mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > Hardware

Reply
 
Thread Tools
Old 2005-04-15, 13:19   #1
Peter Nelson
 
Peter Nelson's Avatar
 
Oct 2004

232 Posts
Default Intel processor lineup

With recent announcement of details of 8xx series dual core processors, and availability of 6xx, I thought I would summarise processor speeds and pricing.

I would still welcome if anyone can run a Prime95 benchmark test on any 6xx or 8xx processor they get their hands on.

UK street prices are quoted from www.scan.co.uk and include VAT but high-street shops have slightly higher prices. Dollar prices are unit cost when buying 1000 chips from Intel's website. The 8xx series pricing was from an IT journalist.

I have omitted the "EE" chips as they are "extremely expensive".

CeleronD

2.26GHz Model 315
2.4GHz Model 320 £46.11 $?
2.53GHz Model 325 £49.64 $73
2.66GHz Model 330 £52.01 $79
2.8GHz Model 335 £60.25 $89
2.93GHz Model 340 £69.67 $103

3.06GHz Model 345 £79.08 $117
3.2GHz forthcoming model 350 TBA

Pentium 4 (generally 775 socket J suffix)

2.8GHz 520 £110.69 $163
3.0GHz 530 £116.68 $178
3.2GHz 540 £143.73 $218
3.4GHz 550 £181.35 $278
3.6GHz 560 £261.27 $417
3.8GHz 570 £410.52 $637

Pentium 4/64bit EM64T with 2GB cache

3.0GHz 630 £151.93 $224
3.2GHz 640 £186.03 $273
3.4GHz 650 £265.95 $401
3.6GHz 660 £402.27 $605
3.8GHz forthcoming model 670 TBA

Pentium D (dual core without hyperthreading. 2GB cache ie 1GB each core)

2.8GHz 820 £? $240
3.0GHz 830 £? $314
3.2GHz 840 £? $528

Well, I put these in a spreadsheet, with the different processors across and speeds downwards.

Thus it was possible to see the relative cost of these chips at a given clock frequency. This shows how much extra you pay for 64 bit or dualcore versions. In dollar pricing, each of these features is about equivalent to the next speed grade jump in the lower series. eg a 640 costs about the same as a 550.

Hope you find it useful having this info in one place.

It seems to me that faster processors are priced with a high additional cost which is not justified by the small increase in speed. Provided I keep the rest of the system cost low, I would be likely to buy 530/540, 630/640, 820/830 rather than the top models.

When doing performance/price comparisons with AMD chips (which is a valid thing to do when benchmarking Prime95), please compare AMD chips with these low-midrange Intel chips, as the top-of-the-range chips from either vendor are not priced in direct proportion to their speed.
Peter Nelson is offline   Reply With Quote
Old 2005-04-15, 15:18   #2
R.D. Silverman
 
R.D. Silverman's Avatar
 
Nov 2003

1D2416 Posts
Thumbs up

Quote:
Originally Posted by Peter Nelson


Pentium 4/64bit EM64T with 2GB cache

Pentium D (dual core without hyperthreading. 2GB cache ie 1GB each core)

2.8GHz 820 £? $240
3.0GHz 830 £? $314
3.2GHz 840 £? $528
Surely you don't mean GB above? A *gigabyte* cache??????? Wow!!!

What about dual core with hyper-threading?
R.D. Silverman is offline   Reply With Quote
Old 2005-04-15, 16:08   #3
TauCeti
 
TauCeti's Avatar
 
Mar 2003
Braunschweig, Germany

2×113 Posts
Default

Quote:
Originally Posted by R.D. Silverman
What about dual core with hyper-threading?
Afaik only the 840EE (Extreme Edition) offers HT at launch. It's a Pentium D 840 with HT enabled.

I wonder how the upcoming dual-core Opterons will perform at Trial-Factoring in 64 bit mode
TauCeti is offline   Reply With Quote
Old 2005-04-15, 19:33   #4
Peter Nelson
 
Peter Nelson's Avatar
 
Oct 2004

232 Posts
Default Wishful thinking!

Quote:
Originally Posted by R.D. Silverman
Surely you don't mean GB above? A *gigabyte* cache??????? Wow!!!

What about dual core with hyper-threading?
Well spotted, yes 1GB cache would be very nice, but I hold my hands up: sadly the correct units are MByte, sorry for the mistake.

