![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
AMD Barcelona Official Pricing
- Thread starter annaconda
- Start date
pliskin_wolf
n00b
- Joined
- Aug 27, 2007
- Messages
- 30
im not too familiar with how quad works so could someone explain how the ghz work since 2.0 is the max comparied to a amd 6000 just something to help me decide what im going to do. thanks
you'll have to wait for benchmarks. But, quads work by putting four processors on one die.
1.7 ghz means that all 4 cores run at 1.7 ghz.
pliskin_wolf
n00b
- Joined
- Aug 27, 2007
- Messages
- 30
i guess what im trying to say is how do the Ghz work on a quad core compared to a duel core, and if you did a octy set-up how that would be taken into effect all in all would it be better to have 2 1.8 quad or 2 3.0 duel
pliskin_wolf
n00b
- Joined
- Aug 27, 2007
- Messages
- 30
so would that be like 1.7 ghzx 4 or just 1.7 4 ways?
What is the difference between the two? No it does not run at 6.8 GHz.
Wor.d
On a single application (assuming it's not multi threaded) you will see 1.7ghz of usage or power. On a multiple application (or a multithread), you will see 6.8ghz of work.
So yes it will be more powerful and no it wont as it depends highly on the application and what you want it to do.
here is a book you should read
http://davidguy.brinkster.net/computer/
sorry could'nt help myself. but ill attempt to answer your question
it depends on what you want to do? IE a quad core at 1.7 vs a dual at 3.0 (assuming all other feature are equal) if your a gamer the 3.0 is the best bet. But if using multi treaded applications like graphics rendering etc. or you are running many single treaded applications at the same time. IE downloading while burning a DVD, listening to MP3s and gaming all at the same time, then the quad will suit your needs better. Also more and more software will in the future be writen with multi treaded code.
Your question is like asking, what is better four computers that are good or two computers that are better. If you have four people all needing a PC then it is better to have four computers rather than have four people sharing 2 PCs.
but if you want the best of both just wait until quad cores reach the same speed as the dual cores.
Even better wait until the confirmed benchmarks come out for the new chips and see which is faster in certain applications that you use.
For future applications multi core CPUs will be the norm 2,4,8,16,32,64 and so on. Because increasing the core speed much past 3.5ghz (at 90nm 65nm) is very in-efficient due to the amount of heat produced and power used.
- Joined
- Jun 3, 2003
- Messages
- 6,677
95W for the 2 GHz model? Ow.
I'm glad the prices are rather competitive though.
I'm glad the prices are rather competitive though.
It's a quad-core.
This is an incredibly perverse way of putting it. Amdahl's law aside, you might see a linear increase in computational power with number of cores, which might be roughly similar to an increase in clock frequency of the same number, but describing it as "6.8GHz" is pretty inaccurate.
I don't agree with your argument. I was simply putting it in lamens terms.
Also hence where I said "of work", not speed.
- Joined
- Jun 3, 2003
- Messages
- 6,677
Oh really? I had no idea AMD was producing a quad-core CPU at all. I just must be out of the loop.
In all seriousness, 95 W is disappointing to me because Intel was able to achieve that same TDP with a MCM. After all the bragging AMD has done about energy efficiency, about their process, and especially after that whole oven mitt thing with Prescott, I expected better.
Right now, it's starting to look as though AMD is releasing a "native quad-core" processor with at best equal performance and thermal envelopes to Intel's "backwards hack job" of a design. After all the energy AMD put into trying to convince me that the native approach was ten times better, I'm rightfully disappointed.
I sincerely hope I'm proven wrong when the NDA lifts.
Run cooler, yes. Overclock? You realize these are two- and eight-way server processors, right? There aren't that many s1207 boards that let you OC. May as well try and OC an Itanium.
That said, I hope the AM2+ counterparts are priced similarly or better. I'd love to put one of these in my 690G HTPC.
ABCD
A= # of sockets/HT links
B= Revision?
C/D= Family & Rank (Think Opteron 144 vs 148)
Was Opteron 144/250/865 any better?
I'm more amused by the "High Efficiency" drawing "only" 68W. Shall we expect an EE with 50W later?
- Joined
- Jun 3, 2003
- Messages
- 6,677
These are server processors. Phenoms will likely recieve the more familiar PR rating.
Or not. Who knows?
Thier using the B1 core at the start which sucks. It uses too much power for the clock speeds. Wait for the B2 core thats the sweet spot as it uses much less power.
- Joined
- Dec 9, 2000
- Messages
- 16,338
yea but this is the CPU + memory controller in 1 95w package =p, you can compare it to the 95w Intel + xxw generated by the Northbridge, or at least mem controller on the north bridge
I thought it was pretty good myself, at least the HE ones were.
Personally I am not too impressed by the whole "but it is native" approach. I also hope the NDA-lift proves us wrong. For the price I think AMD did a good thing, but Intel is already low at price, they can simply drop a few bucks and they are on the same doorstep or even undercut them a little.
Lou Ceifer
[H]Lite
- Joined
- Jan 27, 2003
- Messages
- 88
And add to the fact Intel's TDP is mean average thermal dissipation, AMD's TDP is rated at MAXIMUM thermal dissipation. The actual TDP on average for AMD could very likely be far less than 95 watts, this is also taking into account it has an on-die memory controller while Intel chips do not.
Also, we have yet to see how the K10's split power planes across it's cores and AMD's improved Cool n' Quiet come into play.
I don't think this is 100% valid. Note that AMD has had 35W EE procs for a while on AM2. Even assuming 50% of that is for the memory controller (which we all know is a GROSS overestimation), your DDR2 controller is 17W. Using similar numbers, the 95W "regular" Barcelona still needs almost 80W for just the processor die, not including the IMC.
