diff --git a/aux/bochsrc.txt b/aux/bochsrc.txt
index c429ee3..f7a0a44 100644
--- a/aux/bochsrc.txt
+++ b/aux/bochsrc.txt
@@ -2,7 +2,7 @@ cpu: model=p4_prescott_celeron_336
 
 memory: guest=4096 host=2048
 
-romimage: file=/usr/share/bochs/BIOS-bochs-latest, options=fastboot
+romimage: file=/usr/share/bochs/BIOS-bochs-latest
 vgaromimage: file=/usr/share/bochs/VGABIOS-lgpl-latest.bin
 
 ata0: enabled=1
diff --git a/aux/qemu_amd64.sh b/aux/qemu_amd64.sh
index 5e627a1..7c5e62d 100755
--- a/aux/qemu_amd64.sh
+++ b/aux/qemu_amd64.sh
@@ -1,3 +1,3 @@
 #!/bin/sh
 
-qemu-system-x86_64 -M q35 -m 4G -serial mon:stdio -cdrom mop3.iso
+qemu-system-x86_64 -M q35 -m 4G -serial mon:stdio -enable-kvm -cdrom mop3.iso
diff --git a/kernel/amd64/bootmain.c b/kernel/amd64/bootmain.c
index 10121db..ab54260 100644
--- a/kernel/amd64/bootmain.c
+++ b/kernel/amd64/bootmain.c
@@ -23,13 +23,14 @@ void bootmain (void) {
   amd64_ioapic_init ();
   amd64_hpet_init ();
 
-  int* a = malloc (sizeof (int));
-  *a = 6969;
-  DEBUG ("a=%p, *a=%d\n", a, *a);
+  /* int* a = malloc (sizeof (int)); */
+  /* *a = 6969; */
+  /* DEBUG ("a=%p, *a=%d\n", a, *a); */
 
-  amd64_hpet_sleep_micro (3000000);
-
-  DEBUG ("woke up!!!\n");
+  for (size_t i = 0; i < 1000; i++) {
+    DEBUG ("i=%zu\n", i);
+    amd64_hpet_sleep_micro (1000000);
+  }
 
   for (;;)
     ;
diff --git a/kernel/amd64/hpet.c b/kernel/amd64/hpet.c
index 7295ce2..c83b971 100644
--- a/kernel/amd64/hpet.c
+++ b/kernel/amd64/hpet.c
@@ -1,6 +1,7 @@
 #include <amd64/hpet.h>
 #include <libk/std.h>
 #include <limine/requests.h>
+#include <sync/spin_lock.h>
 #include <sys/debug.h>
 #include <sys/mm.h>
 #include <uacpi/acpi.h>
@@ -15,9 +16,12 @@
 static bool hpet_32bits = 1;
 static uintptr_t hpet_paddr;
 static uint64_t hpet_clock_nano;
+static spin_lock_t hpet_lock = SPIN_LOCK_INIT;
 
 extern void amd64_spin (void);
 
+/* These functions assume hpet_lock is held by the caller! */
+
 static uint64_t amd64_hpet_read (uint32_t reg) {
   struct limine_hhdm_response* hhdm = limine_hhdm_request.response;
   uintptr_t hpet_vaddr = hpet_paddr + (uintptr_t)hhdm->offset;
@@ -36,13 +40,27 @@ static void amd64_hpet_write (uint32_t reg, uint64_t value) {
 
 static uint64_t amd64_hpet_timestamp (void) { return amd64_hpet_read (HPET_MCVR); }
 
-uint64_t amd64_hpet_current_nano (void) { return amd64_hpet_timestamp () * hpet_clock_nano; }
+uint64_t amd64_hpet_current_nano (bool lock) {
+  if (lock)
+    spin_lock (&hpet_lock);
+
+  uint64_t t = amd64_hpet_timestamp () * hpet_clock_nano;
+
+  if (lock)
+    spin_unlock (&hpet_lock);
+
+  return t;
+}
 
 void amd64_hpet_sleep_micro (uint64_t us) {
+  spin_lock (&hpet_lock);
+
   uint64_t start = amd64_hpet_timestamp ();
   uint64_t conv = us * 1000;
   while (((amd64_hpet_timestamp () - start) * hpet_clock_nano) < conv)
     __asm__ volatile ("pause" ::: "memory");
+
+  spin_unlock (&hpet_lock);
 }
 
 void amd64_hpet_init (void) {
@@ -61,7 +79,19 @@ void amd64_hpet_init (void) {
 
   hpet_32bits = (amd64_hpet_read (HPET_GCIDR) & (1 << 13)) ? 0 : 1;
 
-  amd64_hpet_write (HPET_GCR, amd64_hpet_read (HPET_GCR) | 0x01);
+  /* reset */
+  amd64_hpet_write (HPET_GCR, 0);
+  amd64_hpet_write (HPET_MCVR, 0);
+  amd64_hpet_write (HPET_GCR, 1);
 
-  hpet_clock_nano = amd64_hpet_read (HPET_GCIDR + 4) / 1000000;
+  uint64_t gcidr = amd64_hpet_read (HPET_GCIDR);
+  if (hpet_32bits) {
+    uint32_t low = (uint32_t)gcidr;
+    uint32_t high = (uint32_t)amd64_hpet_read (HPET_GCIDR + 4);
+    gcidr = (((uint64_t)high << 32) | low);
+  }
+
+  uint64_t period_fs = (gcidr >> 32);
+
+  hpet_clock_nano = period_fs / 1000000;
 }
diff --git a/kernel/amd64/hpet.h b/kernel/amd64/hpet.h
index c4aad4f..48e5920 100644
--- a/kernel/amd64/hpet.h
+++ b/kernel/amd64/hpet.h
@@ -3,7 +3,7 @@
 
 #include <libk/std.h>
 
-uint64_t amd64_hpet_current_nano (void);
+uint64_t amd64_hpet_current_nano (bool lock);
 void amd64_hpet_sleep_micro (uint64_t us);
 void amd64_hpet_init (void);