![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
The Official Prescott discussion Thread
- Thread starter USMC2Hard4U
- Start date
Vagrant Zero
Gawd
- Joined
- Jan 27, 2004
- Messages
- 898
Didn't Bill Nye get eaten by a crocidile? Ah wait...I'm thinking of the wrong guy. 
IncogNE@ro
Limp Gawd
- Joined
- Jul 18, 2002
- Messages
- 330
How come no one did the HT testing isnt prescott supposed to have to HT?
As far as the first benchies well talk about taking a step back but hey if wha anand said is tru then maybe by the time they hit 4ghz they will be alot faster then the woody( i doubt but hey)
Can wait to see what a 4ghz p4 prescott EE does 1meg L2 cache 2megs L3 cache proably alot higher fsb too now that will be a cpu.
Edit.. also wtf did anand say well after it reaches 3.6ghz and above it gets intersting show ous what you have seen you bastard dont just show us the bone, let ous atleast pick at it alittle.
As far as the first benchies well talk about taking a step back
but hey if wha anand said is tru then maybe by the time they hit 4ghz they will be alot faster then the woody( i doubt but hey)Can wait to see what a 4ghz p4 prescott EE does 1meg L2 cache 2megs L3 cache proably alot higher fsb too
now that will be a cpu.Edit.. also wtf did anand say well after it reaches 3.6ghz and above it gets intersting show ous what you have seen you bastard dont just show us the bone, let ous atleast pick at it alittle.
F
fallguy
Guest
Browsed thru the post, but didnt see any mention of it. Does the Prescott work in most current 875 chipsets? Say the IC7 MAX3? I saw a few reviews with the P4C800 Deluxe, and said it just needed a BIOS update.
edit, well it seems after reading Abits forums, they've posted a beta 14 bios for Prescott support. Dated 1/30/04.
edit, well it seems after reading Abits forums, they've posted a beta 14 bios for Prescott support. Dated 1/30/04.
the new HT extensions require the operating system to be changed to use them. so need a Wind'ohs service pack if it is possible, or maybe even have to wait for a new version o windows (which will be out after duke nukem forever)
I was reading some where else they took a PS up to 3.6.
So sure they r not a NW at overclocking they WILL BE
soon I'm sure of it.
I'm still very happy with the PS. I might get a 2.8
and take it up to 3.5+ on my water cooling.
I'm sure over the next 12 months we are going to
see great things coming from this chip.
The NW core is just about out of gas on the mhz front.
But the PS is just getting started.
I bet the steppings in 6 to 8 months will hit much higher speeds.
The kind of compter user I am HT is a MUST.
Heck a 64 will just not do it for me.
I'm the kind of user who's doing 4+ things at one time.
I'm very glad to see the PS doing just as good in this area
as the EE is.
I had a 2.6b P4 that had no HT it was ok.
But my god when I installed this C with HT
I almost sh!t on myself when I started working on it.
I can do so much more with little to no slow downs now.
I would not trade HT for all the 64's on the planet.
- Joined
- May 18, 1997
- Messages
- 55,772
I do not know.
I just finished reading your review, and I'd like to know if you have any expectations that increasing the core speed of Prescott will make up for some of the poor gaming benchmarks, like Comanche 4. Based upon the comments of everyone in this thread, this CPU sounds like it needs some finagling before it truly becomes the Northwood's successor, due to the poor gaming performance and your vague comments about your ES processor's poor OC abilities.
I guess this is the stuff your talking about. http://www.newegg.com/app/viewProductDesc.asp?description=20-144-124&depa=1. If it is its out of stock also
. You really dont think newegg is going to get any more of that pc3200 ram? Don't flame on me for posting that hear . I am pairing it with a prescott . BTW does anyone have any ideas when newegg is going to get some prescotts?
- Joined
- May 18, 1997
- Messages
- 55,772
Well, I am not sure I understand your question. But YES, faster core speeds should make the game run faster.
