- Sep 28, 2005
- 21,044
- 3,524
- 126
ROFL!
which is why im scratching my head on the comparison.
Sorry guys, deneb will get gobbled by i7. Dont need any reviews to tell you that.
Deneb is not redesigned. Its refreshed with optimization.
i7 is a new platform, and ... well cant tell you guys anymore then that but its HOT.
Very good point, if we all ran stock.
Doesnt apply to the 80% people that post in this section. Once you start overclocking, its more a slaughter then a comparison.
Someone stated that a 3Ghz Phenom should be similar in performance to a 2.5Ghz Deneb. So to get a clock for clock comparison, I looked at a 3Ghz Phenom vs a Q6600 which is the closest to 2.5ghz of the Core 2s. The results make it look like Deneb won't catch Yorkfield, as it barely matches Kentsfield, assuming that the 20% clock per clock performance improvement is correct.
If a near/fully overclocked Phenom cant beat a stock clock Q6600(which is well known for its overclocking ability),AMD will be going bankrupt soon consider they still price them at a similar price and that's already AMD's top of the Quad line.I do hope their Deneb lines will be catching up to Yorkfield(or least Kentsfield) clock for clock given that they dont come up with another TLB Bug issues.
Agreed, in the grand scheme of things overclockers are like a drop in the ocean, its performance/price that will dictate how Deneb compares to i7 (and Yorkfield of course).
Comparison of a current phenom and yorkfield in attempt to find comparison to i7 and deneb is pointless. Kentsfield is still difficult to compare as the cache numbers and setup is different, 8MB on a Q6600 compared to the 2MB on 9850.
first of all the number of threads on the i7 is completely useless unless you use multiple applications that use more than one thread at the same time. mostly hard core enthusiast and pecific applications that take advantage of hyperthreading will have any use for this.
Deneb will be getting some pretty good architectural enhancements for the 45nm switch. adding much more cache and enhancing the hypertransport for ddr2 also.
I believe that Deneb will be better than yorkfield. No it will not equal i7 for many operations but depending on programs yorkfield/deneb will beat i7 in some places also.
If you don?t realize that a second gen 45nm architecture is going to do better than a the first 45nm switch you need some help. The point is to have something that is going to be a good price to performance ratio getting closer to competing on the higher end than they currently are. There is a good chance that deneb architecture will not be fully enhanced until the ddr3 switch, so don?t hold your breath on the first release on January 8th.
So in the end just wait for the final release benchmarks before making an informed decision given your budget options.
Everyone should hope AMD stays in the game with at least mid grade CPU's and high end GPU's or your system will triple equivalent price in the future because nobody else will be able to fill the competition gap for many years if they die off.
After saying that I think i7 will be totally awesome in the next 1-2 years as increasing multi threaded apps (more than 2-3 threads) become the norm.
Thank you for reading through all my crap =D
first of all the number of threads on the i7 is completely useless unless you use multiple applications that use more than one thread at the same time. mostly hard core enthusiast and pecific applications that take advantage of hyperthreading will have any use for this.
Deneb will be getting some pretty good architectural enhancements for the 45nm switch. adding much more cache and enhancing the hypertransport for ddr2 also.
I believe that Deneb will be better than yorkfield. No it will not equal i7 for many operations but depending on programs yorkfield/deneb will beat i7 in some places also.
If you don?t realize that a second gen 45nm architecture is going to do better than a the first 45nm switch you need some help. The point is to have something that is going to be a good price to performance ratio getting closer to competing on the higher end than they currently are. There is a good chance that deneb architecture will not be fully enhanced until the ddr3 switch, so don?t hold your breath on the first release on January 8th.
So in the end just wait for the final release benchmarks before making an informed decision given your budget options.
Everyone should hope AMD stays in the game with at least mid grade CPU's and high end GPU's or your system will triple equivalent price in the future because nobody else will be able to fill the competition gap for many years if they die off.
After saying that I think i7 will be totally awesome in the next 1-2 years as increasing multi threaded apps (more than 2-3 threads) become the norm.
