Sign in with
Sign up | Sign in
Your question

Price/Value, AMD vs. Intel, including Theoretical Thuban 1055T

Last response: in CPUs
Share
March 19, 2010 3:45:52 AM

Recently I had a brief discussion with some of my friends about the best CPU to start a build with. Generally there are two ways to think about "the best CPU." The first way is the "budget way." Starting with a total system budget, you find your CPU budget and then pick the best CPU within that budget. This is commonly seen in the System Builder's Marathon. Alternatively, one can find the best performance value chip on the market, then build a system around it. With the first method, you're best off just reviewing the last "Best Gaming CPU for the Money." The latter method required a bit more analysis.

First I have to define a few concepts. The first is that "performance" in the abstract is meaningless. Performance to a graphic designer is different from performance to a folding@home team is different from performance to a gamer. Because I don't have the time or the wherewithal to build my own 'performance' index, I used PassMark from CPU Benchmarks. This will probably invite methodological criticism, what can I say. Also, value is a nebulous concept because prices vary pretty dramatically between vendors. For this study, I used MicroCenter, because it's prices are consistently the lowest (Sorry newegg). Using a $289 price for the i7-920 would have skewed results pretty differently.

The first piece was pretty straightforward, I took the most of the popularly cited gaming CPU's from Tom's Hardware and found their PassMark/Dollar Ratio. Unfortunately, the AMD flagship, the Phenom II X4 965 BE is only $180, meaning that the $180+ sections of the graphs are exclusively Intel. It's not surprising considering that Intel is about a generation or two ahead of AMD technology-wise (just look at process node and energy efficiency).

Looking at all the chips overclocked would have been interesting, and may have shifted the price value a bit (especially clarkdales), but I intended this to be useful for non-overclockers as well. I did include overclocking data on two chips, the i7-920 and the X2 555 BE. Why those two chips? Given that the i7-920 with air cooling can match the performance of chips costing 4 times as much, I felt it was worth including. The Stock Phenom II X2 555 BE is quite frankly a very bad value. At the $100 price point, it is trounced by every other chip I looked at (i3-530 and X4 630). However, as noted in many other places, many if not most 555's can be unlocked to a Quad Core running at a stock rate of 3.2 GHz. I used a 955 as its proxy. An unlocked X2 555 BE is the best value in current CPU's, period, end of story.

To make the results more interesting, I added a "theoretical Thuban", the Phenom II X6 1055T. Without any data other than the clock rate (2.8 GHz), I took the results for the Phenom II X4 925 @ 2.8 GHz and simply scaled it by 50% (6 cores vs. 4 cores)... of course this is very rough, many of the benchmarks probably won't scale, but it is ballpark. The first takeaway, is that at 5100 PassMarks, the Thuban isn't a beast relative to Intel. With 8 logical cores, and an identical clockrate, an Intel i7-930 will walk all over it. Given it's likely performance is between that of the i5-750 and i7-920, AMD does not have a lot of leeway with respect to pricing. I threw out three prices, a low of $190 (placing it above the current flagship), a medium of $220 (above the i7-920), and a high of $260 (gouging people stuck on AM3?).

Here is the first chart... exceptionally good chips were highlighted:


Baring some unexpectedly high performance from the Thuban, to be competitive against intel, it will have to be a sub-$200 chip. Hard to believe a factory Hexacore being introduced for <$200, but above $200 it's simply eclipsed by stronger Intel chips.


The big takeaway here is that the Phenom II X4's are a bad deal considering how close the performance is not far from <$100 chips and price is close to the i7-920



Even with Thuban, AMD won't reclaim leadership.


This piece is not intended to be the final word on price value. For many builders other factors come into play, power consumption, upgradability, motherboard costs, brand affinity. Truth be told, I'm an AMD guy, have been since my first build (an Athlon 850 Mhz in 1999). For a budget build, AMD still dominates, but once you commit to spending more than $150 on a CPU, Skip everything and go directly to the i7-920 (and stop). Historically, the $300+ cost for X58 mobo's strongly weighed in against the i7-920, but now motherboard costs are quite reasonable. Regarding upgrade pathway, the sub-$100 AM3 chips will have a pretty handsome upgrade to a ~3.8 Ghz hexacore chip in 2011. The i7-920 owner will be able to step up to a more reasonably priced gulftown in late 2011.