It did not have overall poor gaming performance compared to Northwood and I would suggest going back and looking at the graphs. For the most part the differences you saw in compared to the Northwood would never be perceptable in actual gameplay.
As for Com4, it just goes to show that pipeline changes can have negative ramifications, but I would not use it as a ruler to judge overall performance. A recompile of the game engine would most likely solve those issues.
Prescott looks to be a good part and Intel did a good job. They have very much avoided the situation they had to deal with last time so they are doing something right. Of course, as pointed out in our conclusion, the exatra cache did help mask some of the pipeline issues in benchmarks.
Originally posted by kur1j
I guess this is the stuff your talking about. http://www.newegg.com/app/viewProductDesc.asp?description=20-144-124&depa=1. If it is its out of stock also
Sorry the link i gave you was crap, this is the last BH-5s newegg has, I have bought at least 20 sticks of this stuff in the last 6 months, All had BH-5's on them.
I suggest you do not wait if you looking for BH-5 sticks, Ratings mean nothing. Same as the other companies, Just change the rating and charge more. Just remember all BH-5's need a lot of volts. As in the max 3.2v your board has and above.
http://www.newegg.com/app/ViewProductDesc.asp?description=20-144-105&catalog=147&depa=1
http://www.newegg.com/app/ViewProductDesc.asp?description=20-144-106&catalog=147&depa=1
http://www.newegg.com/app/ViewProductDesc.asp?description=20-144-123&catalog=147&depa=1
I am shooting for a gig of ram with upgrade room so i am going to get this http://www.newegg.com/app/ViewProductDesc.asp?description=20-144-106&catalog=147&depa=1
What is so wonderful about this BH-5 you talk about? Is it really good? Not trying to push my luck but what kinda avgerage FSB can this ram hit? What was the average FSB your ram hit?
chrisf6969
[H]F Junkie
- Joined
- Oct 27, 2003
- Messages
- 9,011
I dont think I'll be getting Prescott for at least 6 months+. I'm going to have to wait for new socket, dram & pci-express video cards
This could push the computer business back in to contraction b/c the upgrades are going to be too expensive/radical. SO people are going to hold on to their old P4 systems/AGP card.
My 3.6 Northwood should be equivalent to a 4Ghz Prescott (in some cases)
So I dont think I'm going to be upgrading any time soon now. I'm a little disappointed. I was hoping Prescott would just be a die shrink + cache and enable much higher clock speeds. (like Willy to Woody was)
I'm not disappointed in the performance of Prescott as its doing fine considering the pipe increase. Its just that it doesnt do me any good.
This could push the computer business back in to contraction b/c the upgrades are going to be too expensive/radical. SO people are going to hold on to their old P4 systems/AGP card.
My 3.6 Northwood should be equivalent to a 4Ghz Prescott (in some cases)
So I dont think I'm going to be upgrading any time soon now. I'm a little disappointed. I was hoping Prescott would just be a die shrink + cache and enable much higher clock speeds. (like Willy to Woody was)
I'm not disappointed in the performance of Prescott as its doing fine considering the pipe increase. Its just that it doesnt do me any good.
USMC2Hard4U
Supreme [H]ardness
- Joined
- Apr 4, 2003
- Messages
- 6,157
Since my P4 isnt the greatest OC'er.... only can get 3.18Ghz out of it at 55C idle with good cooling... I am gonna venture out and get a Prescott and OC it... get some watercooling too...
They dont seem like bad processors, espically if i can get it to 3.6+ Stable I will be a happy camper
They dont seem like bad processors, espically if i can get it to 3.6+ Stable I will be a happy camper
AthlonXP
Fully [H]
- Joined
- Oct 14, 2001
- Messages
- 20,596
fyi i just talked to pcprogress and they are all out of the prescotts so hopefully some of you all got them, also one guy said they might not ship out today but I am not gonna give that a gurantee. I guess I will wait and see then.
