AMD Carrizo Pre-release thread

[ET]

Im really wondering if the difference is notable in everydays tasks. And as soon as you start gaming, you'll be held back by slow ram and low tdp. The business case / market segment for this 15w Carrizo seems very meagre because low segment will be covered by Carrizo-L and lower main by I3.

By this measure there's no point for Core CPU's at low TDP and everyone will go for Atom tech. I don't think that's the way it really goes.

 

[mpc007]

By this measure there's no point for Core CPU's at low TDP and everyone will go for Atom tech. I don't think that's the way it really goes.

That is not true, the Intel U tier models are still a whole class faster than 4core Bay Trail Atoms. When you say overpriced low tdp parts i think of Core M (the previous Y tier), but if I'm not mistaken they don't indeed sell very well! So that only proves my point that this is a meagre segment. The average customer isn't that smart, but they understand its a bad idea to pay more than 500 dollar for a CPU which doesn't even make it to 2.5 GHz.

It seems to come down to a difference in opinion between chipmakers and customers about what's important. I really think TDP improvements should get less priority than performance improvements, and so do customers. However, Intel and AMD seem to force lower TDPs through our throats every time again. If I buy a new notebook, I'm first looking up the percentual performance gains, and not if it will run 20 minutes longer on the battery. What weighs a lot in this case, is that OEMs decide how long it will run, because they will pick the accu size. Because that factor can't be controlled, I'd like to focus on things that CAN be compared, such as performance gains. Par example: Its very nice that a new cpu will use 5w less, but if OEMs completely blow that nice effect because they mount a smaller battery (what we already are seeing for a few years) I'm not being helped AT ALL, unless the chip also has something else to offer: more performance (OEMS can't blow that, however that may change with the cTDP on Carrizo..)

 

[Stennan]

Not been tracking all the threads lately, but it would be good to know how these Carrizos compare to a intel + nvidia 840/940? It creates a lot of uncertainty for me as a buyer to have so much variables. Does it have dual channal? Is it full speed memory? Will TDP be configured to 15w or 35w? Gah!!

Right now I can get a 14 inch FHD Lenovo Flex 2 with intel i5 and 840m for around 6200 Swedish Crownes and that includes a 6 hour battery. looks more attractive in my eyes as I know the performance I will be getting.

 

[Enigmoid]

Not been tracking all the threads lately, but it would be good to know how these Carrizos compare to a intel + nvidia 840/940? It creates a lot of uncertainty for me as a buyer to have so much variables. Does it have dual channal? Is it full speed memory? Will TDP be configured to 15w or 35w? Gah!!

Right now I can get a 14 inch FHD Lenovo Flex 2 with intel i5 and 840m for around 6200 Swedish Crownes and that includes a 6 hour battery. looks more attractive in my eyes as I know the performance I will be getting.

Carrizo is comparable to an i3U in terms of CPU performance. It is significantly weaker than a 940m.

http://www.notebookcheck.net/HP-Pavilion-17-g054ng-Notebook-Review.147794.0.html

Looks like a 940m is around 50-100% stronger.

With a dgpu you also have fewer concerns about tdp limitations. Just make sure that notebook doesn't overheat.

 

[Abwx]

Carrizo is comparable to an i3U in terms of CPU performance.

That means nothing, i5/I7consume more than i3s to get higher scores, that s how they get higher perfs, there s no miracle.

Carrizo is comparable to i3/i5/i7 at the same time, the only thing differentiating thoses chips being real TDP, 15W officialy for everything and real TDPs that range from 15W to 28W depending of how the OEMs did set the power management, of course if its available on the used chips.

 

[mpc007]

Not been tracking all the threads lately, but it would be good to know how these Carrizos compare to a intel + nvidia 840/940? It creates a lot of uncertainty for me as a buyer to have so much variables. Does it have dual channal? Is it full speed memory? Will TDP be configured to 15w or 35w? Gah!!

Right now I can get a 14 inch FHD Lenovo Flex 2 with intel i5 and 840m for around 6200 Swedish Crownes and that includes a 6 hour battery. looks more attractive in my eyes as I know the performance I will be getting.

You're asking the right questions. Because Carrizo has so much variables, performance stretches across a wide field.

