The New AMD Building Thread

With a brand new technology, how can you calculate minimal?

 

Are you having a laugh? This is just starting to sound like denial...

 

Sorry to hear that. I can't deny that coding isn't up to date. I haven't denied anything. I'm very disappointed with the power hungry side of the bulldozer. However, I can't deny it's architecture, and it being considerably different enough, to effect how windows 7 and previous OS's handle it. Until I hear windows 7 is handling it properly, and I've seen a proper bench, I will laugh at current benchmarks. Remember, neutral party here. Fanboism is childish, and ignorant. I have to be objective, it's in my nature.

Fact of the matter is, if it were running properly, it's very likely more power hungry...

 

Let's be honest, blindly hoping that a product is going to suddenly massively improve, with absolutely no scientific basis to justify this, is not objective.

 

That's just it man, there is basis LOL! I guess time will tell man. And when I'm proven wrong, you can call me an idiot for blindly hoping. Which i'm not. I simply see a potential, that isn't being tapped. ANd I'm not expecting huge performance increases. I've stated that. I would however expect nothing less than 5 - 10%.

 

Oman7,

And lookie here, who we got making the claim. My old friends at Bit-Tech. I knew the 586w was incorrect the minute I saw it, and was nice enough to figure it was a typo. 586W is impossible unless the toaster is plugged into the computer!

I'm not sure who said it, but I think it was Sam that said you would need high end water cooling to deal with the heat, and that a 120mm radiator wasn't enough to handle the heat. Either AMD have come up with a super heat dissipating Radiator, or those wattage's are incorrect. In fact my whole computer only draws 228w at 3.6GHz while running IBT for 5 passes, all cores at 100% load. I was also suspicious of the wattage for the 1100T. As it happens Bit-Tech ran their 1100T at the same speed I ran my 1090T at, and I was most curious as to why it took over 1.50v to reach 4.2GHz since my 1090T took just 1.394v to reach 4.2GHz. Another reason I know that the wattages for the two AMDs are wrong is that at 4.2GHz, my CPU and NorthBridge temps only went up a couple of degrees, and only hit 53C with 5 passes of IBT. After seeing the much higher CPU voltage, it's not hard to imagine where all the excess wattage came from, and why it would overwhelm an 120mm radiator if it was in a case!

I looked at a large number of high quality 750w PSUs, and even with 4 40 Amp rails, the biggest load a 750w PSU +12v can handle is 62w. At 586w, Bulldozer would be drawing 48.8 amps. on my 1090T at the wattage stated for the 1100T by Bit-Tech, which is fundamentally identical to the 1090T, you get 41.5 Amps. That's considerably more than the 37.5 Amps than my PSU is capable of producing with 2 25 Amp rails (450w Max). Those numbers for the 8150 and the 1100T are either fakes, or a mistake. I'm not going to speculate on which! Suffice to say they are incorrect! Another reason is the i7 920 has always run hotter in every review I read on the 1090T before I bought it. The i7-920 is always the most power hungry and runs the hottest. I really think that this review will prove that I am right! Here's Bit-Tech's own review of the 1100T and the 1090T last December 10.
http://www.bit-tech.net/hardware/cpus/2010/12/07/amd-phenom-ii-x6-1100t-review/7

Oman, I can assure you that the wattages Sam posted are not correct! The AMDs were also tested on an AM3 motherboard, not an AM3+, which makes a big difference in how much voltage you need to overclock, and how much further you can OC it on the AM3+ at a lower voltage. It still doesn't jive with what Bit-Tech posted that Sam just linked.
to! The defense rests! :<)

Best Regards,
Russ

 

Well, the wattages do seem extreme, and perhaps they were exaggerated. But it is an 8 core cpu. And when overclocked, I guess it isn't surprising that the power demand should be great.
I really wonder if it needs a high voltage for overclocking though. I've read posts in the past, that suggest cooling is better than overvolting. My 965 in particular becomes unstable when overvolting. Although the northbridge did want more voltage.
And if it is true, that this beast of a CPU requires serious power, I'm sure there are boards that can handle it, for people that are AMD diehards. Of course very good cooling will still be pertinent!