The dual cores listed above will not do hyperthreading. It is only the EE chip which will boast both dual core AND hyperthreading giving a total of four virtual processors (of which two are real cores, obviously). This is only of academic interest, because they are priced very expensively. Also that EE model is based on prescott (with big pipeline) rather than Xeon architecture so already suffers a little speed hit. I only find it interesting because it points the way to future processors which will be dualcore and hyperthreading eg what will be later rolled out as some "desktrino" models, perhaps Q4 or next year.

Hyperthreading can give some increase in work done, but a real second core will give much more increase. Therefore in view of the price I could happily live without hyperthreading. On my existing Northwood P4 I have the HT feature disabled, so I'm hardly going to miss it!

I'm much more interested in the ability to support 64-bit instruction set.

For my next upgrade I am waiting for motherboards with Intel 945 and 955 chipsets (which support dualcore 8xx cpus) as they are likely to have other improved features. For example I heard that the 945G will have GMA950 integrated graphics rather than the 915G's earlier GMA900. Also I suspect they will have taken the opportunity to support recent developments like the dual speed SATA interface (300 vs 150 speed). Like 915/925 Intel will be trying to launch these with DDR2 to encourage people to switch. Many mobo manufacturers resisted this in response to consumer demand by making 915/925 mobos with conventional DDR memory slots. I believe the new chipset (maybe just the 955 variant) will support 533 ddr2 not just 400 ddr2 which means it may finally give some advantage, and I might therefore change memory type. However, the 8xx only have 800FSB whereas the EE has the faster 1066FSB.

In addition I'm also tempted on the AMD side by the Nforce4 Ultra based motherboards with an Athlon64.

I also heard that AMD were hoping to get as many as 8 cores in a chip with their Opteron series (dual to start with). They are working from high-end servers downwards whereas Intel are going dualcore aimed at conventional machines.
Peter Nelson is offline   Reply With Quote
Old 2005-04-15, 23:14   #5
moo
 
moo's Avatar
 
Jul 2004
Nowhere

11001010012 Posts
Default

Amds 64 bit is designed for duel cores they just didnt have the duel cores in the begining because it would make the chip larger and thats not good ;) no socket a support so they waited and contiuned devoloping it. from what i understand the 2 cores we reperesent them as a and b travel though the hyper transport so there connected at 1600 mhrz. the .... is for spaceing.
a---|---b
.... .|
board

Last fiddled with by moo on 2005-04-15 at 23:14
moo is offline   Reply With Quote
Old 2005-04-22, 09:03   #6
garo
 
garo's Avatar
 
Aug 2002
Termonfeckin, IE

22×691 Posts
Default

Peter,
Thanks for a very informative post. A couple of questions. Does the 8xx series have 64 bit support? And does the 6xx series support HT?
garo is offline   Reply With Quote
Old 2005-04-22, 16:41   #7
Peter Nelson
 
Peter Nelson's Avatar
 
Oct 2004

232 Posts
Default

Quote:
Originally Posted by garo
Peter,
Thanks for a very informative post. A couple of questions. Does the 8xx series have 64 bit support? And does the 6xx series support HT?
The answers are....

8xx processors DO have EM64T (Intel's new 64 bit capability)
6xx processors DO have hyperthreading but are single core, whereas the non-EE 8xx models have dualcore but no hyperthreading.

Also....
955X chipset and mobo is now available in small numbers (has been reviewed by some London mags) and it does support the faster SATA I suspected. There are now some PDF about this on the Intel website, but as yet no Intel specs on the 945 chipsets. 955X will take socket 775 processors of type 5xx, 6xx or 8xx. If you do not have 945/955 chipset (and supporting bios), an 8xx processor will just shut down on system bootup.

I cannot confirm or deny whether socket 775 Celeron D @ 533FSB would work with the 945/955 chipsets. The Intel datasheets do not explicitly specify this.

I did wondered if gigabit LAN might interface directly with the chipset, but they appear to think having PCI express lanes means they can be added this way. I suppose this is acceptable.

