Calling all Windows users who can use a DOS prompt

[grinvader]

What information do we need from him exactly?

Well, stuff (hints) that would improve your cpu detection algorithm (seems like you're having some slight issues, but nothing significant)

Issues ? What issues ? If you still think Celerons aren't properly detected I know a board where you'd fit right in.

[kevman]

Now for a rare CPU:

Code: Select all

vendor_id: AuthenticAMD
cpu family: 5
model: 13
model name: AMD-K6(tm)-III Processor
flags: fpu vme de pse tsc msr mce cx8 pge mmx syscall 3dnowext 3dnow

C:\>cpuopt
athlon-tbird

PWND.

Its a k6-2+

[kevman]

One more CPU.

Its a little rare, too...

Code: Select all

C:\>cpuinfo
vendor_id: CyrixInstead
cpu family: 5
model: 4
model name: Cyrix MediaGXtm MMXtm Enhanced
flags: fpu tsc msr cx8 cmov mmx

C:\>cpuopt
i586

Its a complete piece of shit. Almost uselessly show.

[Cyrus]

Arg I forgot I have a Pentium 1 @ 200mhz lurking around somewhere. Not to mention my mythical 12mhz colour laptop from the stoneage. I'll post the info once I find the P1 and once I find the adapter for the laptop.

[kevman]

Arg I forgot I have a Pentium 1 @ 200mhz lurking around somewhere. Not to mention my mythical 12mhz colour laptop from the stoneage. I'll post the info once I find the P1 and once I find the adapter for the laptop.

I wouldn't go out of my way. Remember, a Pentium 1 MMX was already posted.

[senserd]

vendor_id: AuthenticAMD
cpu family: 5
model: 8
model name: AMD-K6(tm) 3D processor
flags: fpu vme de pse tsc msr mce cx8 mmx syscall 3dnow

c:\test>cpuopt
k6-2

also

vendor_id: AuthenticAMD
cpu family: 6
model: 7
model name: mobile AMD Duron(tm) Processor
flags: fpu vme de pse tsc msr pae mce cx8 sep mtrr pge mca cmov pat pse36 mmx fxsr syscall mmxext 3dnowext 3dnow

c:\test>cpuopt
athlon-tbird

BTW, what's the difference between k6-2+ and k6-2?

[Nach]

Now for a rare CPU:

Code: Select all

vendor_id: AuthenticAMD
cpu family: 5
model: 13
model name: AMD-K6(tm)-III Processor
flags: fpu vme de pse tsc msr mce cx8 pge mmx syscall 3dnowext 3dnow

C:\>cpuopt
athlon-tbird

PWND.

Its a k6-2+

That is one wild processor. Interesting it has 3dnowext...
However I find that hard to believe it's a k6-2+ when the CPU has it written inside it that it is a K6-III.
Although got to fix the tbird check, thanks.

Edit:
Please redownload the test and see if it now says k6-2 or k6-3 instead of athlon-tbird.

[Aerdan]

Now for a rare CPU:

Code: Select all

vendor_id: AuthenticAMD
cpu family: 5
model: 13
model name: AMD-K6(tm)-III Processor
flags: fpu vme de pse tsc msr mce cx8 pge mmx syscall 3dnowext 3dnow

C:\>cpuopt
athlon-tbird

PWND.

Its a k6-2+

That is one wild processor. Interesting it has 3dnowext...
However I find that hard to believe it's a k6-2+ when the CPU has it written inside it K6-III.
Although got to fix the tbird check, thanks.

I was arguing with him about that last night, actually. Even dxdiag says it's a K6-III.

[Deathlike2]

I was arguing with him about that last night, actually. Even dxdiag says it's a K6-III.

The + cpus are probably the laptop versions...

Too bad though.. the K6-3 died a silent death. It didn't really take off.

[Ichinisan]

I would be curious to see the results from the CPU in my first PC. It was an Acer Aspire computer that had an "IBM P150" CPU. FPU performance was pitiful and games like Quake and Fury3 were pretty slow by comparison to an Intel Pentium 133. From what I've been able to tell, IBM-branded CPU's of that time were actually Cyrix. It's too bad I no longer have that computer

[Noxious Ninja]

I was arguing with him about that last night, actually. Even dxdiag says it's a K6-III.

The + cpus are probably the laptop versions...

Too bad though.. the K6-3 died a silent death. It didn't really take off.

Yeah. I heard that the K6-3 finally had a decent FPU.

[Deathlike2]

Actually.. it wasn't the FPU that made the K6-3 great...