 

How about this one then Russ?


This is a slightly different test to that used by Bit-tech, so it uses lower power all round - as proven by the lower i5/i7 tests - it's worth noting, [H] could only get their sample to 4.6Ghz, because their Corsair H100 wasn't up to the job of cooling it any further - Bit-tech aren't lying (neither are they about the performance tests), the CPU is an absolute dog when it comes to power consumption.

Bit-tech are also using an AM3+ board (Crosshair 5), not AM3, as you like to keep pretending they use.

 

As you well know Russ, all tests are carried out at the mains outlet, so this value includes the losses taken by the PSU in stepping down the voltage, as well as the load used by other components - you'll notice I said the CPU used 450W, not 586W - as that takes into account what part of that 586 the CPU is using. At 450W, this is 37.5A - well within the means of most single-rail units. It could prove problematic for certain modular units of course, but usually when you have 4+ rails, the CPU power is split between the first two rails (one 4-pin to each rail).

 

sammorris wrote:

 

Why are you looking at old reviews? The X6 1100T has nothing to do with this, we're discussing the power consumption of Bulldozer here, which was tested in an AM3+ board. When the X6 CPUs were tested, AM3+ board didn't exist, so give them some credit for goodness sake.
Also, the argument about the white vs. black socket is irrelevant, because plenty of AM3 boards can support AM3+ CPUs with a BIOS update.
The defense rests...

 

Sam,

This all started with the post from Shaff and the Jan. 3 Bit-Tech review, that's why. My specific complaint was that the AMD CPUs supposedly use so much wattage. I was looking because I wanted to see how the 1100T tested out in one of Bit-Tech's reviews. The Dec. 7, 2010 article on the 1100T was all I could find. To see why so much difference between Shaff's Bit-Tech review and the one you got your chart from is disturbing to say the least because it takes less voltage to run stock or overclock on the AM3+ motherboard. I've found that to be the case with all all 6 CPUs I've tried in my computer so far

As it happens, I know exactly how much wattage my 1090T CPU consumes, and exactly how much wattage each individual component draws, it makes that wattage list you provided very suspect. I do not believe 586 watts for Bulldozer, and I do not believe the 498 watts for the 1100T.

I am aware that certain socket AM3 motherboards can be flashed to allow Bulldozer. It will primarily be the higher priced motherboards though with more adjustments. It also depends on the bios as well. I had 2 Callisto's, both 555s. One ran fine at stock speeds, but wouldn't turn on the other two cores, and the other one let you turn on the cores and then clocked to 4.4GHz @1.40v. It comes up in CPUz as a B55, but it's actually a Phenom II 955BE. The 940 Propus 3GHz Quad ran at 4.0GHz pretty easy for a locked multiplier CPU. All benefited from the AM3+ chips. I'm beginning to wonder if we aren't seeing history repeat itself.

Remember when GigaByte had their 785G motherboards figured out, months before Asus, Asrock, BioStar, and MSI. Now GigaByte has some serious AM3+ motherboards. The GigaByte GA-990FXA-UD5 AM3+ is amazing. The GigaByte GA-990FXA-UD7 AM3+ is a Tour De Force. I think GigaByte is getting more serious about product development and R&D.

Sorry I got distracted there for a moment. I know we've had avoidable differences on both sides on this subject, and I see many ways to avoid confrontations in the future. It's easy to see now that our articles were months apart in what we posted, but maybe in the future if we just try and give each other a little more information we could avoid mistakes and misunderstandings.