Thank you for reading through all my crap =D
This kid is a typical AMD fanboy,most people realize as a fact that Deneb is not a architecuture make-over from the Phenom,it's just a step up from 65nm-45nm with some extra speed bump(but who knows how much here,dont take AMD's marketing ploys especially those of you have witnessed the promising improvement for Phenom and then they just came up with the TLB bug issues.)Even with the TLB bugs fixed lately,their top of the line 9950BE is still not comparable with the 2 yr old Kentsfield,not to mention the upper class Yorkfield.Realistically the most i can expect from Deneb would be somewhere better than Kentsfield but slower or Equivalent to(which less likely here)Yorkfield clock for clock(that's if they dont have another bug issues like TLB).
He actully knew he's talking crap so forget about his intelligence
One more post like this, and you may be getting a vacation from Anandtech. We don't tolerate personal attacks.
Markfw900 Anandtech Moderator.
One more post like this, and you may be getting a vacation from Anandtech. We don't tolerate personal attacks.
Markfw900 Anandtech Moderator.
Yup yup.
Comparing a 3Ghz Phenom to a 2.4Ghz Kentsfield(To simulate a 2.5Ghz Yorkfield vs a 2.5Ghz Deneb) is a VERY generous comparison in AMD's favor, and the AMD chip still loses about half the time. There's no way that Deneb will match Yorkfield, you can count on that.
Originally posted by: tim924
He actully knew he's talking crap so forget about his intelligence
Please don't make personal attacks. They are not warranted.
Dadofamunky
Platinum Member
- Jan 4, 2005
- 2,184
- 0
- 0
Originally posted by: Martimus
Originally posted by: tim924
He actully knew he's talking crap so forget about his intelligence
Please don't make personal attacks. They are not warranted.
+1
Markfw
Moderator Emeritus, Elite Member
- May 16, 2002
- 27,096
- 16,014
- 136
Originally posted by: Dadofamunky
Originally posted by: Martimus
Originally posted by: tim924
He actully knew he's talking crap so forget about his intelligence
Please don't make personal attacks. They are not warranted.
+1
See above, handled.
i presume you are talking about gaming here (nehalem is launching as an extreme desktop chip for enthusiast first, so you must be - or are you not?).
1. deneb can compete with i7 if you dont need 8 threads to play games.
2. deneb can compete with i7 if you dont need 3 channel bandwidth to play games.
in other words, if games dont take advantage of i7's supposedly superiority, it can actually becaome disadvantageous to it. first the cost, you are going to pay for something not useful, not to mention it being new platform. 2nd, not sure about this - you can correct me if im wrong- power, its going to still suck power even not useful.
now if you are talking about i7 being a server chip, well i dont know, based on intel roadmap, it will almost be a year from now before we see 2-way and 8-way server chips.
also, i am not sure many people here -being an enthusiast forum -will be interested in server applications, but i will be cause im interested in how amd perform from now onwards.
btw, when i7 nda lifts, are you going to bench your i7? or have you done benching and just come up with the result?
i think i agree to this. deneb will probably match kentsfield, assuming that 20pct ipc improvement is correct. now maybe when deneb shifts to am3, it will improve a few percentage points then it could be neck and neck with yorkie, assuming yorkie does not improve by then. but how much is kentsfield and how yorksfield? maybe deneb can be priced in between?
I would love it if Deneb is priced a good bit cheaper than Kentsfield. It would totally heat things up big time.
Some very reasonable assumption based on unbiased review:
Fastest Phenom can barely match the performance of slowest Kentsfield, aka Q6600. Clock to clock, it lost ugly.
Kentsfield -> Yorkfield is 65nm to 45nm shrink. Phenom to Denab is also process shrink. So my best guess is, due to it's cache size increase (3x vs 1.5x of Intel cpu's), it could pull Denab to match or even ahead of Kentsfield. It could be between Kentsfield and Yorkfield or could even close to Yorkfiled, though I highly doubt that would happen.
