Also, I would not underestimate the impact of the overclocking community. Obviously OCers themselves account for a pretty small portion of CPU buyers, but OCers are also the one who guide the purchasing decisions of friends and family in many cases. To say that Intel and AMD "hardly care" whether they are the enthusiast favorite at any given time is a major understatement. There are many many benefits to being the leader in that segment outside of the actually number of CPUs sold directly to those users.
Before you get more off track here, let's review the posts...
1. The OP announced the L5320 and L5310...50w TDP Cloverdales.
2. Yellowbeard's reaction to the announcement of these CPUs was "It appears that Intel is in the process of releasing a CPU that can beat AMD in any catagory AMD wants to compete in".
3. I pointed out that he was comparing an (as yet) unreleased/untested Intel CPU to an older AMD CPU...
4. SexyK then avoided the topic by making a snide remark and making it about the entire current Intel line vs the entire current AMD line...and of course my remarks weren't about that at all. That's why I corrected him with my "apples to apples" comment.
Now you appear to be doing the same thing...can we stay on topic?
The comments were on the unreleased and untested chips Intel just announced...the L5320 and L5310.
At 1.6 and 1.86 GHz, I am dubious that they will be power performance competitive with AMD's LP versions of Barcelona. I say this because of the new power management that's being released with Barcelona...
That's the "apples to apples" comparison...
A valid point, but then we get back to what he was saying about power costs...
But weren't you the one saying that we should compare what is available (referring to FBD)?
Have a good listen to the flash audio interview here.
Barcelona will be able to instantly shut off any individual core that isn't required (no stepping needed), and they have isolated the On-die Northbridge clock from the cores as well, so that it runs at an independant voltage.
Overall, Barcelona should be one of the most efficient power/performance chips for real-world apps ever made...
- Aug 23, 2003
- 25,375
- 142
- 116
Such as? In Intel's implementations thus far, FB-DIMMs consume more power and offer less real-world bandwidth than cheaper, lower-power registered DDR2.
Intel typically jumps on newer technology earlier, while AMD waits for technology to mature and become cost-effective. It was the same situation for the DDR --> DDR2 transition. And now AMD will be able to tout lower costs and lower power consumption because their platform uses registered DDR2.
Here's another review:
http://www.gamepc.com/labs/view_content.asp?id=o2000&page=4
The tables turn and even a dual Clovertown system uses less power at idle and load than a dual Opteron system. So, power usage also depends on the motherboard and I'm sure AMD didn't do the Intel system any favours when it build it for the Techreport review.
And there's no point in repeating the old TDP generalization which may have been relevant in the P4 era, but not longer true for the P-M and Core processors, where power measurements show that if anything, it's Intel that has more conservatively rated TDPs.
- Aug 23, 2003
- 25,375
- 142
- 116
Do you even bother reading the articles you quote? It has nothing to do with the motherboard, and everything to do with enabling the power-saving features of the processor and chipset. As far as your accusation of "AMD didn't do the Intel system any favours when it build it for the Techreport review"...:roll: The Intel machine was provided by Hewlett Packard.
Read the first page of the Techreport review. The two systems were provided by AMD, Techreport pulled the Clovertowns from the HP system to test on the Intel system provided by AMD.
We accepted AMD's challenge, and they shipped us a pair of systems from server vendor Colfax International.
Fortunately, the quad-core Clovertown Xeons are intended to be drop-in replacements for dual-core Woodcrest Xeons, so we flashed the BIOS of the SuperMicro motherboard in our Colfax box to the latest revision, popped in the Clovertowns from the HP workstation, and were almost instantly in business.
- Aug 23, 2003
- 25,375
- 142
- 116
Here are the configuration details, right down to the BIOS revision and chipset driver version.
If you're going to accuse The Tech Report of using rigged computers in their articles, come back with some proof.
Yes and the Supermicro motherboard used for the Xeon system is better featured then the Tyan motherboard used for the Opteron system. Earlier Techreport article, different motherboard, same Woodcrest processors:
http://www.techreport.com/etc/2006q2/woodcrest/index.x?pg=9
And its 50W less on idle and load.
The FB-DIMM to Registered DDR2 isn't a directly comparable comparision as FB-DIMM has a richer feature set compared to Registered DDR2.
Intel jumps on technologies that it feels will enrich it's platforms and to help drive it to the mainstream.
Not quite as significant when your comparing a Pentium D 915/925 vs a Athlon 64x2 3600+/3800+.
Yes, but it's not a direct comparison, as FB-DIMM provide additional features over Registered DDR2, so what the platform provides in features is different. I am not saying not compare but keep this point in mind.
I am still curious if this will contribute enough to lower power consumption levels, we will have to see.
I am hardly doing anything remote close to the same thing.