- Joined
- Dec 9, 2000
- Messages
- 16,338
so 80w for a new QC chip is bad
? I guess the Perf/w will tell it all in the end ^__^I'd say yes, when the 4000+ (also 65nm and 2GHz) is 65W including that same "17W" memory controller.so 80w for a new QC chip is bad
Realistically, let's say it's only 7W- probably closer to the truth. Then they went from a 58W dual-core die to an 88W quad-core die, or just over 50% extra. By contrast, Intel went from a 65W e6600 to a 105W q6600, or roughly 60% of the second die. Considering that Intel's design requires some extra circuitry to communicate and synchronize between the two dies, there's your 10%. How is AMD's native quad more energy-efficient, exactly?
debello_64
Gawd
- Joined
- Oct 18, 2004
- Messages
- 672
Source?
- Joined
- Dec 9, 2000
- Messages
- 16,338
Well I don't know what 4000+ is 65nm 65w and 2ghz
but the X2 3600 is 65/65/1.9 you can compare that, dual core vs quad, 65-15 (make it a bit easier) = 60w x 2 = 120w + 15 ~ 135w, while the Barc equiv is 95w thats a considerable drop this is also the cpu at full tilt, not idle, not half or 3/4 load which will all come into play in the server field no?Also AMDs proclaimed power states will probably make a huge diff in energy-efficiency. and if the AMD chip out performs the same clocked Intel chip, that will play even more into the Perf Per Watt
- Joined
- Jun 3, 2003
- Messages
- 6,677
Power states mean nothing to those who fold on all cores, or those who disable such things to overclock. That covers a *lot* of people here.
- Joined
- Dec 9, 2000
- Messages
- 16,338
I'm sure alot of servers are going to "Overclock" and disable power saving features, aswell as fold 24/7. ;p
datacenters which have 100 servers running will be happy to save 25W on each machine for the same or more performance, thats 2500w if they were running at full throttle.
Sorry, I meant 3800. This 2.1GHz-with-10.5x-multi BS is irritating. Not a huge difference, given that they are all 65W between the 3600 @ 1.9GHz and the 5000+ at 2600.Well I don't know what 4000+ is 65nm 65w and 2ghz
Also AMDs proclaimed power states will probably make a huge diff in energy-efficiency. and if the AMD chip out performs the same clocked Intel chip, that will play even more into the Perf Per Watt
Although you want to re-check your math.
Shall we say we're even about the 3800/4000 comment? More importantly, your 1+1 energy equation is inaccurate- that's why I brought up the Intel numbers earlier. e6600+e6600 = q6600, but 65W + 65W is not 105W (not G0, to keep it fair). As I said, Intel's dual-to-quad pulled an extra 60% in wattage, where AMD's is drawing 50%. I attribute that extra 10% to Intel's two-dies-in-one approach, and nothing else.
- Joined
- Dec 9, 2000
- Messages
- 16,338
Sorry, I meant 3800. This 2.1GHz-with-10.5x-multi BS is irritating. Not a huge difference, given that they are all 65W between the 3600 @ 1.9GHz and the 5000+ at 2600.
Although you want to re-check your math.
More importantly, your 1+1 energy equation is inaccurate- that's why I brought up the Intel numbers earlier. e6600+e6600 = q6600, but 65W + 65W is not 105W (not G0, to keep it fair). As I said, Intel's dual-to-quad pulled an 60% in wattage, where AMD's is drawing 50%. I attribute that extra 10% to Intel's two-dies-in-one approach, and nothing else.
yea my bad on the math =) but still my point is, AMDs jump isn't much different =p, I think the bonus of native QC wasn't the energy efficiency but the loss in performance
I think I heard 1H '08. Can't wait, myself. As long as it isn't SLOWER than X2 clock-for-clock, I'm upgrading as soon as they're affordable.
InorganicMatter
[H]F Junkie
- Joined
- Oct 19, 2004
- Messages
- 15,461
Prices seem a tad high, even if it is quad core. Especially considering the preliminary performance reports.
- Joined
- Dec 9, 2000
- Messages
- 16,338
Virtually all of them can be OC to a degree. My main system runs a SuperMicro H8DCi with two Opteron 280 processors (2.4GHz) OC to 2.8GHz. That's a 400MHz OC with stock voltage. How is this accomplished? Through the employment of software OC tools such as CrystalCPUID or ClockGen, which is what I use. There are other apps that can OC server boards as well.
In the Intel camp, you can 'pin-mod' Xeon processors to utilize higher bus speeds the processors don't support natively, and some boards have FSB adjustments, FWIU, but I don't know much about the new Intel socket-771 platform to make knowledgeable comments based on personal experience - yet.
I only run high end workstation/server boards and have been for the better part of a decade. I will refuse to invest in a new multi-socket platform that does not possess the potential of at a least mild OC, through either hardware or preferably by software. My Tyan Thunder K7 server board, although 6 years old now, still runs OC dual Athlon 2600+ XP-M processors @~2.3Ghz folding 24/7. This was done through a combination of pin-modding and bridge-modding, as many people on this forum will recall. What works for single processor systems, should also work for SMP, to an extent.
Where there's a will, there's definitely a way.
Sooner, Q4’07 for the Phenom desktop processors unless something changes.
InorganicMatter
[H]F Junkie
- Joined
- Oct 19, 2004
- Messages
- 15,461
Let's hope it's the third suggestion, but considering the monkey who is running AMD right now, I wouldn't be surprised if it's option one.