R Brent Mattis
a b à CPUs
March 19, 2010 5:18:02 AM

Interesting/fresh way of looking at things. I concur that if one wants to spend more than $150 on a chip, it is probably easiest to go with say, an i7-920, etc... and that MOBO costs and DDR3 costs have leveled out. Yet, for the budget builder and those who feel that AMD gives them the best performance for what they use a computer for, they are the cheapest/easiest way to go.
a c 83 à CPUs
March 19, 2010 5:37:12 AM

This is interesting, but your price/performace is skewed if you ask me. Using microcenter store only pricing on Intel chips sold below retail value doesn't skew the results? However if you used actualy online etailer pricing that 90%+ of us have to follow as we don't have access to microcenter will skew price/performance results?
Related resources
March 19, 2010 6:17:34 AM

Not a bad read is all i can say... There rest is about choice and desire if you ask me...
a b à CPUs
March 19, 2010 8:09:24 AM

It is nice to see a first post that is not a question which has been answered over 100 times I must say.

Now we just need to wait for the fanboys to arrive and the party can get started. *Grabs popcorn*
March 19, 2010 11:55:54 AM

AMD prices don't change radically at NewEgg. Looking at the two that matter, the i5-750 and the i7-920, the i5 goes to $200 and 21.07 Passmarks/Dollar. The $289 i7-920 drops to 19.34 Passmarks/Dollar. At $289 dollar the i7-920 is a worse performance value than the i5 and the Phenom II X4's. However, the difference is so marginal, one might as well go to the i7-920 and buy a few more months until obsolescence. Another thing weighing in towards the i7-920 is that it's a monster overclocker, so down the line when it's no longer competitive, with some straightforward tweaking one can move back up to a respectable performance level.
a b à CPUs
March 19, 2010 4:50:18 PM

Hmm, I didn't think prices were out on Thubans yet. Looks pretty inexpensive if what the OP lists is true. Last I read (Semi-accurate), prices were >$300.
March 19, 2010 6:09:40 PM

Prices haven't been published on Thuban yet. AMD might indeed try to price the Thuban at $300, but why anyone considering a ground up build would pay $300 is beyond me. Along almost any dimension Bloomfield or Lynnfield is a better choice. Plus, by the time the hotter 3.2 and 3.4 GHz Thuban's hit shelves, Intel's quad-core westmere's should be on the market. Enthusiasts will probably have to wait til Orochi to get a competitive AMD high-end.

Looking on the bright side for AMD, until Intel cuts prices on it's 32nm dual-cores, AMD strictly dominates the low-end of the desktop market.



March 19, 2010 6:52:38 PM

A good read, but your prices seem a little off. I could just be out of the loop on them.
a b à CPUs
March 19, 2010 8:45:59 PM

rbmattis said:
Prices haven't been published on Thuban yet. AMD might indeed try to price the Thuban at $300, but why anyone considering a ground up build would pay $300 is beyond me. Along almost any dimension Bloomfield or Lynnfield is a better choice. Plus, by the time the hotter 3.2 and 3.4 GHz Thuban's hit shelves, Intel's quad-core westmere's should be on the market. Enthusiasts will probably have to wait til Orochi to get a competitive AMD high-end.

Looking on the bright side for AMD, until Intel cuts prices on it's 32nm dual-cores, AMD strictly dominates the low-end of the desktop market.


Supposedly Intel's i7-970 (the slower, locked multiplier version of the 980) will be out next quarter (April - June) and will range from $500 - $800 in price, so there would be room for a 6-core between $300 & $450.

And AMD will need to become profitable again, which it has pretty much not done in the last 3-1/2 years. 6 cores on 45nm has got to use up a lot of silicon real estate, so more expensive to fab. So I would not be surprised to see AMD charging >$300 for Thuban, at least initially, in an effort to make some money.

I also have not seen any quad Westmere's on Intel's roadmaps yet. I'm going to be building an LGA-1366 rig soon and would be interested in either the 970 or a quad Westmere, depending on price of course..
March 19, 2010 8:51:43 PM

I agree AMD needs to make money, however charging >$300 for Thuban doesn't make sense unless Thuban can at least match the $289 i7-920 in along some dimension.