The K6-2 had a name for itself because of 3DNow! (see Quake 2 w/3DNow! performance reviews)

The K6-3 had integrated L2 cache (like the P3 Coppermine CPUs)... the marketing became the "tri-level cache" system because since the K6-3 had integrated L2 cache (256kb of it), the standard onboard mobo cache (the cache was embedded) became the L3 cache (the mobo cache was L2 because it was the second level of cache at the time)... you ideally want to have the levels of cache from fastest (L1, which usually had a small amount) to slowest (L3 like in some server chip systems, you usually want the most cache here, particularly because fast cache is EXPENSIVE and takes up a lot of space on the chip)...

I forget which was the actual problem.. I think it was either production (low yields) or pathetic marketing gave it its own grave (I doubt it was cost, but I don't recall clearly)

The Athlons (K7) had great FPU power (better than Intel's at the time).. and definately one of the factors as to why the Athlons took off...

[Noxious Ninja]

Wow, I didn't know that about the integrated L2 cache. That would've really been huge.

[kevman]

Code: Select all

vendor_id: AuthenticAMD
cpu family: 5
model: 13
model name: AMD-K6(tm)-III Processor
flags: fpu vme de pse tsc msr mce cx8 pge mmx syscall 3dnowext 3dnow

C:\>cpuopt
k6-3

Close enough.

BTW, what's the difference between k6-2+ and k6-2?

The + cpus are probably the laptop versions...

They were designed for laptop, yes. However, a mobile k6-2 is not per se a K6-2+

The difference between the plus and nonplus? The plus uses a smaller (.18 um versus .25) die size. The extra space is then used to add 128kilobyte of L2 cache.

My K6-2+ runs at a VCORE of 1.8, a half volt below that of the original. It only consumes 14watts of power at 500Mhz. Aerdan will atest that it supports on-the-fly CPU multiplier changes via a little app called K6speed.
Also, I do not believe the CXT core K6-2 has 3dnowext. Mine clearly does.
The k6-III does NOT have 256k of l2; it has the same as my k6-2+.

Of course, this is a desktop. It popped right in.

Wow, I didn't know that about the integrated L2 cache. That would've really been huge.

I replaced a k6-2 550 with this k6-2+ 450. Despite the lower speed, I had a 19% increase in benchmarked performance in Gaints: Citizen Kabuto. Then I overclocked it to 500. The difference climbed to about 25%, thanks to the full-speed L2.

@ Nach and Aerdan:

It is a little insulting to question the person who BUILT and OWNS and USES the computer over such a cut-and-dry matter as what CPU is installed. It is simple, if it was a k6-III and not a k6-2+, I wouldn't be able to adjust the CPU multiplier with k6speed.

Would you like me to remove the heatsink, clean the goo off, and take a picture?

EDIT: Read, and be enlightened: http://en.wikipedia.org/wiki/K6-2#K6-2.2B_.28180_nm.29

EDIT2: I ran CPUz:

Prehaps its tagged like that to help BIOS-level compatibility.

Also, do you think a binary compiled with the athlon-tbird optimizations would run on my PC?

[Deathlike2]

The difference between the plus and nonplus? The plus uses a smaller (.18 um versus .25) die size. The extra space is then used to add 128kilobyte of L2 cache.

Oh well, that sounded about right with the amount of cache... but after seeing what the benchmarks were at the time.. it was very good (fast cache = great performance)

The CPU-Z version you used is a bit outdated (1.30 is out)

Also, do you think a binary compiled with the athlon-tbird optimizations would run on my PC?

Sure it would.. but it would run horribly.

See, it's OK to have backward compatibility.. just not forward compatibility (optimizing for a Pentium would be OK for a P3, but optimizing for an X2 would be horrible for an original Athlon.. an even better comparison would be optimizing for a P4 would be terrible for a Athlon64...)

It's hard to give an analogy to why it doesn't work...

[Aerdan]

Actually, optimizing for an arch higher than the one your CPU uses *will* cause problems, guaranteed. Mainly stability-related, but still..

[Nach]

It is a little insulting to question the person who BUILT and OWNS and USES the computer over such a cut-and-dry matter as what CPU is installed. It is simple, if it was a k6-III and not a k6-2+, I wouldn't be able to adjust the CPU multiplier with k6speed.

Would you like me to remove the heatsink, clean the goo off, and take a picture?

Yes, perhaps you should.
I'm not deciding it's a K6-3. The CPU is reporting it's a K6-3. Why would a K6-2 report it's a K6-3?

EDIT: Read, and be enlightened: http://en.wikipedia.org/wiki/K6-2#K6-2.2B_.28180_nm.29

Be enlightened about what exactly? I don't see any relavent information there.