You mean Clovertowns and we are discussing those processors so we are plenty on topic.
It is looking likely that these will come in before the release of Barcelona LP's, so they would have to be compared against the Santa Rosa HE's.
The initial release of Barcelona is of the standard power variants at 95W TDP up to 2.3GHZ, which should do well against the Xeon DP E5345.
Its a good step by Intel providing the processors which requires low electricity and also the cooling costs. I think all the comapnies should come with energy saving processors.
- Aug 23, 2003
- 25,375
- 142
- 116
Same Woodcrest processors, different configuration. The article you linked used 4GB of RAM (4 DIMM), while the article I linked used 8GB of RAM (8 DIMM). The difference in the amount of memory accounts for the difference in idle power consumption.
It's not some conspiracy.
I compared it with Techreport's number using 4 DIMMs for the Colfax Xeon system. Its Supermicro motherboard is better featured than the AMD Tyan motherboard and AMD is comparing the top-speed Xeon processors versus its second-fastest Opteron processor, even though it's much slower.
So while its possible to get an dual-socket Opteron system to have better power consumption than a Core-based Xeon system, it's a lot easier to get a lower power Xeon system due to the large power advantage the Core-based Xeons have.
- Aug 23, 2003
- 25,375
- 142
- 116
You keep reverting back to the power usage of the processor. The fact is that Opterons have better low-power states than Xeons...period. At idle, an Opteron can clock down to 1GHz, much lower than Xeons. Barcelona will improve on an already class-leading design, with individual power states for each core.
FB-DIMMs use significantly more power than registered DDR2. As long as Intel uses FB-DIMMs for their server platform, AMDs server platform will remain competitive on a power-usage standpoint.
And just because it clocks lower doesn't mean it use less power. The Techreport article using non-equivalent AMD assembled systems is not enough evidence to prove that.
The difference in power usage due to memory is not enough to overcome's Intel's CPU performance/watt advantage.
Not significant to who? For companies buying say 5000 systems for a new HQ, the power is much more significant than a once-off savings of a few dollars on a CPU...
A fair point, but the additional RAS is quite minimal...certainly of a much lower criteria than the power/performance savings (with the possible exception of some niche highly mission critical apps). And while FBDs do have the inherent design for larger memory, it seems that mem manufacturers haven't (yet) really been taking advantage of that fact...
I think this is what AMD is talking about when they say they will accomodate FBDs "when it makes sense to do so". Until we start seeing the MMs actually producing those 8GB sticks, the upsides seem to be outweighed by the downsides...
I'm surprised...I would have believed that you might have doubted the veracity of AMD's ability to manufacture it, but I never would have thought you would fail to see the degree to which this will change the power/performance curve.
First, thanks for the correction on Clovertons
I have seen it spelled both ways now, and a Rose by any other name...
We weren't discussing all Clovertowns, just the newly announced ones...the L5320 and L5310. So I stand by my point of comparing apples to apples...
On the rest, that sounds about right to me.
Pentium D 915 and 925 aren't critically higher energy consuming then 3600+/3800+ so that a corporation would overhaul their systems when they have been using Intel support and the processors give adequate performance for most tasks.
Well it would be up to the corporations to decide what is more important to them the features of FB-DIMM's of lower power consumption.
Oh, I don't doubt AMD's ability to manufacture Barcelona, though it will put a major strain on their production capacity and yields, as well as considering the price Intel will be charging for their Clovertowns not too long after Barcelona is released, and while AMD will further strengthen their position in the 4P+ space with Barcelona, Intel will finally release their update for the 4P+ space with Tigerton as well, which should be a substantial improvement over Tulsa, more so then Barcelona could provide against the current Santa Rosa cores.
Barcelona's improvements to power consumption only help when the system isn't doing very much and are in lighter workloads when you can turn off a cores such as single threaded applications.
Still I would like to see Barcelona's power improvements in action rather then rely on a bunch of paper specs.
Originally posted by: Viditor
First, thanks for the correction on Clovertons
I have seen it spelled both ways now, and a Rose by any other name...
We weren't discussing all Clovertowns, just the newly announced ones...the L5320 and L5310. So I stand by my point of comparing apples to apples...
On the rest, that sounds about right to me.
No my original point still stands as Clovertown is the far more prevalant spelling at Intel, do a search for Clovertown and you will find 21 hits on Intel's site with Cloveton recieving 3.
And I was comparing as apples to apples as possible Intel LP Quad Core's to what AMD's has in the LP DP segment.
Use Google as to what the difference as to Registered DDR2 vs FB-DIMM's are, but Intel doesn't adopt a more power hungry memory standard for absolutely no reason at all if it didn't provide features that Intel felt were useful to the server environment.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 22K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
Facebook
Twitter