BUT, there is no comparison to i7.
yorkfield is barely faster than kentsfield to begin with. the main reason yorkfield is faster is cache. you can see that the smaller cache yorkfields (the q9x00) ones, are about the same as a kentsfield.
If deneb has some architectual enhancements, and SSE5 it would be probably on par with yorkfield. The fact that the motherboards will be cheaper and it will use proably less power than a nehalem and has the option of using ddr2 ram, might make it a compelling choice especially if its cheap.
I mean if they release a 2.8 or 3.0 ghz deneb, it'd be very competitive at most things with the nehalem especially given it will have 6mb L3 cache, with the current 512k L2 cache
If deneb has some architectual enhancements, and SSE5 it would be probably on par with yorkfield. The fact that the motherboards will be cheaper and it will use proably less power than a nehalem and has the option of using ddr2 ram, might make it a compelling choice especially if its cheap.
I mean if they release a 2.8 or 3.0 ghz deneb, it'd be very competitive at most things with the nehalem especially given it will have 6mb L3 cache, with the current 512k L2 cache
You just made Deneb not only = Yorkfield, but = i7. I think you are taking a couple leaps there. I would wait for more information on Deneb. We have seen leaked benchies on Deneb, and noone was blown away.
Denab would have helped bridge the gap with Core2, but with Nehalim the chasms is growing mighty big. If you can really use quadcore, it's very likely you can use HT. AMD still has the integrated graphics trump, until Larabee comes out. My money is on Bulldozer, which is rumored to have reverse HT, which sounds nice. If you have lots of money and getting some huge graphics cards in SLI/Xfire than i7 looks to be the best bet. But if you aren't, a Deneb with a 790 and integrated graphics is a very good deal for non gamers.
DrMrLordX
Lifer
- Apr 27, 2000
- 22,701
- 12,652
- 136
Perhaps so. One of the things that turned me off to Agena (and may do the same with Deneb) was the gimped memory controller clock speed. K8 processors of every generation had a memory controller that scaled upwards, clock per clock, with the clock speed of the processor, a fact which could be demonstrated by taking a K8 and raising the HTT in steady increments while testing memory latency as measured in cycles (something I did in an AM2 build thread ages ago). As long as your CPU multi and memory settings stay the same at every HTT speed, you wind up with the same memory latency across the board. Your system memory scales perfectly with the processor, which is something you would normally only expect from on-die cache. Of course, the party stops as soon as you've pushed your system memory to its limits and can no longer clock it any higher.
If you start overclocking your Agena without messing with the on-chip northbridge speed, your memory latency in cycles will increase. Only the cache latency will remain steady (and even then, maybe only the L1 and L2? Not having OCed an Agena, I can't say for sure).
Of course, AMD never really took advantage of that fact with K8; x2-3600+ procs had the same default memory speeds as an x2-6000+ or what have you. Overclockers, however, could enjoy that benefit provided they tweaked properly, however miniscule that benefit might or might not have been.
With Agena, and presumably Deneb, higher clock speeds almost certainly amount to higher memory latency, even with proper tweaking. You can only get the nb speed up so high before things get unstable (on Agena anyway). More cache will provide Deneb processors the same benefits it presumably conferred to Penryn and Conroe, i.e. the ability to scale better at higher clock speeds versus a CPU of a similar or identical architecture with less L2. Of course, comparing Intel's and AMD's different cache architectures is a bit like comparing apples to oranges but still . . .
Keysplayr
Elite Member
- Jan 16, 2003
- 21,211
- 50
- 91
Interesting.. Well, I'm a potential Deneb candidate. I'd like to pop something better into this 780a mobo. This Phenom 9550 X4 just doesn't cut it. And I can't even get this thing to 2.4GHz without it crashing. I get better scores from my 2.13GHz e6420 than the Phenom with both systems equipped with 9800GTX's and same drivers. Phenom isn't "bad", but it just needs to be soooo much better in light of it's competition. I hope Deneb delivers performance at least close or equal to the Core 2's (Conroes) at the same clock speed. At least that would be something.
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: 21K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter