Printable View
well TDP is usually for the highest rated chip in the line (for AMD at least).. by setting such a high number, AMD has a lot of room to grow before running into heat "problems"
assuming between 85 and 90% efficiency, AMD can go up to 3ghz with the venice core before going over the 67w limit.. for this particular chip. problem is, not all chips are the same. some are less efficient than others and put out more heat.
I guess that's sometimes true, especially for older XP chips and mobile A64. They want to make some headroom for future models. But now we see strange numbers like the 3500+D I mentioned. They probably said 67 W because they had a faster revision D in mind but it never showed up.Quote:
Originally Posted by ozzimark
Another story is Opteron E. AFAIK, the 2.4 GHz and the 2.6 GHz showed up at the same time. The 2.4 GHz have 85.3 W. Now is that high number there to get some headroom for the next model? Well no, the 2.6 is already there too and got 92.6 W. Very strange.
yepp, the tdp is the rating for the mobo and heatsink builders, it means that the mainboards and heatsinks have to be able to supply the cpu with that much electricity and be able to handle the heat comming from the cpu, thats all :)
sometimes i wonder if that opteron number includes dual core, because it is definitly very high :confused:
just figure i would post again with prob my last major update for this turion. nothing really new, cpu seems to run everything i throw at it at 2620 @ 1.65. anything higher and it will fail 3dmark2003.(prime95 blend/fft,super pi 32m, and 3dmark2001/2005 are fine higher at around 2670) thought this was a power supply problem so tried a 500w ttgi (not sure if it is better than my antec but figured i would try it) with a 30a 12v. (my antec has only 24a) still fine at 2620 but problems at 2630.(same as my antec) really don't want to throw anymore voltage at it. (even though it is still running only around 40c under load after taking off my smartfan and just using the 22db silentcat) i am going to run it 24/7 with 1.65v at 2600. (with my fsb, about as fast as a stock fx55)
i will contine to monitor and comment on the 90nm mobile chips though. hopefully those chips will turn out to be what i/we thought these turions would turn out to be. (e core san diegos on s754)
Mats, from your diagram a 3500+ Winny would be closer to 48w than 42w, where did you get the 42w figure? And a Rev.E would be bout 56w... all this is on standard voltage & standard HTT... raising HTT and memory speed must also increase power consumption.
Either way, I dont think these Turions are Rev.E procs so maybe they consume even less power than a Winny... who know?
tristancarton, when 3Dmark starts crashing could it be related to memory (speed or timings)?
that's measured from the power plug, thus you must apply the inefficiency of the PWM to the draw. assuming 90%, the actual power draw is ~42w :)
And 90% efficiency is probably a fairly optimistic figure for a consumer board. From what I've read, server/industrial boards are generally the only ones with 90%+ efficiency, and generally the consumer motherboards are between 80-85%, of course depending on the quality of components used by the manufacturer and what kind of load they are experiencing (i.e. if they're anything like PSUs efficiency is lower with extremely high or extremely low load demands).
Thanks for the update tristancarton. I got the Turion up to ~2.55GHz at 1.6v, but didn't bother testing more than a couple hours of P95. Will probably try to find a nice middle ground at 1.3-1.4v.
Rabbi NZ:
How I first got the 42 W figure: Calculated from a 1.35 V, 35 W, 3000+ Oakville running at 1.40 V, 2200 MHz which gave me 41.4 W TDP for a 3500+D. We get 41.4/48 = 86.25 % efficency for that EPoX 9NPA+ Ultra mobo in the review.
Remember that some A64E runs at 1.35 V (not the one in the review), which almost makes up for the extra power needed compared to D.
For a 3500+E(1.4 V): 86.25% * 55.9 = 48.2 W
For a 3500+E(1.35 V): 44.8 W
But if the efficency is lower like ff says, then it makes the CPU even more low power. I guess they rather optimize the mobo power regulators efficency for max output (X2 or 130 nm FX) rather than a 3500+...
Turion is more low power than A64D, since they do 2.4 GHz at 35 W and 1.35 V (2.2 GHz at 25 W and 1.2 V). Oakville did 2.0 GHz at 35 W and 1.35 V.
TDP is valid for a certain CPU (or groups of CPU's, present or future) running at stock speed and stock Vcore, nothing else. If you raise the HTT or mem speed then you're going out of specs, but we all know that...
You should be able to accurately calculate any revision E Turion wattage using the Benchtest CPU calc, because Turion has a very defined thermal limit, and is not as open ended as desktop models. Just plug in max speed for the 25w line (2200MHz), voltage (1.2), and wattage (25w) as the known variables, then calculate your unknowns. To verify the accuracy of the calculator, I plugged in 2400MHz and 1.35v (which is the highest 35w Turion) and sure enough wattage came out to 34.52w, so I'd say the calculator is pretty reliable.
So when I plug 1.55v and 2.4GHz into the calc with the Turion knowns I only get 45.5w. Apparently voltage has very little to do with power consumption on K8, which makes sense, because higher clocks seem to raise temps for me much more than higher voltage (was the same with my Oakville, and to some extent my Newcastle).
Even at 3GHz with 1.6v according to the calculator it's only using 60.6w.
Not sure if these figures can be carried over to the desktop chips, though, because it's possible Turion does use different transistors with different power characteristics at the same settings.
If only I had a clamp-style pickup with my multimeter I would be more than happy to test current on my +12v and either back up or shoot down these claims. But I don't, and I'm not about to hack up my PSU wires to measure amperage in series. And the cheapest DVOMs I've seen with inductive pickups were still pretty expensive.
LOL, so strange...Quote:
Originally Posted by ozzimark
single core 2.4 GHz: 85,3 W :stick:
single core 2.6 GHz: 92,6 W :stick:
dual core 1.8-2.2 GHz: 95 W Ok, I believe this one, (VERY close to 2*48.2 :eek: from my last post)
I use CPUPower for calculating this.Quote:
Originally Posted by fr0stedflakes
No, it's not like that. It's the same for all CPU's. Frequency affects the wattage less than Vcore. Try calculate 3 GHz at stock Vcore and you'll see.Quote:
Originally Posted by fr0stedflakes
Pnew = Pold*(Vnew / Vold)²
Pnew = Pold*(fnew / fold)
....IIRC.
Will motherboard makers post us from using turions in desktop boards.
Thanks
Yeah I guess your right, somehow I think I knew that in the back of my head I just got my thoughts mixed up when I wrote them out. :confused:
Anyways that CPUPower program is slick. Guess I can go ahead and take the benchtest calculator off my bookmarks. :)
What do you mean? :confused:Quote:
Originally Posted by elec999
If you're lucky, your mobo will POST if that's what you mean...
Or is this a Soviet Russia joke, the mobo maker post you?
Interesting....
http://img141.echo.cx/img141/6104/turione9tr.jpg
DH9E0 hmmmm...
Interesting, CBI picks up the revision information.
If someone with a desktop San Diego wouldn't mind checking their revision info with CBI maybe we would be able to determine if these are actually SDs or a different type of beast. I'm guessing if they are just SD cores stuck on a 754pin package revisions would be exactly the same (much like Oakville and Winchester, both DH8-D0).
Also kind of wierd that it detects hyperthreading. Has to be a glitch.
well the HT bit was enabled on the dual cores... maybe amd decided to put it in all the revE cores and we just didn't notice before? :confused:
I dunno, but would that show up in the task manager? I just didn't think they were capable of running two threads.Quote:
Originally Posted by ozzimark
how'd amd get rights to hyperthreading???
i don't think intel can do anything about amd turning on a bit in the CPU-ID ;)
mats, i have no idea. but the HT bit has been enabled since the early northwood days.. before it had HT (not 100% sure, but i believe so. check the toledo thread)
AMD turning on a HT bit ?? What HT bit? Do I miss something ?Quote:
Originally Posted by ozzimark
K8 arch. certainly isn't capable of running HT because of pipe-depth
Multithreading is what AMD can easily do but not HT ...
It is an obvious misdetection ;)
vandi, nice eye. ff, yeah very intriguing. i am going to try and follow up on the cbid san diego revision codes because i really do want to know what the deal is with the turion line as well as the new 90nm mobile line that will be forthcoming. i posted on dfi-street asking rg or ag if they could please download the new cbid and post a ss of their 3700 and 4000 san diego cores. assuming they do that i will update here.
as to the hyperthreading, i am sure it is an error. (or perhaps amd having some fun)
also @ i found nemo. i think amd and intel have rights to each others property and patents for the next few years from an 10-15 year old settlement.(not that amd would use ht. just that they were they could use it without having to pay royalties or asking intel for approval if they wanted to.)
update: ag has already gotten back to my request. generation is 15.0 model is 7.2. revision is SH9E4. don't think the turion is based on san diego. (now the wait for the 90nm new mobiles)
AMN3200BKX5BU
AMN3400BKX5BU
AMN3700BKX5BU
AMN3000BKX5BU
You mean these are available now? As in ready to ship?
Sign me up for that 3400+ if the price is still between $250-275. I want another shot at these new CPUs....
tristancarton, looks like Turion are a different version of the San Deigo desktop core. Thanks for the confirmation, although I think were sure enough about this based on our poor results. :(