Last fiddled with by Peter Nelson on 2005-04-22 at 16:44
Peter Nelson is offline   Reply With Quote
Old 2005-04-26, 23:27   #8
Peter Nelson
 
Peter Nelson's Avatar
 
Oct 2004

232 Posts
Default INTEL 630 BENCHMARK

I have for some time attempted to get someone to benchmark prime95 on one of the new Intel 6xx or 8xx processors.

Today I gained access to a Compaq/HP machine with a (3GHz) 630 processor.

This identifies as type 0 family F model 4 step 3 rev 4 and as 2MB of L2 cache.

IMPORTANT NOTES when you read the results.....

It had 512MB of SINGLE CHANNEL (NOT DUAL CHANNEL) DDR RAM.

And it was running Windows XP HOME (not professional).

And I only ran ONE instance of the program (although the processor supports hyperthreading).

And although it is a 64-bit capable processor this test was done in 32 bit mode because I was not at liberty to install either the microsoft or suse linux 64 bit operating system, and there does not yet exist a linux 64 bit client from George anyway (subtle hint lol). I might have been able to boot suse 64 bit LIVE edition from DVD which would not alter the hard drive but I would still then have not client app for 64 bit mode yet.

With all that in mind, here are the results using the latest (third) version of 24.11 client.....

The speed was 3GHz but when idle, CPU-Z reported the speed management technology had dropped it to about 2.8GHz. Once busy with priming the speed is 3GHz.

[INTEL 630]

Intel(R) Pentium(R) 4 CPU 3.00GHz
CPU speed: 3000.78 MHz
CPU features: RDTSC, CMOV, Prefetch, MMX, SSE, SSE2
L1 cache size: 16 KB
L2 cache size: 2048 KB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
TLBS: 64
Prime95 version 24.11, RdtscTiming=1
Best time for 512K FFT length: 18.358 ms.
Best time for 640K FFT length: 21.796 ms.
Best time for 768K FFT length: 26.365 ms.
Best time for 896K FFT length: 31.650 ms.
Best time for 1024K FFT length: 35.298 ms.
Best time for 1280K FFT length: 46.438 ms.
Best time for 1536K FFT length: 55.953 ms.
Best time for 1792K FFT length: 67.795 ms.
Best time for 2048K FFT length: 75.294 ms.
Best time for 58 bit trial factors: 9.258 ms.
Best time for 59 bit trial factors: 9.919 ms.
Best time for 60 bit trial factors: 9.899 ms.
Best time for 61 bit trial factors: 9.947 ms.
Best time for 62 bit trial factors: 13.751 ms.
Best time for 63 bit trial factors: 13.761 ms.
Best time for 64 bit trial factors: 15.693 ms.
Best time for 65 bit trial factors: 15.589 ms.
Best time for 66 bit trial factors: 15.661 ms.
Best time for 67 bit trial factors: 15.525 ms.

AS COMPARED TO MY

[3GHZ NORTHWOOD]

which does... (on dual channel memory, XP Pro, smaller cache, obviously)

Intel(R) Pentium(R) 4 CPU 3.00GHz
CPU speed: 2992.77 MHz
CPU features: RDTSC, CMOV, Prefetch, MMX, SSE, SSE2
L1 cache size: 8 KB
L2 cache size: 512 KB
L1 cache line size: 64 bytes
L2 cache line size: 128 bytes
TLBS: 64
Prime95 version 24.11, RdtscTiming=1
Best time for 512K FFT length: 16.609 ms.
Best time for 640K FFT length: 19.976 ms.
Best time for 768K FFT length: 24.745 ms.
Best time for 896K FFT length: 29.202 ms.
Best time for 1024K FFT length: 32.546 ms.
Best time for 1280K FFT length: 42.480 ms.
Best time for 1536K FFT length: 52.942 ms.
Best time for 1792K FFT length: 62.410 ms.
Best time for 2048K FFT length: 70.334 ms.
Best time for 58 bit trial factors: 9.726 ms.
Best time for 59 bit trial factors: 9.796 ms.
Best time for 60 bit trial factors: 9.838 ms.
Best time for 61 bit trial factors: 9.770 ms.
Best time for 62 bit trial factors: 10.721 ms.
Best time for 63 bit trial factors: 10.707 ms.
Best time for 64 bit trial factors: 13.081 ms.
Best time for 65 bit trial factors: 12.981 ms.
Best time for 66 bit trial factors: 12.995 ms.
Best time for 67 bit trial factors: 12.969 ms.