Russ

 

Because, for the 10th time, that's full system draw, not the CPU draw by itself. Applying 90% efficiency on their PSU (and it will do that on a UK voltage), subtracting the rest of the system draw we get 448W for Bulldozer, and 369W for the Phenom II X6.
The i5s run about 70W at full load when they're stock, and about 200W when they're overclocked, so when you clock the pants off them they use between 2 1/2 and 3 times as much power as they normally do. This because, fortunately, they don't require too much of a voltage increase to overclock well.
Meanwhile, the Phenom II X6 uses 130W at stock, so if we apply the same 2 1/2 to 3 times ratio to that, we get 325-390W, which their test falls within range of. The Bulldozer chip is closer to 140W at stock, so applying the same logic, we get 350 to 420W. Not quite the 448W figure, but when we consider that to get that 4.8Ghz an enormous voltage was used, it makes sense. I believe the Bit-tech test, and you're not going to change that no matter how many times you call it false.

For what it's worth, my PC's i5 750 (another efficient CPU that only pulls about 70W at full load normally) - runs at 165W idle, from the wall, because of the pair of graphics cards, and its considerable overclock for which I have raised the Vcore. At stock it's much less, more like 120. When I ran IntelBurnTest with it stock, the power usage at the wall went up from 120 to around 175. When I ran it at the full overclock, it jumped to 335. Considering that my CPU uses half as much power as Bulldozer when it's at stock, I have no trouble believing Bit-tech's 586W figure.

 

1100t vs fx-6100. I'd have to say these two are fairly close to each other. If updated coding can increase the Fx-6100 5 - 10%, it just may overtake it in all benchmarks eh? It's certainly less power hungry than 1100t at 3.3ghz. And it actually wins in a few benchmarks.
http://www.techspot.com/review/452-amd-bulldozer-fx-cpus/page11.html

 

So a brand new 3.6Ghz 8-core processor can get 'fairly close to' the 3.2Ghz 6-core processor it was meant to replace? This is why the Phenom II X6 is better than Bulldozer - it's faster, slightly less power hungry, and much cheaper, never mind the fact that because it only uses 6 cores, its performance is better still in applications that aren't fully multithreaded.

Also, the power consumption graph there is stock only - the early i7s lose out big time there because they too are 140W CPUs like Bulldozer, but the X58 chipset is a power hog as well. Newer i7s (and all i5s) don't have a chipset because it's built into the CPU.

 

The 6100 is a six core processor. That's why I'm curious how the 1100/1090t stack against 6100. The fx-6100 should be the successor to 1100t. In my opinion, it isn't yet because of coding. 1100t is to AM3, as 6100 is to AM3+.

1100t may be faster now, but given time, I feel 6100 will win out. On the other hand, I may be full of sheet LOL!

 

The FX-6100 will be a lot more interesting than the FX-8150 as a cheaper CPU. $190 is still a hefty price tag though, one which puts it head to head with the i5 2500K and X6 1100T. Currently, the 3.3Ghz X6 1100T is 6% faster than the 3.3Ghz FX-6100 at x264 encoding, so 6v6 is a lot closer (6% per-core basis) than 8v6 (17% per-core basis). Then again x264 is what Bulldozer is best at out of all the tests, so they may not fare so well overall. Even at a whopping 4.2Ghz, the FX-4170 is smashed by the X4 980, which despite only running at 3.8Ghz, is a full 17% faster. The i5 2500K takes an astonishing 61% lead over the FX-4170.

 

Indeed, I'll be watching reports about 6100 for some time.

I don't recall anyone every saying "Holy cow" to the i7 extreme edition, as far as power requirement. If there are stable boards for that processor, isn't it logical to assume, there will be good boards for Bulldozer 8 core as well? No doubt the 8 core is gonna inspire board makers boost potential. As well as power supplies. But then, if the 8 core beast doesn't sell much, perhaps not eh?
Perhaps AMD should have concentrated on Dye shrink. Actually I think they are. The question is, will it be intel to release 22nm first, or AMD. Mmmm, a 22nm bulldozer. THAT could be interesting. I believe they'll be calling that Steamroller though.

 

Bulldozer was a dieshrink - 45 to 32nm.

 

I meant more concentration on shrinking. 22nm - 28nm. But I'm sure they can get lower volts on the current technology. I'm sure new releases will do just that.