quad westmere
March 21, 2010 6:09:37 PM

Firstly the Thubans are related to Opterons, a related comparison between popular desktops vs Istanbul’s can be seen here http://www.xbitlabs.com/articles/cpu/display/amd-istanb... .
It makes more sense to try to project an Istanbul as the Thuban platform (Leo) than the quads. Be aware that the comparisons in the above article pit a server processor against desktop cpu's so it's not apples to apples but it gives a clear understanding that AMD has tweaked the Thubans for desktop performance, hence the C state performance boost, which again points to tweaking. If the 2 extra cores can manage a 33% increase alone that will place the 2.4GHz Thubans around a i7 975, when you start to add the possible gains with DDR3 mem, Turbo boost, 890GX chipset, the percentages start to add up nicely at the 2.4ghz end. Now add in that desktops(workstations) don’t need HT assist like the servers, or the 3 HyperTransport busses, and that again suggests modifications, which will probably include ‘Fusion’ with their GPUs (its all about graphic professionals). As for pricing I would guess at $250 $450 $600 $1000.
asH
March 22, 2010 4:42:21 AM

You clearly know a lot about CPU technology but I'm a bit confused. The article you linked to shows the Hexacore Opterons getting trounced by modern desktop quads. I read the article, and it seems like the server chips have several "features" which make them less suited for desktop apps. Is it really realistic to expect i7-975 level performance from the thuban?
a b à CPUs
March 22, 2010 6:05:13 AM

I'd expect Thuban to be $300-350. Even then it's going to have zero competition for up to 300% of its price range (as far as 6 core computing and its perks goes).

I see this thing selling for certain uses. Some of those multi-core optimized apps are going to love it. Even with i7's HT I'm sure Thuban will win in some apps.

i3 530 vs Ph2 720 might give you a peak. (of course core scaling won't be the same 4vs6 on most of these apps, and especially the games)
http://www.anandtech.com/bench/default.aspx?p=83&p2=118...

PS: Where are you getting your prices from? They do seem a bit low. (i7 $200? AII 620 $82?)
March 22, 2010 11:48:21 AM

rbmattis said:
You clearly know a lot about CPU technology but I'm a bit confused. The article you linked to shows the Hexacore Opterons getting trounced by modern desktop quads. I read the article, and it seems like the server chips have several "features" which make them less suited for desktop apps. Is it really realistic to expect i7-975 level performance from the thuban?


I'm ready. As stated the comparison is apples to oranges, consider it an untweeked desktop. Trounced is a relative term, the chip was handicapped, 2.4 ghz, and server tuned. Question is how can AMD get at least 70% (my success rate) more performance from Istanbul(for desktop) ? How can AMD call this an 'High End' platform? To answer your question I say yes.If you consider/understand the significants they made over past chips with ie TWRK, Mangy from the success of istanbul, the choice of istanbul for Thuban, makes me want to dig a bit deeper.
March 22, 2010 4:06:07 PM

Raidur said:
PS: Where are you getting your prices from? They do seem a bit low. (i7 $200? AII 620 $82?)


Quote:
For this study, I used MicroCenter, because it's prices are consistently the lowest (Sorry newegg). Using a $289 price for the i7-920 would have skewed results pretty differently.
a b à CPUs
March 22, 2010 4:18:12 PM

^ Actually according to your article it was 2.6GHz.

I would be really astonished to see AMD tweak "70 % more performance" out of the DT version compared to the server version.

Wasn't "TWKR" the super-leaky oc specials from AMD? They were great for overclocking but IIRC AMD didn't want to sell them because of their huge power requirements. How this would help with Thuban I dunno - it looks like AMD is doing all they can to keep Thuban's 6 cores in the same power envelope as their quads.
a b à CPUs
March 22, 2010 4:19:10 PM

Ic. Most of us aren't as lucky to have a MC near by. :) 
March 22, 2010 4:36:11 PM

If u could add the ways of buying CPU to apply to different genres of gaming and overall gaming would be awesome. For example You can make a RTS genre, FPS genre, and a RPG genre.

great read.
March 22, 2010 4:54:49 PM

Raidur said:
Ic. Most of us aren't as lucky to have a MC near by. :) 



As I said before:
Quote:
AMD prices don't change radically at NewEgg. Looking at the two that matter, the i5-750 and the i7-920, the i5 goes to $200 and 21.07 Passmarks/Dollar. The $289 i7-920 drops to 19.34 Passmarks/Dollar. At $289 dollar the i7-920 is a worse performance value than the i5 and the Phenom II X4's. However, the difference is so marginal, one might as well go to the i7-920 and buy a few more months until obsolescence. Another thing weighing in towards the i7-920 is that it's a monster overclocker, so down the line when it's no longer competitive, with some straightforward tweaking one can move back up to a respectable performance level.



a c 131 à CPUs
March 22, 2010 5:20:54 PM

Passmark seems to be able to use the multiple cores well but doesn't seem to be able to utilize the cache very well. The difference between the phenom IIx4 910 at 2.6GHz and the Athlon IIx4 620 at 2.6GHz is only 74.