If you have the A10-8700 with single channel ram at 1600, at 15w tdp, a slow hdd, and no dGPU, you shouldnt spend more than 450 euros on it. Performance cpu wise should be similar to i3 u series, with graphics performance comparable to a GeForce 820 / 920m (but stronger than Intel HD).

If you have the FX-8800 with dual channel ram at 2133, at 35w tdp, with ssd, and additional Radeon R7 M360 in dual graphics (which makes it a R7 365DX), its worth around 650 euros. Performance cpu wise should be close to i5 u series, with graphics performance comparable to GeForce GT840 / 940m).

If you can only get the first config (which it looks like at the moment), please, go for the Lenovo Flex. I have it too (with AMD Beema), and its very decent for its price. If you could get the second config (in 14 inch), which I'm very excited about and really crave for, get that.

 

[Enigmoid]

That means nothing, i5/I7consume more than i3s to get higher scores, that s how they get higher perfs, there s no miracle.

Carrizo is comparable to i3/i5/i7 at the same time, the only thing differentiating thoses chips being real TDP, 15W officialy for everything and real TDPs that range from 15W to 28W depending of how the OEMs did set the power management, of course if its available on the used chips.

One really doesn't care about TDP per se. One cares about the absolute performance and things like battery life and heat.

i3-i7 will throttle down to 15W after the long term turbo ends (less than 10 seconds) unless the manufacturer has configured the chip to use more power.

You're asking the right questions. Because Carrizo has so much variables, performance stretches across a wide field.

If you have the A10-8700 with single channel ram at 1600, at 15w tdp, a slow hdd, and no dGPU, you shouldnt spend more than 450 euros on it. Performance cpu wise should be similar to i3 u series, with graphics performance comparable to a GeForce 820 / 920m (but stronger than Intel HD).

If you have the FX-8800 with dual channel ram at 2133, at 35w tdp, with ssd, and additional Radeon R7 M360 in dual graphics (which makes it a R7 365DX), its worth around 650 euros. Performance cpu wise should be close to i5 u series, with graphics performance comparable to GeForce GT840 / 940m).

If you can only get the first config (which it looks like at the moment), please, go for the Lenovo Flex. I have it too (with AMD Beema), and its very decent for its price. If you could get the second config (in 14 inch), which I'm very excited about and really crave for, get that.

See the notebookcheck review of carrizo. Dual graphics is simply a bust at this point in time and has been for the last several years.

 

[mpc007]

Well, dual graphics MAY fail to deliver the promised performance in some cases, due to bad driver optimisation, but there are enough benchmarks which prove it's a promising concept, which actually delivers good power when playing certain games. For me, I'd just be glad to have the option to use DX. After all, having a IGP and dGPU which are almost identical and NOT be able to use them together, would be a waste of potential.

 

[Abwx]

One really doesn't care about TDP per se. One cares about the absolute performance and things like battery life and heat.

Performance for Carrizo in Cinebench 11.5 range from 2.4 at 15W to 3.2 at 35W, i guess that it cover the whole i3/i5/i7 line..

i3-i7 will throttle down to 15W after the long term turbo ends (less than 10 seconds) unless the manufacturer has configured the chip to use more power.

You are left repeating what i said but in an apologist way, fact is that Intel use the 15W moniker for CPUs whose frequency is in a 1.6 ratio, that s a TDP range that is in a 2.5 ratio, so it s clear that their ratings are complete bs and are underspecced for most of the line.

 

[Yuriman]

That is not true, the Intel U tier models are still a whole class faster than 4core Bay Trail Atoms. When you say overpriced low tdp parts i think of Core M (the previous Y tier), but if I'm not mistaken they don't indeed sell very well! So that only proves my point that this is a meagre segment. The average customer isn't that smart, but they understand its a bad idea to pay more than 500 dollar for a CPU which doesn't even make it to 2.5 GHz.