I think he was asking if you observed anything like what Anandtech did. That an x mhz gain Prescott has a bigger performance impact that the same gain on Northwood.
Or maybe he wasn't...
But it would be somewhat interesting, especially if we can get some clock for clock comparisons while overclocked (even modestly so).
'improved ht' is just two new instruction Monitor and MWait,
these require operating system and probably software support (not currently availibe, they will need to be patched / recompiled) to take advantage of the new instructions.
No reviewer has said support, so you couldn't draw a conclusion on any improvments from the new HT. All you would be able to see is if HT (Northwood 'style') helped Prescott make up more ground than HT (Northwood 'style') helped Northwood make up. And really with all the theoretical benches stressing raw MIPS/FLOPS already it would not be very exciting i would think.
chrisf6969
[H]F Junkie
- Joined
- Oct 27, 2003
- Messages
- 9,011
I'm wondering if that better scaling has to do with Prescott's improvements to prevent clock skewing across the chip. As you get faster it becomes more of a problem and may be why Prescott's performance continues up at a faster rate. Or that Northwood's performance starts to degrade slightly(not as linear) as it goes up b/c of the clock skew problem.
Yes, that's exactly what I was asking FrgMstr about. I could have phrased it a little more clearly but I was under the impression that he had read the Anandtech review and noted that particular section near the end.
But why hasn't anyone (including FrgMstr) really said anything conspicuous about the new HT instructions? The Anandtech review doesn't explicitly discuss HT anywhere (that I could have seen), and based upon the comments of others, the many other reviews linked to from the [H] front page don't seem to say much about it either. What precisely is the purpose of these two new instructions?
Icewindius
Gawd
- Joined
- Sep 30, 2003
- Messages
- 802
Prescott, the little CPU that could, but didn't.
Athlon 64 this summer for me.
Athlon 64 this summer for me.
I am reading the Anandtech review now, already read the [H] review...
Anand says Prescott's default voltage was 1.385v. I wonder if the 1.5v that Kyle's board defaulted to is the reason it runs warm?
Sounds like someone needs to release a BIOS update...
Anand says Prescott's default voltage was 1.385v. I wonder if the 1.5v that Kyle's board defaulted to is the reason it runs warm?
Sounds like someone needs to release a BIOS update...
chrisf6969
[H]F Junkie
- Joined
- Oct 27, 2003
- Messages
- 9,011
Also that it wont boot over 1.6v.
Quick link to the Intel PDF (which usually applies to retail steppings, until an addendum stepping arrives)
ftp://download.intel.com/design/Pentium4/datashts/30056101.pdf
2.80 A/E and 3E are 89 watts
3.20E and 3.40E are 103 watts. I guess the Inq was right on this one.
Suggested clock throttling is now 30 to 50 percent of CPU max speed.
Interestingly the VID's go from 0.8375 to 1.600Volts, which tends to make me think that Intel is hoping that the 90nm process will get them down to the sub-1volt level someday. Current prescott VID range is 1.25 to 1.40 with absolute max at 1.55Volts.
ftp://download.intel.com/design/Pentium4/datashts/30056101.pdf
2.80 A/E and 3E are 89 watts
3.20E and 3.40E are 103 watts. I guess the Inq was right on this one.
Suggested clock throttling is now 30 to 50 percent of CPU max speed.
Interestingly the VID's go from 0.8375 to 1.600Volts, which tends to make me think that Intel is hoping that the 90nm process will get them down to the sub-1volt level someday. Current prescott VID range is 1.25 to 1.40 with absolute max at 1.55Volts.
MWait (monitor Wait) tells a thread to halt execution, similar to the hlt instruction, but i can be issued seperatly to different threads. So if one thread is blocked and unable to continue execution (I/O dependancy, resource conflict, ect) an MWait instruction can be issued and the CPU will continue to execute the unblocked thread. (and presumable allow the OS to schedual another thread or the CPU will reallocate resources to the active thread).