However it might enlighten you to know that the K6-2+ is family 5 model 12 while the k6-3+ is family 5 model 13.

Also, do you think a binary compiled with the athlon-tbird optimizations would run on my PC?

As long as all the opcodes used in a build is supported by the CPU in question, it will run.
Since you have MMX, 3DNow, and 3DNow!, we don't have to worry about the special instructions. However your CPU is i586 while the K7 line is i686, so there is the possibility of some instruction such as "cmov" ending up in the binary thus giving you an illegal instruction.

Considering how well will it run when it's for the wrong CPU. First consider if it's truly the wrong CPU, if you look at the GCC optimization options, you'll see that k6-2 and k6-3 are the same thing as are athlon and athlon-tbird, and k8, opteron, athlon64, athlon-fx.

Then if you compare Athlon vs. Athlon XP for example, they are the exact same architecture, except that the XP also supports the full SSE instruction set. So using Athlon would just mean best for the CPU without using SSE.

However if you optimize for P4, it should run on a K8, as a K8 supports all the instructions the P4 (not prescott or nocona) does. But it will chose certain opcodes over others that are faster for the P4 instead of what is faster for the K8. However you will achieve better performance than a vanilla i585 build.

[sinamas]

AFAIK a k6-III has 256k of L2 cache, while a k6-2+ has 128k. A k6-2 doesn't have on-chip L2 cache. I guess this makes the k6-2+ closer to a k6-III than a k6-2 anyway, and it seems that binaries compiled with arch set to k6-2 has the same md5sum as when compiled with arch set to k6-3, on gcc 3.4. There was a k6-III+ too with 256k L2 cache.

My k6-III claims to have 256k L2 cache (and k6-III is engraved on the cpu):

Code: Select all

$ cat /proc/cpuinfo
processor       : 0
vendor_id       : AuthenticAMD
cpu family      : 5
model           : 9
model name      : AMD-K6(tm) 3D+ Processor
stepping        : 1
cpu MHz         : 401.142
cache size      : 256 KB
fdiv_bug        : no
hlt_bug         : no
f00f_bug        : no
coma_bug        : no
fpu             : yes
fpu_exception   : yes
cpuid level     : 1
wp              : yes
flags           : fpu vme de pse tsc msr mce cx8 pge mmx pni syscall 3dnow k6_mtrr
bogomips        : 792.57

Look to http://en.wikipedia.org/wiki/K6-III for more info.

[kevman]

It is a little insulting to question the person who BUILT and OWNS and USES the computer over such a cut-and-dry matter as what CPU is installed. It is simple, if it was a k6-III and not a k6-2+, I wouldn't be able to adjust the CPU multiplier with k6speed.

Would you like me to remove the heatsink, clean the goo off, and take a picture?

Yes, perhaps you should.
I'm not deciding it's a K6-3. The CPU is reporting it's a K6-3. Why would a K6-2 report it's a K6-3?

My guess? Some BIOSes might check the CPU report to help optimize itself to the correct CPU. If the BIOS is not designed to handle the plus series, the K2-6+ is closer in operation to the k6-III, so the peformance on these boards would be better. Some motherboards ignore the L2 cache completely if it doesn't think its a k6-III.
There is a massive amount of evidence I have that points out that this is a k6-2+. Most of it, though, you will have to trust me on:

1.Cpuz reports it as a "k6-2+ or k6-3+"
2.Cpuz reports that it has 128k of Level 2 Cache.
3.I can change the cpu multiplier on the fly.
4.It has 3dnowExt!
5.My BIOS bootscreen reports it has a k6-2+ (Yes, it says "K6-2+")
6.When I boot into Knoppix, it is reported as a k6-2+
7.When I boot into Knoppix, it is reported to have 128Kilobyte of l2 cache.

2,3 and 7 are IMPOSSIBLE to occur without it being a k6-2+. There is simply no other k6 series that has that amount of cache, and none besides the pluses that have PowerNow!

If you are still too stubborn to believe me, I will take a picture. But that involves ruining the silver grease that is on the top of the CPU, so I don't really want to do it.

[Deathlike2]

Hmm... then I was right about the K6-3... in any case... I don't recall a K6-2+ or a K6-3+.. but in the end, they implement the integrated L2 cache (which is a GOOD thing (tm))

I wouldn't doubt the K6-2 you have judging by the screenshots.. a K6-3 is the next closest profile to that processor anyways (since the major difference between the versions is the cache)

Though.. the 3dnowext is the same that is used in the original Athlon (new 3dnow instructions called 3dnow+ are used)... I guess this was a model when the Athlons were in and the K6-2+/3/3+ were the low end (supporting the extra extensions).. it does seem you have a rare chip though.. congrats

I don't think there's too much of a benefit if you optimized for the newer 3dnow+ extensions

[Nach]

My guess? Some BIOSes might check the CPU report to help optimize itself to the correct CPU. If the BIOS is not designed to handle the plus series, the K2-6+ is closer in operation to the k6-III, so the peformance on these boards would be better.