It seems to come down to a difference in opinion between chipmakers and customers about what's important. I really think TDP improvements should get less priority than performance improvements, and so do customers. However, Intel and AMD seem to force lower TDPs through our throats every time again. If I buy a new notebook, I'm first looking up the percentual performance gains, and not if it will run 20 minutes longer on the battery. What weighs a lot in this case, is that OEMs decide how long it will run, because they will pick the accu size. Because that factor can't be controlled, I'd like to focus on things that CAN be compared, such as performance gains. Par example: Its very nice that a new cpu will use 5w less, but if OEMs completely blow that nice effect because they mount a smaller battery (what we already are seeing for a few years) I'm not being helped AT ALL, unless the chip also has something else to offer: more performance (OEMS can't blow that, however that may change with the cTDP on Carrizo..)

mpc007, I'm going to have to respectfully disagree with you regarding CPU advancement. Lower TDP CPUs allow for laptops to either have significantly longer battery life at the same size/weight, or the same battery life with much lower size/weight. I remember when laptops that exceeded 2 hours of battery life were exceptional, but they did so by having huge and heavy batteries.

Performance is already there for me. I frequently use my HP Chromebook with a Haswell Pentium over my Asus notebook with an Ivy Bridge Core CPU because the Chromebook is lighter and has longer battery life with the same screen size. I definitely put a dollar value on battery life and size.

 

[Enigmoid]

Well, dual graphics MAY fail to deliver the promised performance in some cases, due to bad driver optimisation, but there are enough benchmarks which prove it's a promising concept, which actually delivers good power when playing certain games. For me, I'd just be glad to have the option to use DX. After all, having a IGP and dGPU which are almost identical and NOT be able to use them together, would be a waste of potential.

As of now (NBC review) dual graphics worked on two games. Even synthetics such as 3dmark saw no significant games. Its broken now and it was broken in the past. Furthermore, you need at least 40% higher framerates over a single gpu to get rid of stutter and the like. Gaining 10-20% fps from dual graphics will not give a better experience (and NBC notes this in other reviews).

Performance for Carrizo in Cinebench 11.5 range from 2.4 at 15W to 3.2 at 35W, i guess that it cover the whole i3/i5/i7 line..

You are left repeating what i said but in an apologist way, fact is that Intel use the 15W moniker for CPUs whose frequency is in a 1.6 ratio, that s a TDP range that is in a 2.5 ratio, so it s clear that their ratings are complete bs and are underspecced for most of the line.

Intel's CPUs have no trouble running the CPU only at 2.6-2.8 ghz @ 15W and 3.3 point CB11.5. Its when the igp gets used that things get rough.

Here is the lenovo T450.

15W long term average, 25W short term turbo.

Its a good strategy and I think its good AMD copied them.

However, u level chips can have the tdp set at different levels.

20W power usage and can maintain full turbo (2.5/900) under prime + furmark.

Now what AMD also needs to do is release a 45W laptop chip (carrizo should still have some scaling left at that power usage).

 

[mpc007]

mpc007, I'm going to have to respectfully disagree with you regarding CPU advancement. Lower TDP CPUs allow for laptops to either have significantly longer battery life at the same size/weight, or the same battery life with much lower size/weight. I remember when laptops that exceeded 2 hours of battery life were exceptional, but they did so by having huge and heavy batteries.

Performance is already there for me. I frequently use my HP Chromebook with a Haswell Pentium over my Asus notebook with an Ivy Bridge Core CPU because the Chromebook is lighter and has longer battery life with the same screen size. I definitely put a dollar value on battery life and size.

Basically, I think you agree with me. You are saying that longer battery life is nice, I say the same. But thats only provided the OEMs don't start mounting smaller and cheaper batteries, as soon as they can. And if you're saying your need for performance is satisfied with only an Ivy Bridge Core CPU (probably including intel hd graphics) you're not really the user looking for anything with more power, like Carrizo. You could just sell your other notebook, I'd guess, since todays lower end notebooks already perform close to Ivy Bridge Cores.

 

[frozentundra123456]

Not really, if it is one of the 35 watt ivy bridge cpus. Those are pretty close to current 15 watt U performance. For better or worse, intel hasnt so much increased cpu performance (except for the quad i7) as given the same performance in a lower power envelope.

And Carizzo at 15 watts is hardly a powerhouse. It isnt really any faster than ivy cpu wise, but better igpu, although Skylake may come close in a power constrained laptop.

Edit: My work laptop is a full power mobile, (i.e. not U model) sandy bridge i5, and it is plenty fast enough, although if there is a lot of IT crap running in the background it can drag at times.