Monitor allows for monitoring (well duh) of certain resources to decide if/when it's neccessary to issue MWait, or return from that state.
ie if an I/O address is being monitored and currently in use, a thread can 'sleep' via MWait while it waits for access. Or a thread can be woken up when another thread releases a monitored resource.
As for why it's not being discussed.
Well, it's not a major issue to begin with.
And there's really no software to test it out on.
I think there's a Linux kernel patch in the works, (you can google it, an early revision of the code is apparently online). But really that's all i know of right now. I would presume a new CPU driver for windows eventually, but who knows.
Apparently, the current version of Windows XP doesn't support it. Support will come either with a future service pack (probably not SP2), or more likely, Longhorn, the next major Windows release. So, for those waiting for improved HT and multitasking, your wait isn't over.
BaldHeadedDork
Gawd
- Joined
- Jun 18, 2003
- Messages
- 532
What I'm seeing is re-enforcing my earlier decision to stick with my 2.6C until Tejas. The boost over the Northwood C's doesn't come close to justifying a new $200-300 proc in my mind, and spending more money on this generation of chipsets and motherboard standards seems weak when DDR-2, PCI-X and PCI Express are shipping this summer and fall.
BHD
BHD
- Joined
- May 18, 1997
- Messages
- 55,772
That is one of the smartest things I have ever heard come out of a BaldHeadedDork's mouth...er keyboard. The beauty of 939 will be calling you way before then though...
I second that BaldHeadedDork and Kyle.
Always prudent to have a wait and See attitude when new hardware changes occur.
If you already own a 865/875 with a 2.4c-3ghz c wait a few more months.
Too much new tech coming out.
I Like my p4 and all but was I the only one who noticed the pwnage the Fx51 dished out. not to sound like and AMD fan boy. that is one sweet chip. hehehe
Always prudent to have a wait and See attitude when new hardware changes occur.
If you already own a 865/875 with a 2.4c-3ghz c wait a few more months.
Too much new tech coming out.
I Like my p4 and all but was I the only one who noticed the pwnage the Fx51 dished out. not to sound like and AMD fan boy. that is one sweet chip. hehehe
I don't know if this has been posted before, but Ace's review can be read here: http://www.aceshardware.com/read.jsp?id=60000315
I always read their reviews which are very insightful, but this one of Prescott seems more negative than the other half-dozen or so reviews I read to date. The article concludes with some interesting speculations about Intel's true strategy:
Either way it it is hard to believe that a tweaked Northwood could not have done better than Prescott, even it could not reach the same clockspeeds as Prescott. With the current power consumption, it looks like the power budget will limit the upper clock rate and not the number of logic levels per stage. That triggers the second question: as Intel's holy grail seems to be incredibly high clockspeeds, does Intel already know that future killer applications will not be branch intensive? Will most current AI algorithms (branch intensive) be replaced by neural networks (FPU intensive, much fewer branches)?
I always read their reviews which are very insightful, but this one of Prescott seems more negative than the other half-dozen or so reviews I read to date. The article concludes with some interesting speculations about Intel's true strategy:
Either way it it is hard to believe that a tweaked Northwood could not have done better than Prescott, even it could not reach the same clockspeeds as Prescott. With the current power consumption, it looks like the power budget will limit the upper clock rate and not the number of logic levels per stage. That triggers the second question: as Intel's holy grail seems to be incredibly high clockspeeds, does Intel already know that future killer applications will not be branch intensive? Will most current AI algorithms (branch intensive) be replaced by neural networks (FPU intensive, much fewer branches)?
all right - i've read this entire thread, as well as about 10 reviews.
my reaction? - disappointed... frankly speaking, i thought Prescott will be all that and a bag of chips, but it's pretty much a Northwood, just wearing a shorter skirt (get it? - hotter Northwood?).
