one.

Not sure what this partner dispute has to do with "got good mileage out of Trinity".