 

[Abwx]

Intel's CPUs have no trouble running the CPU only at 2.6-2.8 ghz @ 15W and 3.3 point CB11.5. Its when the igp gets used that things get rough.

15W long term average, 25W short term turbo.

Its a good strategy and I think its good AMD copied them.

It isnt at 15W at 2.6-2.8, it is benched at 25W the time it takes to do Cinebench, this is obvious in the pic above, the cores are using up to 20.5W and that s without the uncore wich take 5W at least...

Anyway that s funny how you are eluding the fact that it s not 15W, hey, it will be 15W once the bench is over...

What you call a good strategy is actualy just a cheat from Intel to make their CPUs look better than what they really are.

Sure that AMD followed suite as they are not stupid to let their products being falsely outmatched thanks to these practices that you are, is it a surprise, branding as valuable, certainly they are for whom wants to mislead the public and get undue credits.

As for AMD they published all possible infos about Carrizo power usages while we have nothing from Intel, this say that their numbers are total BS, one who has valuable tech does not hide those numbers, quite the contrary, just look at AMD....

 

[Shivansps]

i can see "CURRENT CPU Package Power = 14.9W" clear as water.

 

[dahorns]

i can see "CURRENT CPU Package Power = 14.9W" clear as water.

*Abwx*

If you did the math instead of relying on measurements, you would realize the measurements are impossible.

*Abwx*

 

[Abwx]

i can see "CURRENT CPU Package Power = 14.9W" clear as water.

Yes, and it s enough for whom is of blind faith...

If 13W were enough at 2.8 then 17W would be more than enough at 3.2, yet at this latter frequency i read 20.5W, so how is this possible since this imply that at 2.8 the cores would drain 15.7W..?..

Or is it a FFT of different legnth that is used to get the final "good" numbers..?.

And where is the uncore, because that s only the cores power, the rest of the SoC take at least 50-70% of a what a core use...

And Enigmoid didnt link the review, here it is, with the device consuming as much as 40W, indeed he should had read the review before linking those pics :

The Core i5 can temporarily take in considerably more than the specified 15 watts, provided that certain parameters are fulfilled

http://www.notebookcheck.net/Lenovo-ThinkPad-T450-Ultrabook-Review.139988.0.html

Now if you can do yourself your homework and stop asking to others to do it for you it would surely help clean some muddy waters that you deem as clear, wich is possible after all, but not without a lot of wishfull thoughts..

 

[ZaphodBbrox]

It's all "urban legands"

 

[dahorns]

Yes, and it s enough for whom is of blind faith...

If 13W were enough at 2.8 then 17W would be more than enough at 3.2, yet at this latter frequency i read 20.5W, so how is this possible since this imply that at 2.8 the cores would drain 15.7W..?..

Or is it a FFT of different legnth that is used to get the final "good" numbers..?.

And where is the uncore, because that s only the cores power, the rest of the SoC take at least 50-70% of a what a core use...

And Enigmoid didnt link the review, here it is, with the device consuming as much as 40W, indeed he should had read the review before linking those pics :




http://www.notebookcheck.net/Lenovo-ThinkPad-T450-Ultrabook-Review.139988.0.html

Now if you can do yourself your homework and stop asking to others to do it for you it would surely help clean some muddy waters that you deem as clear, wich is possible after all, but not without a lot of wishfull thoughts..

Uh, the numbers seem perfectly consistent. The 40 watt number is for full system load.

According to the stress test numbers is cpu package power was 15 watts average and 25 watts peak. That is a 10 watt difference.

According to the system power numbers it peaked at 40 watts for 30 seconds before settling down at just below 30 watts. That is also 10 watt difference.

Seems pretty reasonable? It suggests the rest of the system uses ~15 watts.

 

[Abwx]

*Abwx*

If you did the math instead of relying on measurements, you would realize the measurements are impossible.

*Abwx*

Not at all, things are not difficult to sort out...

Measurements are rights generaly, in this case the laptop start the benches at 40W apparently before getting back to 30W.

I say apparently because the PSU is only 45W and that the laptop could well have started to extract power from the battery to not overload its DC input circuitry, you realize that in principle one should reject such a review from the start due to this element being incompatible with a reliable measurement of the power seen from the main.