i was planning to get Prescott as soon as i could, but then a lot of people said that it runs very hot and doesn't overclock well. then i was screenies from VRZone and AnandTech - those impressed me. maybe those were gems, maybe all prescotts can do ~4ghz on air. i have decided to wait until [H] gets their hands on some retail boxes and overclocks those.
my 2.8C is hardly pushing 3.3 now... - i must get something faster! if Prescotts turn out to be mediocre overclockers, i might just get a 3.2C (with hopes of hitting 3.6 - 3.7)...
my reaction? - disappointed... frankly speaking, i thought Prescott will be all that and a bag of chips, but it's pretty much a Northwood, just wearing a shorter skirt (get it? - hotter Northwood?
).i was planning to get Prescott as soon as i could, but then a lot of people said that it runs very hot and doesn't overclock well. then i was screenies from VRZone and AnandTech - those impressed me. maybe those were gems, maybe all prescotts can do ~4ghz on air. i have decided to wait until [H] gets their hands on some retail boxes and overclocks those.
my 2.8C is hardly pushing 3.3 now... - i must get something faster! if Prescotts turn out to be mediocre overclockers, i might just get a 3.2C (with hopes of hitting 3.6 - 3.7)...
well considering that I own a chip that can hit 3.6ghz on stock volt with air. I could not care less what intel releases.
unless i hear it can say double or hell even gime 25% more than what i have now.
in other words 4ghz or more. and even still, would 4ghz on a prescott really be 4ghz. I am getting tired of these bloated ghz scores.
I guess you can judge from my reaction that i am dissapointed with the prescott. I love hardware, I update just as much as most of you. I pass on this one.
My next upgrade will have a dual something cpu or dual core whatever in there with pci express. simple as that.
USMC2Hard4U
Supreme [H]ardness
- Joined
- Apr 4, 2003
- Messages
- 6,157
The Intel Pentium 5
"Power.... Performance.... Perfection"
That should be in the P5 Commercial
Agreed. I'm looking into dual 1 or 2MB Xeons of the current variety - not Nocona. The only thing I'm concerned about is PCI Express. It is not supported by any current chipset, and I want a new video card with the best performance when Doom III hits the shelves. Tough choice.
Originally posted by APOLLO
Agreed. I'm looking into dual 1 or 2MB Xeons of the current variety - not Nocona. The only thing I'm concerned about is PCI Express. It is not supported by any current chipset, and I want a new video card with the best performance when Doom III hits the shelves. Tough choice.
why not get an opteron, why faster!
leukotriene
2[H]4U
- Joined
- Sep 18, 2002
- Messages
- 2,664
Ace's is a great site for in-depth discussion and Johan makes some excellent points there. The Power Consumption numbers are truly astounding; TDP of 103W (at 91 amps!) for the 3.2 and 3.4E? Painful.
This is of course not including that Intel's spec'd TDP is not the maximum power output, either:
Backed up by Johan's observations:
Holy Crap!
If it were not for Intel's brilliant and sophisticated thermal protection mechanisms, this could get ugly real fast. In fact, it's pretty ugly even now. Supposedly the second revision will drop current draw to mid-70 amp range for the 3.2/3.4, which is much, much better, but that won't be real soon.
I sincerely hope Intel does it's second revision of Prescott to cut power output and ramp clock speed quickly so that AMD has some incentive to get fast socket 939 chips out in a timely fashion.
The point about future AI algorithms is really fascinating, but I cannot help but think that AI will not be the preeminent "killer app" for some time to come. By designing chips that excel at clock speed and stall badly with extensive branching however, Intel may be in a way encouraging the development of such apps
edit:
Well, it's clock speed will really be 4GHz, and since Prescott appears to scale better than Northwood (how much of this is due to cache I dunno, and how much is due to potential improvements in prefetch/speculative execution I also dunno), it will probably perform better than a NW@4GHz (assuming equalized bus speeds, which wouldnt happen unless you had an engineering sample NW).
GHz of course isn't a score, but a clockspeed