But why is it reporting K6-3+ instead of K6-3 then?

1.Cpuz reports it as a "k6-2+ or k6-3+"

I fail to see the evidence here.

2.Cpuz reports that it has 128k of Level 2 Cache.

Okay, AFAIK no K6-3+ came with only 128 (unless some mobile ones did that I don't know about).

But if you think the BIOS is forcing a lie about the CPU name, how do you know it's not forcing a lie about this?

To be tripply certain, I would boot Knoppix, apt-get install x86info and then run x86info -c.

3.I can change the cpu multiplier on the fly.

So? That's generally motherboard dependant.

4.It has 3dnowExt!

The K6-2+ and k6-3+ both have that, how is this a proof it's K6-2+?

5.My BIOS bootscreen reports it has a k6-2+ (Yes, it says "K6-2+")

Where does your BIOS get the info from?

6.When I boot into Knoppix, it is reported as a k6-2+

I ask the same here. For all I know this means at least K6-2+

7.When I boot into Knoppix, it is reported to have 128Kilobyte of l2 cache.

Again, is your BIOS doing something funny here or is it a mobile CPU?

2,3 and 7 are IMPOSSIBLE to occur without it being a k6-2+. There is simply no other k6 series that has that amount of cache

I would also believe that it's impossible for a K6-2+ to report it's internal name as being K6-3+ and model 13.

But if you want to go the BIOS changed it route, then I could question the other information using that same principal.

and none besides the pluses that have PowerNow!

So it's a K6-3+, that fits very well with it being model 13.

If you are still too stubborn to believe me, I will take a picture. But that involves ruining the silver grease that is on the top of the CPU, so I don't really want to do it.

I'm not trying to be stuborn here.
You have two very strong pieces of evidence direct from the CPU, "AMD-K6(tm)-III Processor" and model 13.
You say you have other evidence. I think this evidence outweighs the other evidence.

But I think no matter what your CPU really is, you definitly have something fishy going on by you.

[Deathlike2]

http://en.wikipedia.org/wiki/AMD_K6-2

The little-known K6-2+ was an enhanced K6-2 with 128 KB of integrated L2 cache and built on a 0.18 micrometre process. (Essentially, a smaller version of the K6-III+.) The K6-2+ was specifically designed as a low-power mobile CPU, and released at a time when mainstream desktop machines were fast moving on to newer platforms like the Athlon. It sold in modest numbers to its target market, and although AMD made no attempt to publicise this, it was also made available as an orthodox desktop CPU.

http://www.tomshardware.com/cpu/20000713/k6-02.html

If you look at the WhatCPU picture in this link.. it says "Model 13"... which substantiates the claim..

In any case.. some apps detect CPUs incorrectly when they are not well publicized...

Edit: New link
http://www.muc.de/~hm/linux/AMD-K6-2+.html

Someone else has a similar system.. and it gets reported as such.. so you can optimize for a K6-2 or a K6-3 (K6-3 is the probably the closest because of the integrated cache)

[kevman]

My guess? Some BIOSes might check the CPU report to help optimize itself to the correct CPU. If the BIOS is not designed to handle the plus series, the K2-6+ is closer in operation to the k6-III, so the peformance on these boards would be better.

But why is it reporting K6-3+ instead of K6-3 then?

Nothing I have ever seen has called a k6-3+. Where are you getting this?

2.Cpuz reports that it has 128k of Level 2 Cache.

Okay, AFAIK no K6-3+ came with only 128 (unless some mobile ones did that I don't know about).

But if you think the BIOS is forcing a lie about the CPU name, how do you know it's not forcing a lie about this?

To be tripply certain, I would boot Knoppix, apt-get install x86info and then run x86info -c.

That's not what I meant about the BIOS. I meant that AMD changed the Identifier So that more BIOSes would use it correctly.

As for Knoppix, I can't boot Knoppix on this machine and have Internet. I get Internet via WLAN, and my WLAN NIC is not supported by Linux.

3.I can change the cpu multiplier on the fly.

So? That's generally motherboard dependant.

It is NOT motherboard dependant. That's my point. In order for such a feature to work, the CPU has to support it, the motherboard doesn't come into play at all with this feature. Read up on k6speed and Central Tweak Unit. All they do is change an internal CPU register.

4.It has 3dnowExt!