Indeed the battery runtime suggest that on battery the laptop consume roughly 30W, wich is not compatibe with the 30W on the main as the PSU would add 4-4.5W at least, this add to the probability that it drain more than 40W on peaks but that the PSU is helped by the batteries.

 

[Shivansps]

Yes i know what that means exactly, it works like that since SB in both desktop and mobile.

It means the Bios has 2 TDP settings, TDP1 and TDP2, TDP1 is the normal TDP of what the cooling system can handle for sure and its always set to the maximum TDP of the installed processor, unless the OEM decided to cut corners and put a cheapper cooling system than what should go with the processor, OEMs has been doing this for a while(IE: set TDP1 to 29W instead of 35W for a mobile Pentium), TDP down is just a way to "legalice" that.

TDP2 is a "burst" tdp, meaning, the CPU and use up to TDP2 bios setting until the temperature reachs a certain level, generally 60-65°C, thats why you see 25W Package Power as a MAX, it means it went to that level until temperature got too hot and them went to TDP1.

TDP1 and TDP2 are checked against "CPU Package Power", there is no magic math in this.
If you even had a Intel K CPU and a Z motherboard and experimented with these TDP value on bios you whould known it.

Also Intel cant fake it, TDP value is what OEM use to calculale the apropiate cooling system, there whould be massive problems if Intel where faking TDP in any way.

 

[Abwx]

Uh, the numbers seem perfectly consistent. The 40 watt number is for full system load.

Yes, and how much use the rest of system assuming those 40W are genuine.?.

A TN 15.6" 768p screen use 4-5W, wifi will take 1W , RAM 2-3W, HDD 2W, what else.?..

Because we re barely at 10W, there s still a lot to find to explain the remaining 25W to get to 35W..

Ah, those physicals laws according to Intel s fairy tales, what a wonderfull world where 1 + 1 = 3 today after 4 yesterday...

And since you didnt read my post above here what the reviewer said, again :

The Core i5 can temporarily take in considerably more than the specified 15 watts

Lol.....

 

[dahorns]

Yes, and how much use the rest of system assuming those 40W are genuine.?.

A TN 15.6" 768p screen use 4-5W, wifi will take 1W , RAM 2-3W, HDD 2W, what else.?..

Given that we only need to find 15 watts, 10 plus PSU inefficiency has to be pretty close, right?

Ah, those physicals laws according to Intel s fairy tales, what a wonderfull world where 1 + 1 = 3 today after 4 yesterday...

And since you didnt read my post above here what the reviewer said, again :

Lol.....

Uh, the fairy tale based on actual measurements? You realize you are doing science backwards, right?

And I did read the review and what you wrote. 25 watts is considerably more than 15 watts. We already knew the processor could take in 25 watts for short periods of time (during a stress test mind you, not Cinebench). This isn't surprising.

 

[ZaphodBbrox]

It's all "urban legands"

Actually, it's "fairy tales" and "urban legends"!

 

[Enigmoid]

It isnt at 15W at 2.6-2.8, it is benched at 25W the time it takes to do Cinebench, this is obvious in the pic above, the cores are using up to 20.5W and that s without the uncore wich take 5W at least...

Anyway that s funny how you are eluding the fact that it s not 15W, hey, it will be 15W once the bench is over...

What you call a good strategy is actualy just a cheat from Intel to make their CPUs look better than what they really are.

Sure that AMD followed suite as they are not stupid to let their products being falsely outmatched thanks to these practices that you are, is it a surprise, branding as valuable, certainly they are for whom wants to mislead the public and get undue credits.

As for AMD they published all possible infos about Carrizo power usages while we have nothing from Intel, this say that their numbers are total BS, one who has valuable tech does not hide those numbers, quite the contrary, just look at AMD....

There really isn't much room to use more power.

I mean you are seeing Prime at 2.8 ghz and 15W. It seems prefectly reasonable that that same chip can run the much less demanding CB at 3.1 ghz at 15W. A bit of logic easily shows that the chip is capable of 2.8 ghz in CB at <15W.

And again package power is everything on the package including uncore, pch, memory, igp, and FIVR.