There may also have been some differences in the motherboard specs between machines, but the Northwood appears a little faster than the 630 in this test. As 630 is based on Prescott it may be closer to those figures in performance.

The extra L2 cache does not appear to make a great difference, possibly because George has optimised the code to keep everything in the cache available on normal machines. Obviously, changing to dual channel memory might help speed up the 630 too.

In future 630 does offer the opportunity to migrate to 64-bit operation. If this gave a similar acceleration factor as on Athlon64 in 64 vs 32bit mode then it could be attractive. Unfortunately I was not in a position to install a different OS and 64 bit client.

Availability update: Here in north east of England, 630 processor is now available locally in SOME retail outlets (either standalone or in a whole PC system). Local street price of the processor alone is about £15 on top of the Scan pricing above (but I would not have to pay shipping charges because I can pick it up in the store).

I will look out for opportunities to benchmark an 8xx processor.
Peter Nelson is offline   Reply With Quote
Old 2005-06-06, 23:20   #9
cheesehead
 
cheesehead's Avatar
 
"Richard B. Woods"
Aug 2002
Wisconsin USA

769210 Posts
Default

Have a look at what moo found on Tom's Hardware Guide about a Pentium M desktop:

http://www.mersenneforum.org/showthread.php?t=4167

Maybe a clue to future Intel lineup?


Richard "cheesehead" Woods (on vacation)

Last fiddled with by cheesehead on 2005-06-06 at 23:27
cheesehead is offline   Reply With Quote
Old 2005-06-07, 02:47   #10
moo
 
moo's Avatar
 
Jul 2004
Nowhere

80910 Posts
Default

Hmm someone found it yay... Yes but talking with primecruncher he says they have a reletively low fp. We need results people post them in benchmark threads.

Last fiddled with by moo on 2005-06-07 at 02:47
moo is offline   Reply With Quote
Old 2005-07-03, 05:58   #11
Peter Nelson
 
Peter Nelson's Avatar
 
Oct 2004

232 Posts
Default

Quote:
Originally Posted by cheesehead
Maybe a clue to future Intel lineup?

Richard "cheesehead" Woods (on vacation)

Hehehe I'm in the official Intel reseller programme.

Therefore I do have some clue as to future Intel lineup.

Of course stuff is under NDA so I can't tell you.

(Or if I did I'd have to kill you)

Anyway you probably already noticed the strong emphasis on EM64T.

You may have noticed that the new 5xx WITH EM64T is officially now made EXACTLY the same price as the older 5xx models without 64 bit capability.

You may also have noticed the very recent announcement of some versions of the Celeron D WITH EM64T too.

I think people would be very foolish to invest in any hardware which is not 64-bit capable, regardless of what timescale they will or intend to upgrade their OS in.

Things are rapidly heading in the socket 775 direction, with PCI-express and DDR2 as opposed to AGP and DDR. Expect to see the volumes of socket 478 processors reduce quickly for pentium and also later this year for Celeron.

Oh, and I don't think it is giving anything away to say look out for price changes during the year (these tend often to be downwards, obviously).

From reviews you may have noticed the 8xx are quite HOT.

Intel appear to see BTX formfactor/cooling as the solution and have quite a few BTX products planned, (although ATX versions will remain) including a pico-BTX board.

Also, as for availability, we can finally get hold of 8xx chips and 955 motherboards.



Of course the rival AMD X2 dualcore chips pose good competition, albeit at quite a high price. These are also just becoming available through distributor channels.

Last fiddled with by Peter Nelson on 2005-07-03 at 06:02
Peter Nelson is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
The Secret CPU Inside Your Intel Processor ewmayer Tales From the Crypt(o) 21 2017-11-23 03:02
Intel Processor Speculations Mark Rose Hardware 109 2017-10-13 16:55
64 bit intel processor? Unregistered Hardware 2 2006-08-30 22:21
Intel Core Duo processor drew Hardware 5 2006-05-29 07:00
Which type of Intel processor to choose? Mike Hardware 11 2004-12-21 04:10

All times are UTC. The time now is 16:29.


Wed Oct 20 16:29:20 UTC 2021 up 89 days, 10:58, 0 users, load averages: 1.34, 1.27, 1.40

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.