The K6-2+ and k6-3+ both have that, how is this a proof it's K6-2+?

Up until now, the arguement has been whether its a k6-2+ or a k6-3. Noone has mentioned the k6-III+ until you did just now.

5.My BIOS bootscreen reports it has a k6-2+ (Yes, it says "K6-2+")

Where does your BIOS get the info from?

6.When I boot into Knoppix, it is reported as a k6-2+

I ask the same here. For all I know this means at least K6-2+

No idea.

7.When I boot into Knoppix, it is reported to have 128Kilobyte of l2 cache.

Again, is your BIOS doing something funny here or is it a mobile CPU?

Huh? You mean that my BIOS is shutting off half the l2 cache? Why and how the hell would it do that?

2,3 and 7 are IMPOSSIBLE to occur without it being a k6-2+. There is simply no other k6 series that has that amount of cache

I would also believe that it's impossible for a K6-2+ to report it's internal name as being K6-3+ and model 13.

My CPU doesn't ID itself as a k6-III+

and none besides the pluses that have PowerNow!

So it's a K6-3+, that fits very well with it being model 13.

If you are still too stubborn to believe me, I will take a picture. But that involves ruining the silver grease that is on the top of the CPU, so I don't really want to do it.

I'm not trying to be stuborn here.

Sure you are.

The person who purchased, installed, uses, and has ACTUALLY SEEN THE CPU IN PERSON tells you what it is, and you refuse to believe them, saying that you know exactly what AMD was thinking when the CPU was designed? I call that stubborn.

But I think no matter what your CPU really is, you definitly have something fishy going on by you.

http://www6.tomshardware.com/cpu/20000713/k6-02.html
http://www.muc.de/~hm/linux/AMD-K6-2+.html


Wow. A ton of people are having fishy goings on when it comes to the CPU.

[Nach]

My guess? Some BIOSes might check the CPU report to help optimize itself to the correct CPU. If the BIOS is not designed to handle the plus series, the K2-6+ is closer in operation to the k6-III, so the peformance on these boards would be better.

But why is it reporting K6-3+ instead of K6-3 then?

Nothing I have ever seen has called a k6-3+. Where are you getting this?

http://cvs.sourceforge.net/viewcvs.py/x ... &view=auto

I quote:

Code: Select all

	case 0x5c0:
		sprintf(cpu->name, "%s", "K6-2+ (0.18um)");
		cpu->connector = CONN_SUPER_SOCKET_7;
		break;
	case 0x5d0:
		sprintf(cpu->name, "%s", "K6-3+ (0.18um)");
		cpu->connector = CONN_SUPER_SOCKET_7;
		break;

I also reviewed many other programs I found online, and they basically had the same thing.
Being that something like 6/6 programs I've seen all have basically the same code it makes me wonder.
Of course one could have decided that's how it was, and everybody just copied that...

4.It has 3dnowExt!

The K6-2+ and k6-3+ both have that, how is this a proof it's K6-2+?

Up until now, the arguement has been whether its a k6-2+ or a k6-3. Noone has mentioned the k6-III+ until you did just now.

In all my previous posts I was using k6-2 and k6-3 to refer to k6-2+ and k6-3+ respectively. You decided to add a + onto where I said k6-2 but failed to do the same for k6-3, I decided to correct that oversight in my last post.

7.When I boot into Knoppix, it is reported to have 128Kilobyte of l2 cache.

Again, is your BIOS doing something funny here or is it a mobile CPU?

Huh? You mean that my BIOS is shutting off half the l2 cache? Why and how the hell would it do that?

Why? Perhaps it doesn't support it?

If you are still too stubborn to believe me, I will take a picture. But that involves ruining the silver grease that is on the top of the CPU, so I don't really want to do it.

I'm not trying to be stuborn here.

Sure you are.

The person who purchased, installed, uses, and has ACTUALLY SEEN THE CPU IN PERSON tells you what it is, and you refuse to believe them, saying that you know exactly what AMD was thinking when the CPU was designed? I call that stubborn.

From what you told me earlier about having to remove the gunk off it and stuff, you made it sound like you have never seen the CPU in person. I am not going to argue against the stamp on the CPU, but I am going to wonder all the various software, source code, and comparison sheets I've seen.


Although after more research on the subject it seems that the K6-2+ and the K6-3+ are the exact same CPU. They're a slightly advanced K6-III, but one has more cache than the other.

I find the whole thing bizare. But whatever, this discussion isn't really relavent. Does anyone have a WinChip?

[sinamas]

I guess some failed k6-III+ may have been released as k6-2+ with half of the cache disabled, but wouldn't they've changed the model number then?