All 2.6GHz models:

Athlon: 3123
Phenom II: 3197

And just for kicks the old 9950
Phenom: 3015
March 22, 2010 8:13:14 PM

Interesting... apparently prices were leaked: http://www.techreport.com/discussions.x/18648

I was pretty close. The 2.8 GHz hexacore will be $200 and the 3.2 GHz black edition will be $300. With such aggressive pricing, I'll have to wait to see the benchmarks vs. the i7 series.

*edited to say*

Scratch that... with Microcenter blowing out the i7-920's for $180 and i7-930's for $200... sorry AMD, I'm joining the dark side.



March 23, 2010 4:48:57 AM

Wish List

$...............i7.............AMD >W...........GHz.........X6............PassMark.........$
..........................................95...........2.6...........1035...........5384
$279.00 -920 i (130W).....95...........2.8...........1055...........5589........ $199.00
$379.00............................125..........3...........1075.............5885
$587.00- 960 i (130 W).....125.........3.2...........1095............6430........ $295.00
$969.00- 975 i (130 W).....125.........3.4..............................7201
a b à CPUs
March 23, 2010 5:16:57 AM

rbmattis said:
Interesting... apparently prices were leaked: http://www.techreport.com/discussions.x/18648

I was pretty close. The 2.8 GHz hexacore will be $200 and the 3.2 GHz black edition will be $300. With such aggressive pricing, I'll have to wait to see the benchmarks vs. the i7 series.

*edited to say*

Scratch that... with Microcenter blowing out the i7-920's for $180 and i7-930's for $200... sorry AMD, I'm joining the dark side.


Why? I would wait until Thuban benchmarks are out before making such a hasty decision - especially considering it can drop right into AM3 boards without having to buy triple channel + new mobo just for an i7 with 4 real cores.
a b à CPUs
March 23, 2010 6:41:48 PM

werxen said:
Why? I would wait until Thuban benchmarks are out before making such a hasty decision - especially considering it can drop right into AM3 boards without having to buy triple channel + new mobo just for an i7 with 4 real cores.


and AM2+ if the board vendor updates the BIOS (i am hoping ASUS)
a b à CPUs
March 23, 2010 8:39:50 PM

ash9 said:
you would want a new board to take advantage of HT3


dont know how sound it is but Wikipedia listsTurbo speeds and multipliers
http://en.wikipedia.org/wiki/List_of_future_AMD_micropr...


it would be fine since HT doesn't really affect performance that much

also, my board is the M4A78-E (its the DDR2 variant of the M4A78T-E), so i actually have a newer HT system (also PII X4 810)
March 24, 2010 9:30:04 AM

that would work
March 24, 2010 9:36:37 AM

http://www.tomshardware.com/reviews/usb-3.0-sata-6gb,25...

AMD vs. Intel?

For some reason, all Intel chipsets available today only support PCI Express 2.0 on the primary links that are used for graphics. This applies to the 4- and 5-series chipsets employing the ICH10 southbridge. While Intel claims PCI Express 2.0 support on those links, it limits throughput to PCI Express 1.1-class performance. This is naturally a problem when we start looking at the latest high-speed motherboard-down controllers.

AMD, on the other hand, upgraded the link speeds on its 700- and 800-series chipsets, which means that current AMD mainstream and enthusiast chipsets don’t create bandwidth bottlenecks for high-speed add-on devices.

We took three P55 motherboards from Gigabyte and MSI that all come with different solutions to offer USB 3.0 and SATA 6Gb/s connectivity. We analyzed SATA performance using Crucial’s new RealSSD-C300 and a Seagate Barracuda XT with support for the third-gen standard and found that not all solutions deliver ample bandwidth.
March 24, 2010 9:47:28 AM


http://www.tomshardware.com/reviews/usb-3.0-sata-6gb,25...

Check this out, from the same article above- 1st line : From who??? No one on the web can make a direct statement against Intel without trepidation.


Issues

Existing mainstream chipsets from do not provide sufficient PCI Express bandwidth for USB 3.0 or SATA 6Gb/s controllers because, while PCH-based PCI Express lanes supposedly offer a second-gen interface, they run at first-gen transfer rates (250 MB/s instead of 500 MB/s). Motherboard manufacturers can work around this by routing add-on components through PCIe switching logic or by physically wiring these controllers to PCI Express 2.0 lanes, which typically drive your graphics cards. AMD chipsets (starting with the 700-series) are fully PCI Express 2.0-compliant and consequently don’t exhibit such a limitation

asH
!